kevin/deepseek-harness
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge latest native Windows CI base into coverage follow-up

Browse Source
This commit is contained in:
Tianyi Cui
2026-08-08 20:44:36 +08:00
parent f96be1db13 0ad10e0f9d
commit ded7f8ffd1
86 changed files with 3289 additions and 232 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md
2026-07-22-durable-subagent-catalog-and-list-agents.md: b96d6e1dd36c58af67c8e93e62515672790ad009
2026-07-22-durable-subagent-catalog-and-list-agents.zh.md: 856bac615db84bfe2898ec0838094c6bc29f77b2
2026-07-22-durable-subagent-catalog-and-list-agents.md: 8337a926238bf7fc4395896fcd4ca180c9c1ac1c
2026-07-22-durable-subagent-catalog-and-list-agents.zh.md: bcf2895a9a69c8cff949788c78158bfccd198c5c
.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md
+4 -4
View File
@@ -23,7 +23,7 @@ Parent-to-child enumeration is a service capability with consumer-specific proje
- report corpus activity separately as `running` or `inactive`, without implying completion or resumability;
- return every resulting child in stable `createdAt` ascending, child-id ascending order.
Every ordinary local start receives a `one-shot` descriptor with an optional caller-owned display label, while the continuation manager persists a labeled `continuable` descriptor containing its additional reconstruction fields. The model-facing delegation tool already owns a short `description` and supplies it for one-shot display; lower-level callers such as workflows need not invent presentation metadata. The model-facing `list_agents` adapter filters the service result to continuable children and maps `inactive` to its existing `complete` presentation; a UI can consume both modes and choose an id-based fallback for unlabeled one-shot history. Descriptor persistence, by-id lookup, direct-parent authorization, and provider-independent cold resume remain owned by the implemented Activation contract. Listing consumes those facts but cannot weaken them or invent a second descriptor representation.
Every ordinary local start receives a `one-shot` descriptor with an optional caller-owned display label, while the continuation manager persists a labeled `continuable` descriptor containing its additional reconstruction fields. The model-facing delegation tool already owns a short `description` and supplies it for one-shot display; lower-level callers such as workflows need not invent presentation metadata. The model-facing `list_agents` adapter filters the service result to continuable children and refines status through the live Agent registry (`running`/`idle`/`complete`); a UI can consume both modes and choose an id-based fallback for unlabeled one-shot history. Descriptor persistence, by-id lookup, direct-parent authorization, and provider-independent cold resume remain owned by the implemented Activation contract. Listing consumes those facts but cannot weaken them or invent a second descriptor representation.
### Enumeration decision
@@ -52,7 +52,7 @@ If measured scale later requires an index, that index is derived state: session
A valid descriptor produces one child entry, a per-child inspection failure produces one diagnostic entry, and a candidate without a descriptor produces no entry. `mode` is durable creation policy; `activity` is a process-local corpus snapshot. Activity is neither `AgentStatus`, the manager's internal Activation state, nor a durable outcome, and the result does not expose the internal `createdAt` sorting key. Exact Activation states and durable outcomes such as successful completion, failure, cancellation, and stop reason require a separate durable activation record and are outside this feature.
The model-facing `list_agents` tool takes no arguments, derives `parentSessionId` from the current execution Agent, and is a thin adapter in `@deepseek-ai/dsh-tool-subagent-control`. It keeps diagnostics, drops `one-shot` child entries, maps a continuable child's `running` activity to `running` and `inactive` activity to `complete`, then renders `<id> [<status>] — <label>` or `<id> [diagnostic: <reason>]` in the surviving trace order. An empty projection renders `(no subagents)`.
The model-facing `list_agents` tool takes one optional `scope: 'children' | 'descendants'` argument, derives the root id from the current execution Agent, and is a thin adapter in `@deepseek-ai/dsh-tool-subagent-control`. It keeps diagnostics, drops `one-shot` child entries, derives status from the live Agent registry — `running` for an active driver, `idle` for a resident Agent between turns, and `complete` when no live Agent remains — then renders `<id> [<status>] — <label>` or `<id> [diagnostic: <reason>]` in stable catalog order. The `descendants` scope reads `SubagentService.listDescendants(rootSessionId)`, which flattens the complete tree from one live-preferred corpus in stable pre-order, traverses ordinary and one-shot intermediates so deeper continuable agents are discovered, revalidates each cold candidate against its enumerated lifecycle, and adds `parentId`/`depth` to every entry. The tool inserts ` parent=<id> depth=<n>` before the label; `parent` is the durable direct-parent session id and may name an omitted ordinary session. For the current caller, only depth-1 child rows are `send_message` candidates, while deeper child rows may be selected for `interrupt_agent` ([interrupt contract](2026-08-06-continuable-subagent-interrupt.md)). Discovery is a hint only — follow-up authority stays exact-direct-parent, and interrupt authority stays with the service's live-lineage check. An empty projection renders `(no subagents)`.
Diagnostics use three fixed reasons. Malformed event surfaces, conflicting headers discovered during an exact child load, a read result whose immutable header differs from the traced candidate or no longer names the requested direct parent, a target that is no longer the located descriptor event, malformed descriptor content, and multiple descriptor events map to `corrupt`. An unknown descriptor version maps to `unsupported`. `SESSION_QUERY_SESSION_NOT_FOUND`, `SESSION_QUERY_EVENT_NOT_FOUND`, and `SESSION_QUERY_PERSISTENCE_FAILED` from a per-child read map to `unavailable`. This phase boundary is intentional: a persistence outage during the initial trace fails the operation, while the same outage beginning during candidate reads may produce one identical `unavailable` diagnostic per affected child; the first version neither coalesces those diagnostics nor promotes them to a global failure. A missing descriptor is instead a non-subagent exclusion without a diagnostic. Configuration/window errors and unrecognized failures are not child diagnostics and propagate as operation failures. Each diagnostic identifies the child id and reason without exposing model-hidden descriptor content; the candidate is omitted while healthy siblings remain visible. Sessions outside the trace's direct descendants are never read and produce no diagnostic.
@@ -93,8 +93,8 @@ The first version has no child deletion operation. If later product behavior del
## Testing
- `packages/subagent/subagent/tests/service.spec.ts` pins descriptor v2 parsing for both modes and proves an unlabeled raw start resolves a one-shot descriptor before provider dispatch. `packages/subagent/subagent-inprocess/tests/subagent-inprocess.spec.ts` proves the local driver appends that descriptor inside the initial turn, returns the published id when cancellation lands in the factory-to-run handoff, and keeps result and handle-disposal failures on separate channels. Delegation-tool tests pin propagation of their existing display description and preserve independent result and disposal diagnostics.
- `packages/subagent/subagent/tests/list-children.spec.ts` pins the current read path against a real composition of the session store, JSONL persistence, spawn/fork providers, the subagent service, and the projection registry — no query service — keylessly: live-only listing without persistence; loud `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` and `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` even with zero children; the three-rung ladder (a live child never inspected, a cold child inspected exactly once, and the cache-hit, absent-key, absent-service, and poisoned-row second-rung cases); last-wins over multiple descriptors; malformed payloads and unknown versions diagnosed as `corrupt`; a failed cold inspection as one `unavailable` diagnostic retried on the next listing; a fork seed's ancestor descriptor listed under that identity; foreign-unit fold failures contained per child as `corrupt` on both the live and cold paths; `createdAt`-then-id ordering without ordinary forks; provider absence without child omission; compacted/uncompacted twins listing identically; a persisted-listing failure failing the whole enumeration; cancellation normalized to stable `CANCELLED`; and typed stable error codes. A companion spec (retired together with the query-backed read path) rejected eager evaluation of the optional session-query runtime while importing the ordinary subagent surface.
- `packages/subagent/tool-subagent-control/tests/list-agents.spec.ts` pins the `list_agents` schema (no parameters), the continuable-only projection that omits a healthy one-shot sibling while preserving diagnostics, the fixed child/diagnostic/empty text forms, an end-to-end settled-child listing with its durable label, forwarding of the tool cancellation signal, the no-agent rejection, the narrowed load requirement without `sessionQuery`, and HMR disposal.
- `packages/subagent/subagent/tests/list-children.spec.ts` pins the current read path against a real composition of the session store, JSONL persistence, spawn/fork providers, the subagent service, and the projection registry — no query service — keylessly: live-only listing without persistence; loud `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` and `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` even with zero children; the three-rung ladder (a live child never inspected, a cold child inspected exactly once, and the cache-hit, absent-key, absent-service, and poisoned-row second-rung cases); last-wins over multiple descriptors; malformed payloads and unknown versions diagnosed as `corrupt`; a failed cold inspection as one `unavailable` diagnostic retried on the next listing; a fork seed's ancestor descriptor listed under that identity; foreign-unit fold failures contained per child as `corrupt` on both the live and cold paths; `createdAt`-then-id ordering without ordinary forks; provider absence without child omission; compacted/uncompacted twins listing identically; a persisted-listing failure failing the whole enumeration; cancellation normalized to stable `CANCELLED`; typed stable error codes; and descendant listing's iterative stable pre-order, traversal through ordinary and one-shot intermediates, positioned diagnostics, lifecycle revalidation, and cancellation. A companion spec (retired together with the query-backed read path) rejected eager evaluation of the optional session-query runtime while importing the ordinary subagent surface.
- `packages/subagent/tool-subagent-control/tests/list-agents.spec.ts` pins the `list_agents` schema (one optional `scope` enum), the continuable-only projection that omits a healthy one-shot sibling while preserving diagnostics, registry-derived child/diagnostic/empty text forms, an end-to-end settled-child listing with its durable label, the descendants scope's pre-order parent/depth annotations across a live waiting branch, cancellation forwarding to both scopes, the no-agent rejection, the `agents` load requirement without `sessionQuery`, and HMR disposal.
- The keyless ACP snapshot scenario `subagent-list-agents` (examples/acp-agent) fences its second parent turn on a snapshot-only `subagent/end` marker, then executes `list_agents` for real against the subagent service, the projection registry, and JSONL persistence, rendering `<id> [complete] — <label>`.
- The keyless snapshot scenario `subagent-diagnostic` (examples/headless-agent) pins the current listing's model-visible diagnostic classification, including a descriptor-less settled child surfacing as a `corrupt` diagnostic.
- The keyless ACP snapshot scenario `subagent-published-run-failure` publishes a real one-shot child, injects independent run-result and handle-disposal failures, and preserves both diagnostics in the parent tool result.
.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.zh.md
+4 -4
View File
@@ -23,7 +23,7 @@ parent 到 child 的枚举是一项带消费方专用投影的服务功能。`Su
- 将语料活动状态单独报告为 `running` 或 `inactive`，但不暗示已完成或可恢复；
- 按 `createdAt` 升序、再按 child id 升序稳定返回所有结果 child。
每次普通的本地启动都会收到带可选、由调用方拥有之显示标签的 `one-shot` 描述符，而继续执行管理器会持久化带标签、包含附加重建字段的 `continuable` 描述符。面向模型的委派工具已经拥有简短 `description`，会将其用于一次性显示；workflow 等底层调用方无需凭空构造展示元数据。面向模型的 `list_agents` 适配器会将服务结果过滤为可继续 child，并将 `inactive` 映射为其现有的 `complete` 表示；UI 可以消费两种模式，并为无标签的一次性历史选择基于 id 的回退展示。描述符持久化、按 id 查找、直接 parent 鉴权和不依赖提供方的冷恢复仍归已实现的 Activation 契约负责。列表查询消费这些事实，但不能削弱它们，也不能另行发明第二种描述符表示。
每次普通的本地启动都会收到带可选、由调用方拥有之显示标签的 `one-shot` 描述符，而继续执行管理器会持久化带标签、包含附加重建字段的 `continuable` 描述符。面向模型的委派工具已经拥有简短 `description`，会将其用于一次性显示；workflow 等底层调用方无需凭空构造展示元数据。面向模型的 `list_agents` 适配器会将服务结果过滤为可继续 child，并通过在线 Agent 注册表细化状态（`running`／`idle`／`complete`）；UI 可以消费两种模式，并为无标签的一次性历史选择基于 id 的回退展示。描述符持久化、按 id 查找、直接 parent 鉴权和不依赖提供方的冷恢复仍归已实现的 Activation 契约负责。列表查询消费这些事实，但不能削弱它们，也不能另行发明第二种描述符表示。
### 枚举决策
@@ -52,7 +52,7 @@ subagent 服务将 `sessionQuery` 保持为可选依赖，因此没有该服务
有效描述符产生一个 child 条目，逐 child 检查失败产生一个 diagnostic 条目，缺少描述符的候选不产生条目。`mode` 是持久化创建策略；`activity` 是进程本地语料快照。活动状态既不是 `AgentStatus`、管理器内部的 Activation 状态，也不是持久化结果，结果不公开内部 `createdAt` 排序键。成功完成、失败、取消和停止原因等精确 Activation 状态与持久化结果需要单独的持久化激活记录，不在本功能范围内。
面向模型的 `list_agents` 工具不接受参数，从当前正在执行的 Agent 推导 `parentSessionId`，并作为 `@deepseek-ai/dsh-tool-subagent-control` 中的轻量适配器。它保留 diagnostic，丢弃 `one-shot` child 条目，将可继续 child 的 `running` 活动状态映射为 `running`、`inactive` 活动状态映射为 `complete`，然后按剩余的追踪顺序渲染 `<id> [<status>] — <label>` 或 `<id> [diagnostic: <reason>]`。空投影渲染为 `(no subagents)`。
面向模型的 `list_agents` 工具接受一个可选的 `scope: 'children' | 'descendants'` 参数，从当前执行 Agent 推导根 id，并作为 `@deepseek-ai/dsh-tool-subagent-control` 中的轻量适配器。它保留 diagnostic，丢弃 `one-shot` child 条目，状态取自在线 Agent 注册表——driver 活跃为 `running`，驻留但处于轮次之间为 `idle`，没有在线 Agent 时为 `complete`——然后按稳定目录顺序渲染 `<id> [<status>] — <label>` 或 `<id> [diagnostic: <reason>]`。`descendants` scope 读取 `SubagentService.listDescendants(rootSessionId)`：它从一份实时优先语料按稳定 pre-order 展平完整树，遍历普通与一次性中间节点以发现更深的可继续 agent，依据枚举生命周期重新校验每个冷候选，并为每个条目附加 `parentId`／`depth`。工具会在 label 之前插入 ` parent=<id> depth=<n>`；`parent` 是持久化直接 parent 会话 id，可能指向被省略的普通会话。对于当前调用方，只有 depth-1 child 条目可作为 `send_message` 候选，更深的 child 条目则可供 `interrupt_agent` 选择（[中断契约](2026-08-06-continuable-subagent-interrupt.md)）。发现结果只是提示——follow-up 权限仍仅属于确切直接 parent，中断权限仍由服务的在线 lineage 检查决定。空投影渲染为 `(no subagents)`。
diagnostic 使用三种固定原因。格式错误的事件 surface、精确加载 child 时发现的 header 冲突、读取结果中的不可变 header 与追踪到的候选不一致或不再指向请求的直接 parent、读取目标不再是先前定位的描述符事件、格式错误的描述符内容和多个描述符事件映射为 `corrupt`。未知描述符版本映射为 `unsupported`。逐 child 读取产生的 `SESSION_QUERY_SESSION_NOT_FOUND`、`SESSION_QUERY_EVENT_NOT_FOUND` 和 `SESSION_QUERY_PERSISTENCE_FAILED` 映射为 `unavailable`。这项阶段边界是有意为之：初始追踪期间发生持久化故障会让操作失败，而同一故障如果始于候选读取期间，可能会让每个受影响的 child 分别产生一条相同的 `unavailable` diagnostic；第一版既不合并这些 diagnostic，也不会把它们提升为全局失败。缺少描述符则作为非 subagent 排除，且不产生 diagnostic。配置错误、窗口错误和未识别的失败不属于 child diagnostic，会作为操作失败继续向上传播。每条 diagnostic 都标识 child id 及原因，不暴露对模型隐藏的描述符内容；系统会排除该候选，而其他健康的 sibling 仍然可见。系统绝不会读取不属于追踪结果直接后代的会话，也不会为它们产生 diagnostic。
@@ -93,8 +93,8 @@ diagnostic 是瞬时查询结果，不属于会话事件或目录状态。推导
## 测试
- `packages/subagent/subagent/tests/service.spec.ts` 固定两种模式下的描述符 v2 解析，并证明无标签的底层启动会在分发给提供方之前解析出一次性描述符。`packages/subagent/subagent-inprocess/tests/subagent-inprocess.spec.ts` 证明本地驱动会在初始轮次内追加该描述符，在取消落入工厂到 run 的交接窗口时返回已发布 id，并让结果与句柄释放失败保留在独立通道中。委派工具测试固定其现有显示说明的传递，并保留相互独立的结果与 dispose diagnostic。
- `packages/subagent/subagent/tests/list-children.spec.ts` 针对由会话存储、JSONL 持久化、spawn/fork 提供方、subagent 服务与投影注册表构成的真实组合——不含查询服务——以无密钥方式钉住现行读取路径：无持久化时的仅存活列表；零 children 也响亮报 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 与 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`；三级阶梯（存活 child 从不检查、冷 child 恰好检查一次，以及缓存命中、key 缺席、服务缺席、行中毒四个第二级用例）；多描述符 last-wins 取末者；载荷格式错误与未知版本诊断为 `corrupt`；冷检查失败成一条 `unavailable` diagnostic 并在下次列表重试；fork seed 中的祖先描述符按该身份列出；外部 unit 折叠失败在存活与冷两条路径上按 child 收纳为 `corrupt`；按 `createdAt` 再按 id 排序且不列普通 fork；提供方缺失时不排除 child；压缩与未压缩的孪生 child 列表结果一致；持久化列表失败使整次枚举失败；取消稳定归一化为 `CANCELLED`；以及带类型的稳定错误码。一个伴随规格(已随查询式读取路径一起退役)曾在导入普通 subagent surface 时拒绝对可选 session-query 运行时的 eager 求值。
- `packages/subagent/tool-subagent-control/tests/list-agents.spec.ts` 固定 `list_agents` 的 schema（无参数）、只保留可继续 child 且排除健康的一次性 sibling、同时保留 diagnostic 的投影、child／diagnostic／空结果的固定文本形式、带持久化 label 的已结束 child 端到端列表、工具取消信号的转发、无调用 agent 时的拒绝、收窄后的加载要求（不再注入 `sessionQuery`），以及 HMR dispose。
- `packages/subagent/subagent/tests/list-children.spec.ts` 针对由会话存储、JSONL 持久化、spawn/fork 提供方、subagent 服务与投影注册表构成的真实组合——不含查询服务——以无密钥方式钉住现行读取路径：无持久化时的仅存活列表；零 children 也响亮报 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 与 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`；三级阶梯（存活 child 从不检查、冷 child 恰好检查一次，以及缓存命中、key 缺席、服务缺席、行中毒四个第二级用例）；多描述符 last-wins 取末者；载荷格式错误与未知版本诊断为 `corrupt`；冷检查失败成一条 `unavailable` diagnostic 并在下次列表重试；fork seed 中的祖先描述符按该身份列出；外部 unit 折叠失败在存活与冷两条路径上按 child 收纳为 `corrupt`；按 `createdAt` 再按 id 排序且不列普通 fork；提供方缺失时不排除 child；压缩与未压缩的孪生 child 列表结果一致；持久化列表失败使整次枚举失败；取消稳定归一化为 `CANCELLED`；带类型的稳定错误码；以及后代列表的迭代式稳定 pre-order、穿过普通与一次性中间节点、带位置 diagnostic、生命周期复验与取消。一个伴随规格(已随查询式读取路径一起退役)曾在导入普通 subagent surface 时拒绝对可选 session-query 运行时的 eager 求值。
- `packages/subagent/tool-subagent-control/tests/list-agents.spec.ts` 固定 `list_agents` 的 schema（一个可选 `scope` 枚举）、只保留可继续 child 且排除健康的一次性 sibling、同时保留 diagnostic 的投影、由注册表推导的 child／diagnostic／空结果文本形式、带持久化 label 的已结束 child 端到端列表、descendants scope 在在线 waiting 分支上的 pre-order parent/depth 注释、两个 scope 的取消信号转发、无调用 agent 时的拒绝、要求 `agents` 但不再注入 `sessionQuery` 的加载契约，以及 HMR dispose。
- 无密钥 ACP 快照场景 `subagent-list-agents`（examples/acp-agent）使用仅限快照的 `subagent/end` 标记为第二个 parent 轮次设置边界，随后针对 subagent 服务、投影注册表和 JSONL 持久化真实执行 `list_agents`，渲染 `<id> [complete] — <label>`。
- 无密钥快照场景 `subagent-diagnostic`（examples/headless-agent）钉住现行列表的模型可见诊断分类，包括无描述符的定局 child 以 `corrupt` diagnostic 出现。
- 无密钥 ACP 快照场景 `subagent-published-run-failure` 会发布一个真实的一次性 child，注入相互独立的 run result 与 handle dispose 失败，并在 parent 工具结果中保留两项 diagnostic。
.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md
2026-07-27-web-subagent-conversations.md: edaaa97bbfcf0ce1751db84d8fc2d3ab07af8b03
2026-07-27-web-subagent-conversations.zh.md: 4c5e670cc46621b889d8463daa90afbf43bde30f
2026-07-27-web-subagent-conversations.md: da4de7d4bcd23f61459eba3fee8108b0af1c95f9
2026-07-27-web-subagent-conversations.zh.md: 43d9710ec2b63a09bd3db029f172746d63d91cfe
.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md
+6 -6
View File
@@ -20,7 +20,7 @@ Every opened child carries a catalog-derived address `{ parentSessionId, childSe
The generic Host domain preserves the same ownership boundary. `session.history` and the source side of `session.fork` read an attached Session or inspect persistence without acquiring an Agent; history folds cold projection values from that exact inspected prefix, while a fork publishes an ordinary independent session. Generic Agent-bound session, command, and goal routes return `agent-busy` for session-backed subagents, as do explicit-id `session.create` adoption and attached-only queue controls. The denial classifier accepts the coarse `origin` marker, a `subagent/descriptor` in the session's own suffix, or exact live runtime ownership by the parent; these signals only prevent generic ownership and never replace catalog mode or direct-parent authorization.
The ordinary Stop action is absent from addressed child conversations. `SubagentService.followup()` owns admission only until inbox acceptance and intentionally exposes no public child cancellation operation. A later cancellation design needs an explicit authority and lifecycle contract rather than falling through to `session.cancel`.
Stopping an addressed child never falls through to `session.cancel`. `SubagentService.followup()` owns admission only until inbox acceptance and grants no cancellation handle; a running continuable child is stopped through the dedicated `subagent.interrupt` route under the [current-turn interrupt contract](2026-08-06-continuable-subagent-interrupt.md), which parks pending work instead of discarding it. One-shot children remain uncancellable from the Web.
This decision covers Web discovery, transcript viewing, and parent-authorized human continuation. It does not make a subagent independently user-owned; that product remains [interactive side sessions](../../proposed/feature/2026-07-08-interactive-side-sessions.md).
@@ -45,7 +45,7 @@ Healthy rows reuse the standard session projections retained in the list mirror.
Selecting a row records its exact address before opening the resident client `Session`. History pagination, event folding, tool render intents, titles, and live mux reconciliation reuse the ordinary conversation machinery. Breadcrumbs use catalog labels, follow parent links only through `origin: 'subagent'` rows, include the first ordinary owner, and keep ordinary forks single-level. Forking an addressed subagent creates an ordinary fork with direct source lineage and attaches it to the nearest workspace-owning ancestor. The catalog is an ARIA tree with lazy ArrowRight/ArrowLeft disclosure, linear ArrowUp/ArrowDown navigation, Home/End, Escape, and focus restoration.
A one-shot row always replaces the composer with copy explaining that the execution record is read-only. A continuable row does so only while `parentAvailable` is false. When enabled, its Send action admits another FIFO turn even if the child is currently running; it never becomes Stop. Prompt failures retain the draft through the ordinary error behavior.
A one-shot row always replaces the composer with copy explaining that the execution record is read-only. A continuable row does so only while `parentAvailable` is false and the child is not running; a running parent-offline child keeps the ordinary composer with its input disabled so the same primary Stop stays reachable, and the read-only takeover returns once it stops. With a live parent, Enter admits another FIFO turn even while the child runs, and the single primary action follows the ordinary running convention: Send while idle, Stop while running, routed through `subagent.interrupt` ([interrupt contract](2026-08-06-continuable-subagent-interrupt.md)). Prompt failures retain the draft through the ordinary error behavior.
Agent-bound auxiliary controls are unavailable in addressed child views. In particular, the model selector and `/model` contribution do not call ordinary `session.models` or `session.selectModel`; the Host also rejects any accidental call instead of activating persisted child history outside the direct-parent continuation seam.
@@ -89,7 +89,7 @@ The shipped Web composition mounts SQLite session query beside JSONL persistence
**Auto-resume an absent parent.** Rejected because continuation requires the exact live direct parent. Child navigation must not mutate the parent lifecycle.
**Expose ordinary cancellation.** Rejected because the accepted inbox turn outlives its admission request and the continuation seam exposes no authority-safe cancellation handle.
**Expose ordinary cancellation.** Rejected because the accepted inbox turn outlives its admission request and, at this decision's time, the continuation seam exposed no authority-safe cancellation handle. The later [current-turn interrupt contract](2026-08-06-continuable-subagent-interrupt.md) added that explicit authority as a dedicated subagent route; falling through to `session.cancel` remains rejected.
**Show only continuable children.** Rejected because the durable catalog deliberately describes both session-backed modes. One-shot transcripts remain useful even though they never accept follow-ups.
@@ -103,9 +103,9 @@ The shipped Web composition mounts SQLite session query beside JSONL persistence
- Host protocol tests pin schemas including required boolean expandability, id echoing, mode verification, non-activating history, exact-parent enforcement, FIFO admission receipts, cancellation, and sanitized failure mapping.
- Generic Host tests pin attached and cold history and forks without Agent publication, cold projection folding, descriptor/origin/runtime-owner denial, explicit-id adoption denial, and the direct queue-control fence.
- Client object tests pin retained and restored addresses, one-shot read-only rejection, history routing, continuable prompt routing, no addressed cancellation, suppression of Agent-bound model controls, live activity flips including in-flight response replay and detach fallback, subagent-parent expandability flips, and membership refresh.
- Client object tests pin retained and restored addresses, one-shot read-only and cancel rejection, history routing, continuable prompt and interrupt routing, suppression of Agent-bound model controls, live activity flips including in-flight response replay and detach fallback, subagent-parent expandability flips, and membership refresh.
- jsdom tests pin the aggregate descendant count and activity, sidebar propagation across nested lineage and ordinary-fork boundaries, row-status precedence, token totals, second-precision running and frozen inactive durations, adaptive long-duration units with exact accessible text, the summary-backed root action across absent and stale-empty catalogs, known loading-row shape, mixed-mode rows, pre-click leaf disclosure, diagnostics, lazy descendant disclosure, direct-parent addresses, keyboard behavior, and both read-only reasons.
- The keyless assembled Web snapshot contains an inactive continuable child with durable usage, an inactive one-shot sibling with a deterministic long duration, and a persisted grandchild; it pins the three-descendant trigger across a stale empty catalog response, usage and timing rows, and adaptive long-duration presentation, expands without activation, opens persisted history, admits a human FIFO follow-up, reconciles child mux events, and proves one-shot history remains read-only. A separate assembled scenario holds a real child Agent turn at the model seam while it pins the aggregate running state in both the header and visible idle owner row, then cancels the turn during teardown.
- The keyless assembled Web snapshot contains an inactive continuable child with durable usage, an inactive one-shot sibling with a deterministic long duration, and a persisted grandchild; it pins the three-descendant trigger across a stale empty catalog response, usage and timing rows, adaptive long-duration presentation, and the aggregate running transition, expands without activation, opens persisted history, admits a human FIFO follow-up, reconciles child mux events, and proves one-shot history remains read-only. A separate assembled scenario holds a real child Agent turn at the model seam while it pins the aggregate running state in both the header and visible idle owner row, then cancels the turn during teardown.
- Navigation tests pin subagent-only breadcrumbs, workspace placement for forks created from subagents, and `origin: 'subagent'` sidebar filtering without hiding ordinary forks.
## Consequences
@@ -114,4 +114,4 @@ The shipped Web composition mounts SQLite session query beside JSONL persistence
- Parent availability, child activity, and `hasChildren` are snapshots. Publication, disposal, another sender, or another process may win after listing; typed prompt failure remains expected.
- A child may publish between history fetch and mux subscription, so the existing sequence reconciliation also covers the cold-to-live addressed path.
- Persisted origin adds one deliberately weak product-classification field to child headers and list projections; it cannot become an authorization shortcut.
- The UI has no child cancellation, durable outcome, Activation identity, deletion, or independently interactive offline mode, and its text must not imply those capabilities. Active-turn duration measures logged work rather than Activation residency.
- Beyond the current-turn Stop of a running continuable child ([interrupt contract](2026-08-06-continuable-subagent-interrupt.md)), the UI has no child cancellation, durable outcome, Activation identity, deletion, or independently interactive offline mode, and its text must not imply those capabilities. Active-turn duration measures logged work rather than Activation residency.
.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.zh.md
+6 -6
View File
@@ -20,7 +20,7 @@ Web 产品通过页头操作公开选中会话中由会话支撑的直接 subage
通用 Host 领域遵守同一所有权边界。`session.history` 与 `session.fork` 的源端会读取已附加 Session 或检查持久化存储，而不获取 Agent；history 从所检查的确切前缀归并冷态投影值，fork 则发布一个普通的独立会话。绑定到 Agent 的通用会话、命令与目标路由会对由会话支撑的 subagent 返回 `agent-busy`；显式 id 的 `session.create` 接纳与仅针对已附加会话的队列控件亦然。拒绝分类器接受粗粒度 `origin` 标记、会话自身后缀中的 `subagent/descriptor`，或 parent 对其确切的存活运行时所有权；这些信号只会阻止通用路径取得所有权，绝不取代目录 mode 或直接 parent 授权。
已寻址 child 对话不提供普通 Stop 操作。`SubagentService.followup()` 只负责消息被 inbox 接受前的准入，并有意不公开任何 child 取消操作。后续取消设计需要显式的授权与生命周期契约，而不能回退到 `session.cancel`。
停止一个已寻址 child 绝不回退到 `session.cancel`。`SubagentService.followup()` 只负责消息被 inbox 接受前的准入，不授予取消句柄；正在运行的可继续 child 通过专用的 `subagent.interrupt` 路由停止，遵循[当前轮次中断契约](2026-08-06-continuable-subagent-interrupt.md)，该契约会暂停而非丢弃待处理工作。one-shot child 在 Web 端仍不可取消。
本决策涵盖 Web 端发现、transcript 查看与经 parent 授权的用户继续交互。它不会让 subagent 成为用户独立所有的对象；这类产品仍然属于[交互式 side session](../../proposed/feature/2026-07-08-interactive-side-sessions.md)。
@@ -45,7 +45,7 @@ Figma 中的 [subagent 列表](https://www.figma.com/design/jRBBK7zBgcszdVWQ0Fh5
选择一行后，系统会先记录其确切地址，再打开常驻客户端 `Session`。历史分页、事件 fold、工具渲染意图、title 与实时 mux 归并都会复用普通对话机制。面包屑导航使用目录 label，只会沿 `origin: 'subagent'` 行的父链接逐级回溯，包含第一个普通 owner，并让普通 fork 保持单层。从已寻址 subagent 创建 fork 时，会生成具有直接源谱系的普通 fork，并将其附加到最近拥有 Workspace 的祖先。目录是一棵 ARIA 树，支持懒加载式 ArrowRight／ArrowLeft 展开与折叠、线性 ArrowUp／ArrowDown 导航、Home／End、Escape 以及焦点恢复。
one-shot 行始终会用文案替代输入框，说明执行记录为只读。可继续行仅在 `parentAvailable` 为 false 时如此。启用后，即使 child 正在运行，其 Send 操作也会准入另一个 FIFO 轮次，绝不会变成 Stop。提示词失败会通过普通错误行为保留草稿。
one-shot 行始终会用文案替代输入框，说明执行记录为只读。可继续行仅在 `parentAvailable` 为 false 且 child 未在运行时如此；parent 离线但仍在运行的 child 保留普通输入框，只禁用其输入区，让同一个 primary Stop 保持可达，停止后只读替代恢复。parent 在线时，即使 child 正在运行，Enter 也会准入另一个 FIFO 轮次，而唯一的 primary 操作遵循普通运行约定：空闲时为 Send，运行时为 Stop，经由 `subagent.interrupt` 路由（[中断契约](2026-08-06-continuable-subagent-interrupt.md)）。提示词失败会通过普通错误行为保留草稿。
已寻址 child 视图不提供绑定到 agent 的辅助控件。具体而言，模型选择器与 `/model` contribution 不会调用普通 `session.models` 或 `session.selectModel`；Host 也会拒绝任何意外调用，而不是在直接 parent 继续执行 seam 之外激活持久化 child 历史。
@@ -89,7 +89,7 @@ one-shot 行始终会用文案替代输入框，说明执行记录为只读。
**自动恢复缺失的 parent。** 不予采纳，因为继续执行要求确切的存活直接 parent。child 导航不得改变 parent 生命周期。
**公开普通取消操作。** 不予采纳，因为已获 inbox 接受的轮次会比其准入请求存续更久，而继续执行 seam 不会公开具备安全授权的取消句柄。
**公开普通取消操作。** 不予采纳，因为已获 inbox 接受的轮次会比其准入请求存续更久，且在本决定当时，继续执行 seam 未公开具备安全授权的取消句柄。后来的[当前轮次中断契约](2026-08-06-continuable-subagent-interrupt.md)以专用 subagent 路由补上了这项显式授权；回退到 `session.cancel` 仍被拒绝。
**只显示可继续 child。** 不予采纳，因为持久化目录有意描述由会话支撑的两种 mode。one-shot transcript 即使绝不接受后续消息，仍然有用。
@@ -103,9 +103,9 @@ one-shot 行始终会用文案替代输入框，说明执行记录为只读。
- 宿主协议测试固定 schema（包括必需的布尔可展开性）、id 回显、mode 校验、非激活式历史、确切 parent 强制要求、FIFO 准入回执、取消与脱敏后的失败映射。
- 通用 Host 测试固定在不发布 Agent 的情况下读取已附加与冷态历史及执行 fork、冷态投影归并、按描述符／origin／运行时 owner 拒绝、拒绝显式 id 接纳，以及直接队列控制栅栏。
- 客户端对象测试固定已保留与已恢复的地址、one-shot 只读拒绝、历史路由、可继续提示词路由、已寻址对话不提供取消、屏蔽绑定到 agent 的模型控件、实时活动状态翻转（包括在途响应回放与 detach 回退）、subagent parent 可展开性翻转与成员刷新。
- 客户端对象测试固定已保留与已恢复的地址、one-shot 只读与取消拒绝、历史路由、可继续提示词与中断路由、屏蔽绑定到 agent 的模型控件、实时活动状态翻转（包括在途响应回放与 detach 回退）、subagent parent 可展开性翻转与成员刷新。
- jsdom 测试固定后代聚合计数与活动状态、侧边栏活动在嵌套谱系中的传播与普通 fork 边界、行状态优先级、token 用量总计、精确到秒的运行中耗时与冻结后 inactive 耗时、采用自适应单位的长耗时及其精确无障碍文本、目录缺失或为陈旧空目录时由摘要支撑的根操作、已知加载行的形态、混合 mode 行、点击前的叶子展开控件、diagnostic、后代懒加载展开、直接 parent 地址、键盘行为与两种只读原因。
- 无密钥的组装 Web 快照包含一个具有持久化 token 用量的 inactive 可继续 child、一个具有确定性长耗时的 inactive one-shot sibling 和一个持久化 grandchild；它会固定触发器在一次陈旧的空目录响应后仍显示三个后代，并固定 token 用量与计时行及自适应长耗时呈现，在不激活的情况下展开、打开持久化历史、准入一条用户 FIFO 后续消息、归并 child mux 事件，并证明 one-shot 历史仍然只读。另一个独立的组装场景会在 model seam 处保持一个真实的 child Agent 轮次进行中，同时固定页头和可见空闲 owner 行中的聚合运行状态，随后在 teardown 期间取消该轮次。
- 无密钥的组装 Web 快照包含一个具有持久化 token 用量的 inactive 可继续 child、一个具有确定性长耗时的 inactive one-shot sibling 和一个持久化 grandchild；它会固定触发器在一次陈旧的空目录响应后仍显示三个后代，并固定 token 用量与计时行、自适应长耗时呈现及聚合 `running` 状态转换，在不激活的情况下展开、打开持久化历史、准入一条用户 FIFO 后续消息、归并 child mux 事件，并证明 one-shot 历史仍然只读。另一个独立的组装场景会在 model seam 处保持一个真实的 child Agent 轮次进行中，同时固定页头和可见空闲 owner 行中的聚合运行状态，随后在 teardown 期间取消该轮次。
- 导航测试固定仅含 subagent 的面包屑导航、从 subagent 创建 fork 时的 Workspace 归属，以及 `origin: 'subagent'` 侧边栏过滤，同时不隐藏普通 fork。
## 后果
@@ -114,4 +114,4 @@ one-shot 行始终会用文案替代输入框，说明执行记录为只读。
- parent 可用性、child 活动状态与 `hasChildren` 都是快照。列出之后，发布、dispose、其他发送方或其他进程都可能抢先改变状态；类型化提示词失败仍属预期行为。
- child 可能在历史获取与 mux 订阅之间发布，因此现有序号归并也涵盖从冷态转为存活的已寻址路径。
- 持久化 origin 会为 child header 与列表投影添加一个有意保持弱约束的产品分类字段；它不能变成授权捷径。
- UI 不提供 child 取消、持久化结果、Activation 身份、删除或可独立交互的离线 mode，其文案不得暗示这些功能已经存在。活跃轮次耗时度量的是已记录工作，而非 Activation 驻留时间。
- 除对正在运行的可继续 child 的当前轮次 Stop（[中断契约](2026-08-06-continuable-subagent-interrupt.md)）之外，UI 不提供 child 取消、持久化结果、Activation 身份、删除或可独立交互的离线 mode，其文案不得暗示这些功能已经存在。活跃轮次耗时度量的是已记录工作，而非 Activation 驻留时间。
.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md
2026-07-28-continuable-subagent-conversations.md: a5a900c3bb30a8d965aabc0bf498f8399ee70e7a
2026-07-28-continuable-subagent-conversations.zh.md: bde8e874ba27950dc0fab8449c6ba2ee49737d41
2026-07-28-continuable-subagent-conversations.md: 56d9abb2b09577b2fba14b9a655a941417e0c493
2026-07-28-continuable-subagent-conversations.zh.md: 77f497ceeef35dcc519242dca30c81759cc3b907
.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md
+3 -3
View File
@@ -131,7 +131,7 @@ Parent-originated delivery requires the parent to be live when admitted and keep
### Durability, disposal, and recovery
Without Tasks there is no `task_output`, `task_kill`, Task status, per-message result promise, or public subagent cancellation operation. The caller signal can abort start or follow-up only before inbox acceptance. After acceptance, the parent cannot cancel the message, turn, or Activation through `ctx.subagents`; `Agent.cancel()` remains a lower-level Agent capability that this version does not expose through the subagent service.
Without Tasks there is no `task_output`, `task_kill`, Task status, or per-message result promise. The caller signal can abort start or follow-up only before inbox acceptance. After acceptance, the parent cannot cancel the accepted message or dispose the Activation through `ctx.subagents`; the only public stop is the later [current-turn interrupt](2026-08-06-continuable-subagent-interrupt.md), which cancels the live target's current turn with `keepInbox` and leaves residency, pending work, and descendants intact.
Host and manager teardown remains the lifecycle stop path. Manager unload applies it globally; a host applies it only below the exact top-level Agents it owns. Each form closes the applicable admission scope, stops the selected visible Activations, awaits admitted materializations in that scope, releases child-first, and preserves the durable Sessions.
@@ -145,7 +145,7 @@ Session and descriptor persistence survive restart. Activation state, Agent inbo
This version covers continuable in-process children and leaves one-shot delegation unchanged. Remote providers require a separate Activation handle with equivalent authenticated control and child-first quiescence contracts before they can support the same behavior.
It adds no host-user continuation, subagent steering operation, durable mailbox, cross-process lease, automatic replay of interrupted inbox work, team authority, workflow authority, public subagent cancellation operation, public residency query, new live-Activation or descendant limit, or runtime cache. Existing delegation-depth policy remains unchanged. Optional child-to-parent reporting is a later consumer of this lifecycle rather than part of the base continuable capability.
It adds no host-user continuation, subagent steering operation, durable mailbox, cross-process lease, automatic replay of interrupted inbox work, team authority, workflow authority, public residency query, new live-Activation or descendant limit, or runtime cache; the later [current-turn interrupt](2026-08-06-continuable-subagent-interrupt.md) added the one public stop operation on top of this lifecycle. Existing delegation-depth policy remains unchanged. Optional child-to-parent reporting is a later consumer of this lifecycle rather than part of the base continuable capability.
## Alternatives considered
@@ -186,7 +186,7 @@ The implementation pins these behaviors:
- `followup()` accepts only the exact live direct parent and rechecks that identity at the final no-await inbox-admission boundary after any materialization; durable message provenance cannot authorize delivery.
- Continuation messages always use `Agent.followup()` and share its inbox FIFO, including when the child already has an open turn.
- `ctx.subagents.followup()` and its `send_message` adapter return only the accepted `MessageId`; the continuation layer accepts no delivery target and defines no subagent-specific route result.
- This version exposes no public subagent cancellation operation; caller signals stop start and follow-up only before inbox acceptance, while host-scoped and manager-global teardown retain child-first cleanup.
- Caller signals stop start and follow-up only before inbox acceptance, while host-scoped and manager-global teardown retain child-first cleanup; the [current-turn interrupt](2026-08-06-continuable-subagent-interrupt.md) is the one public stop and does not enter teardown.
- This version exposes no subagent steering operation or current-turn controller state.
- An idle Agent with live owned children yields a `waiting` Activation whose `AgentHandle` remains retained.
- A `next-turn` delivered to `waiting` wakes the same Activation; delivery after completed disposal cold-resumes a new Activation.
.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.zh.md
+3 -3
View File
@@ -131,7 +131,7 @@ activation-owner 作用域之所以存在，是因为普通 Cordis owner effect
### 持久性、dispose 与恢复
没有 Task 后，系统不再提供 `task_output`、`task_kill`、Task 状态、逐消息结果 promise 或公开 subagent 取消操作。调用方 signal 只能在 inbox 接受消息前中止 start 或 follow-up。消息被接受后，parent 不能通过 `ctx.subagents` 取消该消息、轮次或激活；`Agent.cancel()` 仍是底层 Agent 能力，但本版本不通过 subagent 服务暴露它。
没有 Task 后，系统不再提供 `task_output`、`task_kill`、Task 状态或逐消息结果 promise。调用方 signal 只能在 inbox 接受消息前中止 start 或 follow-up。消息被接受后，parent 不能通过 `ctx.subagents` 取消已接受的消息或 dispose 激活；唯一的公开停止操作是后来的[当前轮次中断](2026-08-06-continuable-subagent-interrupt.md)，它以 `keepInbox` 取消在线目标的当前轮次，驻留、待处理工作与后代均保持不变。
宿主和管理器拆卸仍是生命周期停止路径。管理器卸载会全局应用它；宿主只会在自己确切拥有的顶层 Agent 之下应用它。两种形式都会关闭适用的准入作用域，停止选中的可见 Activation，等待该作用域中已获准的物化过程，按 child-first 顺序释放，并保留持久化 Session。
@@ -145,7 +145,7 @@ activation-owner 作用域之所以存在，是因为普通 Cordis owner effect
本版本覆盖可继续的进程内 child，一次性委派保持不变。远程提供方必须具备单独的激活 handle，以及等价的认证控制与 child-first 完全停稳契约，才能支持同样的行为。
它不新增 host-user 继续执行、subagent steering 操作、持久化邮箱、跨进程 lease、中断 inbox 工作的自动回放、团队权限、工作流权限、公开 subagent 取消操作、公开驻留查询、新的在线激活数量或后代总数限制，以及运行时缓存。现有委派深度策略保持不变。可选的 child 到 parent 报告是后续消费该生命周期的功能，不属于基础可继续能力。
它不新增 host-user 继续执行、subagent steering 操作、持久化邮箱、跨进程 lease、中断 inbox 工作的自动回放、团队权限、工作流权限、公开驻留查询、新的在线激活数量或后代总数限制，以及运行时缓存；后来的[当前轮次中断](2026-08-06-continuable-subagent-interrupt.md)在此生命周期之上补充了唯一的公开停止操作。现有委派深度策略保持不变。可选的 child 到 parent 报告是后续消费该生命周期的功能，不属于基础可继续能力。
## 曾考虑的替代方案
@@ -186,7 +186,7 @@ activation-owner 作用域之所以存在，是因为普通 Cordis owner effect
- `followup()` 只接受确切的在线直接 parent，并在任何物化之后的最终无 await 的 inbox 准入边界再次检查该身份；持久化消息来源信息不能授权投递。
- 继续执行消息始终使用 `Agent.followup()` 并共享其 inbox FIFO，包括 child 已有开放轮次的情况。
- `ctx.subagents.followup()` 及其 `send_message` 适配器只返回已接受的 `MessageId`；继续执行层不接受投递 target，也不定义 subagent 专属路由结果。
- 本版本不暴露公开 subagent 取消操作；调用方 signal 只能在 inbox 接受消息前停止 start 和 follow-up，限定到宿主的拆卸与管理器全局拆卸则保留 child-first 清理。
- 调用方 signal 只能在 inbox 接受消息前停止 start 和 follow-up，限定到宿主的拆卸与管理器全局拆卸则保留 child-first 清理；[当前轮次中断](2026-08-06-continuable-subagent-interrupt.md)是唯一的公开停止操作，且不进入拆卸流程。
- 本版本不暴露 subagent steering 操作或当前轮次控制方状态。
- 带有在线所持 child 的空闲 Agent 会产生 `waiting` 激活，其 `AgentHandle` 继续保留。
- 向 `waiting` 投递 `next-turn` 会唤醒同一个激活；完成 dispose 后投递消息会冷恢复新激活。
.agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.i18n.yaml
+6
View File
@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.md
2026-08-06-continuable-subagent-interrupt.md: 0f663ac6a58b9c3d658aa572ca2fea18dbef6449
2026-08-06-continuable-subagent-interrupt.zh.md: b8a5b557b73b2845e3c0b1dabdec0163f56eca48
.agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.md
+48
View File
@@ -0,0 +1,48 @@
# Agent Note: Continuable subagent current-turn interrupt
Status: implemented
English | [中文](2026-08-06-continuable-subagent-interrupt.zh.md)
## Problem
A running continuable subagent could not be stopped without destroying it. The continuation manager cancels child Agents only inside whole-Activation teardown (settlement, drain, scoped drain), `send_message`/`subagent.prompt` only add work, and the Web composer's Stop button was deliberately limited to ordinary sessions. A human watching a continuable child burn tokens on a wrong path had no lever short of killing the parent tree, and when the direct parent Agent was offline the child was entirely untouchable even though its Activation stayed live. One-shot runs have holder-owned disposal and task-kill; continuable children had no analogous current-turn control.
## Decision
`ctx.subagents.interrupt(targetSessionId, authority)` stops only the live target's current turn. The manager primitive authorizes synchronously, calls the existing `Agent.cancel(cause, { keepInbox: true })`, and returns `void` — fire-and-return: the cancel signal is guaranteed issued, target quiescence is not awaited. Nothing else changes: no Activation disposal, no handle release, no descendant cascade, no inbox clearing, and no `AgentLoop` or `CancelOptions` change. Because `keepInbox` parks the unclaimed pending queue at idle, an interrupt never auto-starts the next queued follow-up; work already claimed into the interrupted turn belongs to that turn and is not requeued. Once the interrupted driver is idle, an explicit waking send resumes the preserved FIFO order.
Authority is a closed two-variant union, deliberately wider than delivery authority because stopping a turn is idempotent and delivers no content:
- `{ kind: 'user', parentSessionId }` — a human presents the durable direct-parent address. The live target's `session.header.parentSession` must match; no live parent Agent, catalog read, or persistence access is involved, which is exactly what keeps a live child stoppable while its parent Agent is offline. Cancel cause `user`.
- `{ kind: 'ancestor', agent }` — an exact live ancestor Agent (direct parent or deeper). The caller must be the registry's current entry for its id (stale callers are rejected even for absent targets), must not be the target itself, and must appear in the Activation's materialization-time `ancestry` WeakSet. Cancel cause `parent`.
Targets are resolved only in the manager's process-local Activation map. An absent id — unknown, one-shot, or naturally settled — is an accepted no-op, which uniformly covers completion races and repeat requests without leaking durable-catalog information; a target whose disposal transaction is already open is likewise an accepted no-op after authorization. One-shot lifecycle (holder `dispose()`, task-kill) is untouched. `SubagentService.interrupt()` treats a manager-less composition as an accepted no-op rather than `CONTINUATION_UNAVAILABLE`, because without a manager no manager-owned live Activation can exist.
The Host RPC `subagent.interrupt` takes the continuable `SubagentAddress` and returns `{ accepted: true }`. Its implementation calls only the core primitive with `user` authority — deliberately no `catalogChild()`, `listChildren()`, `sessionQuery`, or parent-registry lookup. A live target with a mismatched parent address maps to `subagent-unauthorized`; unexpected failures map to `internal` without leaking error text onto the wire.
## Alternatives considered
**Route human interrupts through `session.cancel`.** The generic session cancel requires an attached ordinary session and rejects subagent-owned sessions; widening it would entangle subagent authority rules with ordinary session routing. A subagent-domain RPC keeps the address-based authorization and the parent-offline guarantee explicit.
**Await target quiescence and return the turn outcome.** Cancellation is cooperative, so quiescence is unbounded; holding the RPC (and a `ChildLock` slot) open invites timeouts and convoying against delivery and disposal. Acceptance-of-signal is the only fact the caller needs, and races (natural completion, disposal) already settle idempotently.
**Reuse whole-Activation disposal for interrupt.** Disposal cancels without `keepInbox`, flushes, captures, and releases the handle — it destroys queued work and the child's residency. Interrupt is a control operation on one turn, not a lifecycle operation on the Activation.
**Extend `send_message`/`followup` authority to ancestors while at it.** Delivery injects content into a conversation and is not idempotent; its exact-direct-parent authority stays unchanged. Only interrupt gets the wider ancestor and address-based user authority.
**Auto-resume the parked queue after an interrupt.** Immediately starting queued follow-up B after aborting A would make the interrupt look ignored and steal the human's window to redirect the child. Parking until an explicit waking send keeps the stop observable and the FIFO order intact.
## Consequences
A human or ancestor can stop a runaway continuable turn without losing the child, its unclaimed queued work, or its running descendants; the cost is a deliberately weak postcondition (`accepted` means "signal issued", so a target may remain visibly `running` until it observes the signal) that clients must render honestly. The parked-queue rule means an interrupted child sits idle with retained work until a waking message arrives after the driver is idle — an intentional human-in-the-loop pause, not a scheduler defect. A waking send accepted during abort convergence currently remains queued without latching wake; Issue #1838 tracks the shared agent-loop correction.
The address-only RPC exposes one bit of live residency: an absent target is accepted while a live target under a mismatched parent returns `subagent-unauthorized`. The single-user local Host trust model accepts that observability; a future multi-principal Host must revisit both authority and response indistinguishability.
The Web surface reuses the existing primary Send/Stop toggle rather than adding a second action: the client `Session.cancel()` routes a continuable address through `subagent.interrupt` (one-shot addresses stay uncancellable, ordinary sessions keep `session.cancel`), and a running parent-offline continuable child keeps the default composer with disabled input so that same primary Stop remains reachable, returning to the read-only takeover once it stops ([Web subagent conversations](2026-07-27-web-subagent-conversations.md) owns the surrounding catalog and composer contract).
The model-facing `interrupt_agent(agent_id)` tool in `dsh-tool-subagent-control` passes `exec.agent` as the `ancestor` authority and adds none of its own: the core primitive verifies live registry identity and recorded lineage, so the tool can name a direct child or a deeper descendant with the same generic `agent_id` parameter — deliberately not `subagent_id`, which would imply direct children only. Discovery rides `list_agents({ scope: 'descendants' })` over the new `SubagentService.listDescendants()` one-trace pre-order walk with verified `parentId`/`depth` per entry ([durable catalog note](2026-07-22-durable-subagent-catalog-and-list-agents.md) owns the listing contract); discovery is a hint, never authority. `send_message` keeps its exact-direct-parent authority — only interrupt is ancestor-wide.
## Testing
Core coverage in `packages/subagent/subagent/tests/continuation.spec.ts` proves the durable `turn/end` abort, parked-then-FIFO-resumed queue, untouched descendant, both authority kinds with their cancel causes, self/sibling/stale/non-ancestor rejection, absent/one-shot/disposal-race no-ops, and the unchanged `keepInbox` loop behavior. Host coverage in `packages/host/apiproxy/tests` proves the RPC calls only the core primitive (no agents/catalog/history reads), the `subagent-unauthorized`/`internal` mappings, the wire schema's continuable-mode fence, and carrier round-trips. Client coverage pins the address-routed `Session.cancel()`, the InputBar Send/Stop toggle with the parent-offline locked-input state, and the read-only-composer selector's running exception; the keyless assembled Web scenarios (`apps/web/tests/subagent-interrupt.e2e.ts`, `subagent-interrupt-ui.e2e.ts`) hold a real child turn open with a replay hang entry and prove the interrupt transport, the aborted `turn/end`, the parked follow-up, and the FIFO resume end to end. Tool coverage in `packages/subagent/tool-subagent-control/tests` proves direct and deep ancestor interrupts with the `parent` cause and parked queue, self/sibling/stranger rejection without touching the target, absent-target no-ops without cold resume, and the descendants listing's pre-order positions; the keyless ACP snapshots pin the new tool schemas in every recorded request header.
.agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.zh.md
+48
View File
@@ -0,0 +1,48 @@
# Agent Note: Continuable subagent 当前轮次中断
Status: implemented
[English](2026-08-06-continuable-subagent-interrupt.md) | 中文
## Problem
一个正在运行的 continuable subagent 无法在不销毁它的前提下被停止。继续执行管理器只在整个 Activation 拆除（结算、drain、scoped drain）内部取消子 Agent，`send_message`／`subagent.prompt` 只能增加工作，而 Web composer 的 Stop 按钮被刻意限制在普通会话。人类眼看着一个 continuable child 在错误路径上烧 token，除了干掉整个 parent 树没有任何手段；当直接 parent Agent 离线时，即使 child 的 Activation 仍然在线，它也完全不可触及。一次性运行有持有方拥有的 disposal 和 task-kill；continuable child 没有对应的当前轮次控制。
## Decision
`ctx.subagents.interrupt(targetSessionId, authority)` 只停止在线目标的当前轮次。管理器原语同步完成鉴权，调用现有的 `Agent.cancel(cause, { keepInbox: true })`，然后返回 `void`——fire-and-return：保证取消信号已发出，但不等待目标静止。其余一切不变：不 dispose Activation、不释放 handle、不级联后代、不清空 inbox，也不改动 `AgentLoop` 或 `CancelOptions`。由于 `keepInbox` 让尚未领取的待处理队列停在 idle，中断绝不会自动启动下一个排队的 follow-up；已被领取进入中断轮次的工作属于该轮次，不会重新入队。被中断的 driver 进入 idle 后，一次显式唤醒发送会按保留的 FIFO 顺序恢复。
授权是一个封闭的双变体 union，刻意比投递权限更宽，因为停止一个轮次是幂等的且不投递任何内容：
- `{ kind: 'user', parentSessionId }`——人类出示持久化直接 parent 地址。在线目标的 `session.header.parentSession` 必须匹配；不涉及在线 parent Agent、目录读取或持久化访问，这正是 parent Agent 离线时在线 child 仍可被停止的原因。取消 cause 为 `user`。
- `{ kind: 'ancestor', agent }`——一个确切在线的 ancestor Agent（直接 parent 或更深）。调用方必须是注册表中其 id 的当前条目（过期调用方即使目标不存在也被拒绝），不得是目标本身，并且必须出现在 Activation 物化时记录的 `ancestry` WeakSet 中。取消 cause 为 `parent`。
目标只在管理器进程本地的 Activation map 中解析。不存在的 id——未知、一次性或已自然结算——是被接受的 no-op，统一覆盖完成竞态和重复请求而不泄露持久化目录信息；disposal 事务已打开的目标在鉴权后同样是被接受的 no-op。一次性生命周期（持有方 `dispose()`、task-kill）不受影响。`SubagentService.interrupt()` 把未绑定管理器的组合视为被接受的 no-op 而不是 `CONTINUATION_UNAVAILABLE`，因为没有管理器就不可能存在管理器拥有的在线 Activation。
Host RPC `subagent.interrupt` 接收 continuable 的 `SubagentAddress` 并返回 `{ accepted: true }`。它的实现只以 `user` 授权调用核心原语——刻意不调用 `catalogChild()`、`listChildren()`、`sessionQuery` 或 parent 注册表查找。parent 地址不匹配的在线目标映射为 `subagent-unauthorized`；意外失败映射为 `internal`，不把错误文本泄漏到 wire。
## Alternatives considered
**让人类中断走 `session.cancel`。** 通用会话取消要求附着的普通会话并拒绝 subagent 拥有的会话；放宽它会把 subagent 权限规则缠进普通会话路由。subagent 域的 RPC 让基于地址的鉴权和 parent 离线保证保持显式。
**等待目标静止并返回轮次结果。** 取消是协作式的，静止时间无上界；让 RPC（以及一个 `ChildLock` 槽位）保持打开会招致超时并与投递、disposal 形成排队。调用方需要的唯一事实是信号已被接受，而竞态（自然完成、disposal）本就幂等收敛。
**复用整个 Activation 的 disposal 来做中断。** disposal 的取消不带 `keepInbox`，还会 flush、capture 并释放 handle——它销毁排队工作和 child 的驻留。中断是针对一个轮次的控制操作，不是针对 Activation 的生命周期操作。
**顺手把 `send_message`／`followup` 权限扩展到 ancestor。** 投递向对话注入内容且不幂等；其确切直接 parent 权限保持不变。只有中断获得更宽的 ancestor 与基于地址的用户授权。
**中断后自动恢复被暂停的队列。** 在中止 A 后立即启动排队的 follow-up B 会让中断看起来被忽略，并夺走人类重新引导 child 的窗口。暂停到显式唤醒发送为止，让停止可观察且 FIFO 顺序完整。
## Consequences
人类或 ancestor 可以停止一个失控的 continuable 轮次，而不丢失 child、其尚未领取的排队工作或正在运行的后代；代价是一个刻意保持弱的后置条件（`accepted` 表示“信号已发出”，目标在观察到信号前可能仍显示 `running`），客户端必须如实呈现。暂停队列规则意味着被中断的 child 会带着保留的工作停在 idle，直到 driver 进入 idle 后收到唤醒消息——这是有意的 human-in-the-loop 暂停，不是调度器缺陷。在 abort 收敛期间被接受的唤醒发送目前会保持排队而不锁存 wake；Issue #1838 跟踪共享的 agent-loop 修正。
仅凭地址的 RPC 会暴露一位在线驻留信息：不存在的目标会被接受，而 parent 不匹配的在线目标会返回 `subagent-unauthorized`。单用户本地 Host 的信任模型接受这种可观察性；未来的多主体 Host 必须重新审视权限和响应不可区分性。
Web 侧复用现有的 primary Send/Stop 切换而不新增第二个操作：客户端 `Session.cancel()` 将 continuable 地址路由到 `subagent.interrupt`（one-shot 地址保持不可取消，普通会话仍走 `session.cancel`）；parent 离线但仍在运行的 continuable child 保留默认 composer 并禁用其输入区，让同一个 primary Stop 保持可达，停止后恢复只读替代（周边目录与 composer 契约由 [Web subagent 对话](2026-07-27-web-subagent-conversations.md)拥有）。
`dsh-tool-subagent-control` 中面向模型的 `interrupt_agent(agent_id)` 工具把 `exec.agent` 作为 `ancestor` 授权传入，自身不增加任何权限：核心原语校验在线注册表身份与记录的 lineage，因此该工具可以用同一个通用 `agent_id` 参数指定直接 child 或更深的后代——刻意不用会暗示仅限直接 child 的 `subagent_id`。发现依赖 `list_agents({ scope: 'descendants' })`，其底层是新的 `SubagentService.listDescendants()` 单次追踪 pre-order 遍历，每个条目带经校验的 `parentId`／`depth`（列表契约由[持久化目录 note](2026-07-22-durable-subagent-catalog-and-list-agents.md)拥有）；发现只是提示，绝非权限。`send_message` 保持其确切直接 parent 权限——只有中断是 ancestor 级的。
## Testing
`packages/subagent/subagent/tests/continuation.spec.ts` 中的核心覆盖证明了持久化 `turn/end` 中止、队列先暂停后按 FIFO 恢复、后代不受影响、两种授权及其取消 cause、self/sibling/stale/非 ancestor 拒绝、absent/一次性/disposal 竞态 no-op，以及 `keepInbox` 循环行为不变。`packages/host/apiproxy/tests` 中的 Host 覆盖证明 RPC 只调用核心原语（不读 agents/目录/历史）、`subagent-unauthorized`／`internal` 映射、wire schema 的 continuable 模式围栏以及 carrier 往返。客户端覆盖固定按地址路由的 `Session.cancel()`、InputBar 的 Send/Stop 切换及 parent 离线时锁定输入的状态，以及只读 composer selector 的运行例外；keyless 组装 Web 场景（`apps/web/tests/subagent-interrupt.e2e.ts`、`subagent-interrupt-ui.e2e.ts`）用 replay hang 条目保持真实 child 轮次打开，端到端证明中断传输、中止的 `turn/end`、follow-up 暂停以及 FIFO 恢复。`packages/subagent/tool-subagent-control/tests` 中的工具覆盖证明直接与更深 ancestor 以 `parent` cause 中断并暂停队列、self/sibling/陌生调用方被拒绝且不触碰目标、目标不存在时 no-op 且不冷恢复，以及 descendants 列表的 pre-order 位置；keyless ACP 快照把新工具 schema 固定在每个已录制请求 header 中。
.agents/notes/proposed/feature/2026-08-04-task-surface.i18n.yaml
+6
View File
@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/proposed/feature/2026-08-04-task-surface.md
2026-08-04-task-surface.md: 72fc9e12f1a02f66a02d3335ad2e8d0d4c1c3dc2
2026-08-04-task-surface.zh.md: 087c80195f75e145da93c5557afb43a421e1050d
.agents/notes/proposed/feature/2026-08-04-task-surface.md
+289
View File
@@ -0,0 +1,289 @@
# Agent Note: Task Surface for structured session interaction
Status: proposed
English | [中文](2026-08-04-task-surface.zh.md)
## Problem
Some tasks are awkward to finish through alternating prose messages. Comparing several options, reordering a plan, reviewing a table, or filling a small set of related fields all work better as one structured interaction. Today an agent can describe such an interaction, but it cannot ask the Web client to render one without adding a permanent product component or generating executable Client Plugin code.
Those two workarounds put ownership in the wrong place. Product-specific components require a new trigger and release for every task shape. Generated code has far more authority and lifecycle cost than a one-turn form needs. It also makes the presentation, rather than the user's conclusion, the durable artifact.
The missing contract is a bounded, replayable description of a temporary UI that belongs to one Session and one tool occurrence. The product should own validation, placement, interaction mechanics, and submission. The agent should own the task-specific copy, data, and choice of supported components.
## Proposal
Add **Task Surface**, a versioned declarative model rendered by a normal Web Client Plugin. One stable model-facing tool, `show_task_surface`, publishes the model. A successful call ends the current turn. The user edits and submits the rendered panel; the Host records the submission as one ordinary visible user message and starts the next turn.
Task Surface is the default structured-UI path when all of the following hold:
- the interaction belongs to the current Session and current task;
- its behavior fits the declared component set;
- it needs no background execution or new runtime authority; and
- the useful durable result is the user's submitted conclusion, not the panel itself.
This is one trigger, not a family of product heuristics. The agent calls `show_task_surface` explicitly. A user may ask the agent to use a Task Surface in ordinary language. Products do not inspect tool names or task topics to open bespoke panels, and repeated use does not automatically turn a Task Surface into a Plugin.
Short blocking questions remain with [`ask_user_question`](../../implemented/feature/2026-07-29-ask-question-web-presentation.md). Plain explanation remains chat. Cross-Session navigation, background behavior, new services, or durable custom UI belongs to the Generated Client Plugin workflow.
## Declarative model
`TaskSurfaceModelV1` is JSON. It contains content blocks, input fields, and one submit label; it contains no code, callbacks, selectors, HTML, CSS, URLs to executable assets, or expression language. This type is unrelated to core Session's existing `SurfaceManager`/`SurfaceOp` message-reduction types; Task Surface is a product interaction protocol.
```ts
interface TaskSurfaceModelV1 {
version: 1
title: string
description?: string
sections: TaskSurfaceSection[]
fields?: TaskSurfaceField[]
submit: { label: string }
}
interface TaskSurfaceSection {
id: string
title?: string
layout?: TaskSurfaceLayout
blocks: TaskSurfaceBlock[]
}
type TaskSurfaceLayout =
| { kind: 'stack' }
| { kind: 'grid'; columns: 2 | 3 }
type TaskSurfaceBlock =
| { kind: 'markdown'; text: string }
| { kind: 'metrics'; items: { label: string; value: string; detail?: string }[] }
| { kind: 'table'; columns: { id: string; label: string }[]; rows: Record<string, string | number | boolean | null>[] }
| { kind: 'diff'; path?: string; before: string | null; after: string; language?: string }
| { kind: 'notice'; tone: 'neutral' | 'info' | 'warning'; text: string }
type TaskSurfaceField =
| { kind: 'text'; id: string; label: string; multiline?: boolean; required?: boolean; initial?: string }
| { kind: 'choice'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string }
| { kind: 'multi-choice'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string[] }
| { kind: 'toggle'; id: string; label: string; initial?: boolean }
| { kind: 'order'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string[] }
interface TaskSurfaceOption { id: string; label: string; detail?: string }
```
The renderer controls typography, spacing, responsive layout, focus order, keyboard behavior, and theme tokens. An absent layout means `stack`; a `grid` layout owns its column count and collapses when the available width cannot support it. Unknown versions or union arms use the generic tool-result fallback instead of partial interpretation.
The `markdown` block reuses `MarkdownText` with an explicit model-URL policy. `MarkdownText` gains `remoteImages: 'render' | 'alt-only'`, preserving `render` as its ordinary default; Task Surface always passes `alt-only`, so image syntax renders only its alt text. Raw HTML and embedded media remain omitted, automatic link previews are absent, and no model-supplied URL is dereferenced without explicit user activation. Ordinary HTTP(S) links may still navigate when the user chooses them. Fixed application assets such as syntax-highlighting chunks remain under the product's normal loading policy.
Version 1 deliberately omits conditional fields, client-side data fetching, charts, file uploads, and arbitrary event handlers. A new block or field kind is a protocol change with a parser, renderer, accessibility behavior, fallback, and replay fixture in the same change.
Limits are schema-backed configuration on the Task Surface service. The initial defaults are 64 KiB for the normalized model, 64 blocks, 32 fields, 200 table rows, and 32 KiB for a submission. IDs are unique within the model; field values must match their declarations; unknown fields are rejected. The limits bound log, DOM, and prompt costs without changing the protocol.
## Tool and presentation contract
`show_task_surface` accepts `{ model: TaskSurfaceModelV1 }`. The Host parses and normalizes the complete model, rejects the call when that Session already has an open Task Surface, mints `surfaceId`, and returns canonical `{ surfaceId, model }` with the normalized model. `presentationMeta` persists `value.model`, so the projector and executor cannot disagree about normalization. The Native result names the Surface and explains that an ordinary message bypasses it when the client cannot render the panel. The tool then calls `exec.concludeTurn()` so the agent does not continue past the requested human checkpoint.
The tool definition omits `isConcurrencySafe`. Under the existing tool-registry contract, omission classifies every call as an exclusive ordering barrier; no new `ToolDefinition` field is introduced. The tool is composed only in Web profiles that mount both the Host service and Web renderer. Version 1 supports `native` and `both` tool modes; a `code`-only profile does not advertise it because Code Mode dispatch is nested and cannot carry its presentation metadata to the outer result.
The browser-safe domain package imports the type-only `Branded` primitive from `@deepseek-ai/dsh-brand` and owns all three Task Surface IDs. The canonical value is execution-local under the [canonical tool output contract](../../implemented/architecture/2026-07-20-canonical-tool-output-contract.md). Replay therefore uses `output.presentationMeta(args, value)` to persist this tagged payload with `tool/result.meta`:
```ts
import type { Branded } from '@deepseek-ai/dsh-brand'
type TaskSurfaceId = Branded<'TaskSurfaceId'>
type TaskSurfaceSubmissionId = Branded<'TaskSurfaceSubmissionId'>
type TaskSurfaceDismissalId = Branded<'TaskSurfaceDismissalId'>
interface TaskSurfacePresentationMeta {
kind: 'dsh/task-surface'
version: 1
surfaceId: TaskSurfaceId
model: TaskSurfaceModelV1
}
```
The tool keeps a generic [render intent](../../implemented/architecture/2026-07-02-tool-render-intent-union.md). The keyed Web row reads the tagged metadata already retained on `ToolResultNode`; no new render-intent arm or presentation registry is required. Clients without Task Surface support render the ordinary result content.
The Web plugin has two static Session-scoped registrations under the [toolview](../../implemented/architecture/2026-07-23-toolview-dissolution.md) and [slot registration](../../implemented/architecture/2026-07-22-slot-type-chain-implementation.md) contracts. A keyed `conversation.chat.toolview` entry for `show_task_surface` renders the durable transcript occurrence as a compact summary and read-only replay. One `TaskSurfaceDock` entry in the existing `conversation.input.dock` is the only actionable mount: it reads the active projection, calls `getActive` for the exact identity, and owns fields, drafts, submit, and dismiss. Because the Dock is independent of transcript pagination, an active Surface remains actionable when its `ToolResultNode` is outside the loaded history window.
The Dock follows the existing composer-chain fallback semantics. Any `conversation.composer` takeover hides the fallback composer stack, including `TaskSurfaceDock`, without unmounting it; the same draft owner reappears when the takeover resolves. A takeover does not receive Task Surface actions or create another editor.
The model does not choose a conversation tab, dock order, details column, modal, pixel position, or z-index. A later placement change remains a renderer decision and does not alter logged models. The transcript row never becomes a second editor, so one Surface cannot acquire competing draft or submission owners.
## Submission contract
The Task Surface domain exposes three operations through the Host transport. `submit` is the only one that admits a user message:
```ts ignore-check
type TaskSurfaceSubmissionPhase = 'queued' | 'claiming'
interface TaskSurfacePendingSubmission {
submissionId: TaskSurfaceSubmissionId
messageId: MessageId
phase: TaskSurfaceSubmissionPhase
}
interface TaskSurfaceService {
getActive(input: { sessionId: SessionId; surfaceId: TaskSurfaceId }): Promise<GetActiveTaskSurfaceResult>
submit(input: SubmitTaskSurfaceRequest): Promise<SubmitTaskSurfaceResult>
dismiss(input: DismissTaskSurfaceRequest): Promise<DismissTaskSurfaceResult>
}
interface SubmitTaskSurfaceRequest {
sessionId: SessionId
surfaceId: TaskSurfaceId
submissionId: TaskSurfaceSubmissionId
values: Record<string, JsonValue>
note?: string
}
type SubmitTaskSurfaceResult =
| { accepted: true; messageId: MessageId; phase: 'queued' }
| { accepted: false; reason: 'not-open' | 'stale' | 'invalid-submission' | 'submission-pending' }
type GetActiveTaskSurfaceResult =
| {
active: true
callId: CallId
surfaceId: TaskSurfaceId
model: TaskSurfaceModelV1
pending: TaskSurfacePendingSubmission | null
}
| { active: false; reason: 'not-open' }
interface DismissTaskSurfaceRequest {
sessionId: SessionId
surfaceId: TaskSurfaceId
dismissalId: TaskSurfaceDismissalId
}
type DismissTaskSurfaceResult =
| { dismissed: true; eventSeq: number }
| { dismissed: false; reason: 'not-open' | 'stale' | 'submission-pending' }
```
The Host resolves the exact successful `show_task_surface` occurrence, revalidates the submitted values against its persisted model, and admits the response through the normal Session queue. The response becomes a user-role message with a merge-extensible source:
```ts ignore-check
interface TaskSurfaceCorrelation {
version: 1
submissionId: TaskSurfaceSubmissionId
callId: CallId
surfaceId: TaskSurfaceId
values: Record<string, JsonValue>
}
interface TaskSurfaceUserMessageSource {
kind: 'user'
rpcId: RpcId
taskSurface: TaskSurfaceCorrelation
}
```
The `session/queue` wire item already carries the complete `Message`. The client projection is explicitly extended to retain its source instead of dropping the correlation:
```ts ignore-check
interface QueuedMessage {
id: InboxItemId
messageId: MessageId
placement: 'queued' | 'steering'
source: MessageSource
content: readonly ContentBlock[]
preview: string
text: string | null
}
```
The browser-safe domain package owns `TaskSurfaceId`, the submission and dismissal IDs, `TaskSurfaceCorrelation`, and the pending-submission shape. ApiProxy owns the transport augmentation that combines the correlation with `rpcId`. Keeping `kind: 'user'` preserves the ordinary user bubble and prompt semantics while the extra field provides durable correlation. The message content is a product-formatted readable summary: panel title, labels and submitted values, plus the optional note. The model receives that same text. The structured source is not a second hidden instruction.
The product shell owns collapse and dismiss. Collapse is local view state and sends nothing. When no submission is pending, `taskSurface.dismiss({ sessionId, surfaceId, dismissalId })` appends one `task-surface/dismissed` Session event and does not start a turn; the exact event closes the projection and updates the Dock and transcript row. Retries reuse `dismissalId` and return the original result without appending another event. Dismiss is disabled while a submission is `queued` or `claiming`, and the Host rejects such a request with `submission-pending`.
Submission is transactional at the client boundary. Acceptance returns the exact `messageId` in phase `queued`; the Dock disables every mutation through both `queued` and `claiming` and clears the persisted draft only after the matching user message becomes durable. A rejection keeps the values editable and shows the returned reason. Double clicks and transport retries reuse `submissionId` and return the first result; another submission ID receives `submission-pending` while the first is live. The Host admits one user message for one accepted Surface.
The Task Surface service records accepted submission coordination as `pending.phase: 'queued'`, while the client can correlate the still-present queue row through its retained `source`. When the Agent dequeues that occurrence for ordinary prompt admission, the service synchronously changes the same pending record to `claiming` before ApiProxy publishes the ordinary queue snapshot without the claimed row. The service keeps that process-local claim across asynchronous admission and reconnect until a matching durable `user/message` is published or the Agent reports a terminal discard.
The matching `user/message` closes the durable projection and clears the claim. Rejection, cancellation, or disposal before durability reports the discard, clears the claim, and leaves the Surface open. The Dock never interprets queue-row disappearance as either outcome: it re-reads `getActive`; `pending.phase: 'claiming'` stays disabled, `pending: null` restores the draft, and `not-open` closes the Dock. `getActive` joins the log-derived active occurrence with this one process-local pending record. The record is coordination state, not a second durable authority; after a Host restart, an uncommitted claim is absent and the still-open logged Surface becomes editable again.
`session.updateQueue` rejects `edit` and `steer` for a Task Surface-correlated row. Editing would separate formatted content from its source-carried structured values, and steering would persist a `steering/message` that does not satisfy the submission lifecycle. `remove` is allowed while the row is queued; it reports the discard and restores the open Surface. Once claimed, the row has left the generic queue and queue mutations return `queue-item-not-found`. The Task Surface service holds one single-flight pending record until commit or discard.
## Lifecycle and recovery
The Session log is the authority. A small `taskSurface` unit in the existing [Session projection system](../architecture/2026-07-27-session-projection-and-command-log.md) folds successful surface result metadata and later user-message sources into this state:
```ts ignore-check
interface TaskSurfaceProjection {
active: { callId: CallId; surfaceId: TaskSurfaceId } | null
}
```
One Session has at most one open Task Surface. A successful result opens it. A matching Task Surface user message or dismissal event closes it. A later ordinary user message also closes it as an explicit bypass; another `show_task_surface` call fails until one of those events closes the active occurrence. Rewind and fork derive their active occurrence by folding the resulting log; transient queue phase is not copied, and no separate Surface database participates.
The full model remains on its `tool/result.meta`; the projection carries only the active identity. `TaskSurfaceDock` exists independently of history rows and reacts to that identity. `taskSurface.getActive({ sessionId, surfaceId })` reads the exact occurrence from the Session log, revalidates its metadata, joins the Task Surface service's pending coordination record, and returns `{ callId, surfaceId, model, pending }`. A missing or closed occurrence returns `not-open`. Refresh and reconnect therefore recover an actionable Surface and its same-process pending phase even when the result is outside the history tail, without copying the model into every projection baseline.
The Web plugin keeps unsubmitted values in a bounded, per-Session persisted slot store keyed by `surfaceId`; they never enter the Session log, prompt, or long-term memory. Submitted values live in the accepted user message, so losing a browser draft cannot erase a conclusion.
## Package boundaries and dependencies
The capability is split where ownership changes:
| Package | Responsibility |
|---|---|
| `packages/core/agent` and `packages/core/agent-loop` | Generic terminal outcome for a claimed next-turn inbox occurrence, allowing a Host observer to distinguish durable admission from discard without Task Surface-specific types |
| `packages/task-surface/task-surface` | Browser-safe model, branded IDs, correlation and pending types, parser, limits, submission validator/formatter, Session event extension, projection unit, and Host service contract |
| `packages/task-surface/tool-task-surface` | `show_task_surface`, canonical output, presentation metadata, generic render intent, active-Surface check, and `concludeTurn()` behavior |
| `packages/client/runtime` | Generic queued-message `source` projection and Session-scoped active-projection access |
| `packages/client/ui-primitives` | Task Surface-agnostic `MarkdownText.remoteImages` policy, including the `alt-only` image branch and URL-policy tests |
| `packages/client/ui-task-surface` | Static actionable `TaskSurfaceDock`, read-only keyed transcript row, declarative Web renderer that consumes the Task Surface model and `MarkdownText` in `alt-only` mode, per-Session draft store, and submit client |
| `packages/host/apiproxy` | Typed active-read/submit/dismiss transport, user-source augmentation and carriage, queue-action restrictions, and routing of claim and terminal outcomes; delegates validation, pending coordination, and admission to the Task Surface service |
`ui-task-surface` depends on the browser-safe Task Surface domain, client connection and runtime, locale, `ui-conversation` for the declared slot contracts, `ui-slots` for registration, and `ui-primitives`; `ui-primitives` does not depend on Task Surface. ApiProxy depends on the Task Surface service contract and the generic AgentLoop terminal outcome. Core Agent packages do not import Task Surface types.
The implementation depends on the existing message log, canonical tool output, tagged render intents, Session projection, per-Session declared slot stores, and slot lifecycle. It does not depend on runtime Client Plugin creation. The generated Client Plugin workflow may use Task Surface to present a review form, but neither protocol owns or activates the other.
## Delivery stages
1. Land the model/parser, `MarkdownText` model-URL policy, projection unit, `show_task_surface`, presentation metadata, read-only Web row, static `TaskSurfaceDock`, active retrieval, and generic fallback with read-only blocks.
2. Add fields, persisted drafts, Host-validated submit/dismiss, branded correlation, client queued-source carriage, Task Surface `queued`/`claiming` coordination, claimed-occurrence terminal reporting, queue-action restrictions, and visible user-message admission.
3. Add only component kinds justified by real tasks and two consumers or a clear generic fallback. A separate explicit user action may start the generated Plugin authoring workflow, but it creates a candidate; it never promotes code directly.
## Alternatives considered
**Add product-specific triggers and panels.** Rejected because every new task shape would couple agent behavior to a shipped product component. Product code should define one admitted component vocabulary and placement policy; the agent chooses among it explicitly.
**Render arbitrary HTML, CSS, or JavaScript from the tool call.** Rejected because it turns a temporary interaction into executable Client Plugin code without the build, preview, evaluation, approval, or rollback lifecycle that code requires.
**Extend `userInteraction.ask()` with a large form.** Rejected for this contract. `ask()` is a blocking request/response operation used when a running tool cannot continue without a short answer. A Task Surface ends the turn, may remain open across refreshes, and submits its result as the next visible user turn.
**Register one dynamic `conversation.view` per call.** Rejected because the view ledger is global while its render scope is per Session, and because transient task identity would become registration identity. One static Session-scoped Dock owns interaction, and one static keyed row summarizes the logged occurrence; neither registration uses occurrence identity.
**Keep the model only in the canonical tool value.** Rejected because canonical values are not persisted. Replay requires the normalized model in `presentationMeta`.
**Store the panel in long-term memory.** Rejected because layout and draft state are not the reusable fact. Memory may retain the submitted user conclusion under existing memory policy.
## Acceptance criteria
- A real model in `native` or `both` mode can call one stable `show_task_surface` schema, the call ends its turn, and a capable Web client renders the same normalized model live and after replay; `code`-only mode does not advertise it.
- The static `TaskSurfaceDock` is the only editor and remains actionable for an active result outside the loaded history window; the keyed toolview remains a read-only transcript summary and replay. A composer takeover hides the still-mounted Dock, preserves its draft, and reveals the same owner after release.
- Submitting produces exactly one visible user message per `submissionId`, starts the next turn through normal queue admission, and retains exact branded occurrence correlation while keeping `source.kind: 'user'`; dismissing records one log event and starts no turn.
- The queued client row retains the correlated message source. `getActive` exposes `queued` or `claiming` across same-process reconnect; commit closes the projection, while explicit discard clears pending state and leaves the Surface open. Queue-row disappearance alone changes no UI state. Edit and steer are rejected, and remove succeeds only before claim.
- Refresh, reconnect, Session switching, fork, and rewind produce the lifecycle state implied by the log; `getActive` recovers the model and pending phase outside the history tail, and no panel, pending state, or draft leaks across Sessions.
- Unsupported versions, malformed metadata, and absent client capability fall back to readable tool-result content with the ordinary-message bypass; nested calls and calls made while another Surface is active fail without opening a Surface.
- Wire schemas validate ID strings and domain APIs expose the branded ID types throughout. The model parser enforces tagged layout shapes, field values, and configured byte/count limits before the panel becomes actionable. Browser tests show image syntax becomes alt text, raw HTML and embedded media do not render, and no model-supplied URL is requested before explicit user activation.
- Keyboard-only operation, focus restoration, accessible names, narrow layouts, both themes, and zh/en product chrome are covered by component tests.
- Keyless browser composition covers show, Dock and read-only-row ownership, off-window recovery, edit, retry after rejected admission, queued-to-claiming transition, discard, durable handoff without an editable gap, forbidden queue actions, dismiss, reconnect, and double-submit idempotency.
- Prefix snapshots show one stable tool definition regardless of the task-specific model; only the call arguments and later user conclusion vary.
- Unloading the Web plugin disposes its Dock, row, and draft stores through the owning Fiber without changing the durable transcript.
## Risks
The first component set may be either too small for useful tasks or broad enough to become a weak application framework. Usage evidence should decide additions; v1 has no expression language or network behavior.
The Task Surface Markdown policy gives up inline images, media, and automatic link previews. Ordinary links remain useful, but only an explicit user activation may navigate or start a request.
Large tables and Markdown can still create expensive DOM even inside byte limits. The renderer must virtualize or truncate where needed while preserving a readable fallback and explicit counts.
A product-formatted submission can become verbose when many fields are filled. The formatter needs a deterministic compact form and must preserve every submitted value without repeating the complete display model.
Holding a process-local claim until durable handoff adds a terminal-state invariant. Every admission exit must produce either the matching `user/message` or an explicit discard; otherwise a reconnect could retain a disabled Dock indefinitely.
Browser-local draft persistence can retain sensitive unsubmitted text. The store needs the stated byte bound, per-Session keys, explicit clearing after acceptance, and the same storage posture as the existing conversation draft.
The Dock and transcript row show the same occurrence in different roles. Keeping the row read-only and the Dock as the sole mutation owner prevents conflicting drafts at the cost of a second compact representation while the Surface is active.
.agents/notes/proposed/feature/2026-08-04-task-surface.zh.md
+289
View File
@@ -0,0 +1,289 @@
# Agent Note: 用于结构化会话交互的 Task Surface
Status: proposed
[English](2026-08-04-task-surface.md) | 中文
## 问题
有些任务很难通过交替发送文本消息来完成。比较多个选项、调整计划顺序、审阅表格，或填写一小组关联字段，都更适合在一次结构化交互中处理。目前，agent（智能体）可以描述这类交互，但若不增加永久的产品组件或生成可执行的客户端插件代码，就无法要求 Web 客户端渲染这类交互。
这两种变通方案的职责归属都不合理。产品专用组件要求每种任务形态都新增触发方式并发布新版本。对于只需一个轮次的表单，生成代码所拥有的权限和生命周期成本都远超实际需要。这样做还会把展示界面而非用户结论变成持久产物。
目前缺少这样一份契约：用有界、可回放的描述来定义临时 UI，并让它只属于一个会话和一次工具调用实例。产品应当负责校验、放置、交互机制和提交；agent 应当负责特定任务的文案、数据，以及从受支持组件中作出选择。
## 提案
新增 **Task Surface**：一种由普通 Web 客户端插件渲染、带版本的声明式模型。面向模型提供一个稳定工具 `show_task_surface`，用于发布该模型。调用成功后，当前轮次结束。用户编辑并提交渲染出的面板；Host 将提交内容记录为一条普通的可见用户消息，并开始下一轮。
同时满足以下条件时，Task Surface 是默认的结构化 UI 路径：
- 交互属于当前会话和当前任务；
- 行为可以由已声明的组件集合表达；
- 不需要后台执行或新增运行时权限；
- 有价值的持久结果是用户提交的结论，而不是面板本身。
这里定义的是一个触发方式，不是一组产品启发式规则。agent 会显式调用 `show_task_surface`。用户可以通过普通语言要求 agent 使用 Task Surface。产品不会根据工具名称或任务主题打开专用面板；重复使用也不会自动把 Task Surface 转为插件。
简短的阻塞式问题仍由 [`ask_user_question`](../../implemented/feature/2026-07-29-ask-question-web-presentation.md) 处理。纯文本说明仍留在聊天中。跨会话导航、后台行为、新服务或持久自定义 UI 则属于 Generated Client Plugin 工作流。
## 声明式模型
`TaskSurfaceModelV1` 使用 JSON。它包含内容块、输入字段和一个提交标签；不包含代码、回调、选择器、HTML、CSS、可执行产物的 URL，也不包含表达式语言。该类型与核心会话中现有的 `SurfaceManager`/`SurfaceOp` 消息归约类型无关；Task Surface 是一套产品交互协议。
```ts
interface TaskSurfaceModelV1 {
version: 1
title: string
description?: string
sections: TaskSurfaceSection[]
fields?: TaskSurfaceField[]
submit: { label: string }
}
interface TaskSurfaceSection {
id: string
title?: string
layout?: TaskSurfaceLayout
blocks: TaskSurfaceBlock[]
}
type TaskSurfaceLayout =
| { kind: 'stack' }
| { kind: 'grid'; columns: 2 | 3 }
type TaskSurfaceBlock =
| { kind: 'markdown'; text: string }
| { kind: 'metrics'; items: { label: string; value: string; detail?: string }[] }
| { kind: 'table'; columns: { id: string; label: string }[]; rows: Record<string, string | number | boolean | null>[] }
| { kind: 'diff'; path?: string; before: string | null; after: string; language?: string }
| { kind: 'notice'; tone: 'neutral' | 'info' | 'warning'; text: string }
type TaskSurfaceField =
| { kind: 'text'; id: string; label: string; multiline?: boolean; required?: boolean; initial?: string }
| { kind: 'choice'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string }
| { kind: 'multi-choice'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string[] }
| { kind: 'toggle'; id: string; label: string; initial?: boolean }
| { kind: 'order'; id: string; label: string; options: TaskSurfaceOption[]; initial?: string[] }
interface TaskSurfaceOption { id: string; label: string; detail?: string }
```
渲染器控制字体排印、间距、响应式布局、焦点顺序、键盘行为和主题 token。未指定布局时使用 `stack`；`grid` 布局自带列数，可用宽度无法容纳时会折叠。遇到未知版本或联合类型分支时，系统使用通用工具结果回退，而不是只解释其中一部分。
`markdown` 块复用 `MarkdownText`，并显式指定模型 URL 策略。`MarkdownText` 新增 `remoteImages: 'render' | 'alt-only'`，普通场景仍默认使用 `render`；Task Surface 始终传入 `alt-only`，因此图片语法只渲染替代文本。原始 HTML 和嵌入式媒体仍会被省略，不生成自动链接预览；未经用户显式操作，不会解引用模型提供的任何 URL。普通 HTTP(S) 链接仍可在用户选择后导航。语法高亮分片等固定应用资源继续遵循产品的常规加载策略。
版本 1 有意不支持条件字段、客户端数据获取、图表、文件上传和任意事件处理器。新增任何块或字段类型都属于协议变更，必须在同一变更中加入解析器、渲染器、无障碍行为、回退方式和回放 fixture（测试前置数据）。
Task Surface 服务通过受 schema 校验的配置定义限制。初始默认值为：规范化模型不超过 64 KiB、块不超过 64 个、字段不超过 32 个、表格行不超过 200 行、提交内容不超过 32 KiB。模型内的 ID 必须唯一；字段值必须符合其声明；未知字段会被拒绝。这些限制约束日志、DOM 和提示词成本，但不改变协议。
## 工具与呈现契约
`show_task_surface` 接收 `{ model: TaskSurfaceModelV1 }`。Host 解析并规范化完整模型；若该会话已有一个打开的 Task Surface，则拒绝调用；否则生成 `surfaceId`，并返回带规范化模型的规范值 `{ surfaceId, model }`。`presentationMeta` 持久化 `value.model`，使投影器和执行器不会对规范化结果产生分歧。Native 结果会指明该 Surface，并说明客户端无法渲染面板时，可以通过普通消息绕过它。随后工具调用 `exec.concludeTurn()`，防止 agent 越过所要求的人工检查点继续执行。
工具定义省略 `isConcurrencySafe`。根据现有工具注册表契约，省略该字段会将每次调用归类为独占排序屏障，无需新增 `ToolDefinition` 字段。该工具只会组装到同时挂载 Host 服务和 Web 渲染器的 Web profile 中。版本 1 支持 `native` 和 `both` 工具模式；仅支持 `code` 的 profile 不会向模型公布该工具，因为 Code Mode 分发属于嵌套调用，无法把呈现元数据传到外层结果。
浏览器安全的领域包从 `@deepseek-ai/dsh-brand` 以仅类型方式导入 `Branded` 原语，并拥有全部三个 Task Surface ID。根据[规范工具输出契约](../../implemented/architecture/2026-07-20-canonical-tool-output-contract.md)，规范值仅存在于本次执行中。因此，回放通过 `output.presentationMeta(args, value)` 将以下带标签的载荷随 `tool/result.meta` 一并持久化：
```ts
import type { Branded } from '@deepseek-ai/dsh-brand'
type TaskSurfaceId = Branded<'TaskSurfaceId'>
type TaskSurfaceSubmissionId = Branded<'TaskSurfaceSubmissionId'>
type TaskSurfaceDismissalId = Branded<'TaskSurfaceDismissalId'>
interface TaskSurfacePresentationMeta {
kind: 'dsh/task-surface'
version: 1
surfaceId: TaskSurfaceId
model: TaskSurfaceModelV1
}
```
该工具保留通用 [render intent](../../implemented/architecture/2026-07-02-tool-render-intent-union.md)。带 key 的 Web 行读取 `ToolResultNode` 上已经保留的带标签元数据，无需新增 render-intent 分支或呈现注册表。不支持 Task Surface 的客户端会渲染普通结果内容。
Web 插件按照 [toolview](../../implemented/architecture/2026-07-23-toolview-dissolution.md) 和 [slot 注册](../../implemented/architecture/2026-07-22-slot-type-chain-implementation.md)契约，提供两个静态的会话作用域注册项。一个以 `show_task_surface` 为 key 的 `conversation.chat.toolview` 条目将持久 transcript（文本记录）调用实例渲染为简洁摘要和只读回放。现有 `conversation.input.dock` 中的一个 `TaskSurfaceDock` 条目是唯一可操作的挂载点：它读取活动投影，针对确切身份调用 `getActive`，并拥有字段、草稿、提交和关闭操作。Dock 与 transcript 分页相互独立，因此即使 `ToolResultNode` 位于已加载历史窗口之外，活动 Surface 仍可操作。
Dock 遵循现有 composer chain 的回退语义。任何 `conversation.composer` 接管都会隐藏包括 `TaskSurfaceDock` 在内的回退 composer 栈，但不会将其卸载；接管结束后，同一个草稿所有者会重新出现。接管方不会获得 Task Surface 操作，也不会创建另一个编辑器。
模型不能选择会话标签页、Dock 顺序、详情栏、模态框、像素位置或 z-index。以后即使改变放置位置，也只是渲染器的决策，不会改变日志中记录的模型。transcript 行绝不会成为第二个编辑器，因此同一个 Surface 不会出现相互竞争的草稿或提交所有者。
## 提交契约
Task Surface 领域通过 Host 传输层公开三个操作。只有 `submit` 会接纳用户消息：
```ts ignore-check
type TaskSurfaceSubmissionPhase = 'queued' | 'claiming'
interface TaskSurfacePendingSubmission {
submissionId: TaskSurfaceSubmissionId
messageId: MessageId
phase: TaskSurfaceSubmissionPhase
}
interface TaskSurfaceService {
getActive(input: { sessionId: SessionId; surfaceId: TaskSurfaceId }): Promise<GetActiveTaskSurfaceResult>
submit(input: SubmitTaskSurfaceRequest): Promise<SubmitTaskSurfaceResult>
dismiss(input: DismissTaskSurfaceRequest): Promise<DismissTaskSurfaceResult>
}
interface SubmitTaskSurfaceRequest {
sessionId: SessionId
surfaceId: TaskSurfaceId
submissionId: TaskSurfaceSubmissionId
values: Record<string, JsonValue>
note?: string
}
type SubmitTaskSurfaceResult =
| { accepted: true; messageId: MessageId; phase: 'queued' }
| { accepted: false; reason: 'not-open' | 'stale' | 'invalid-submission' | 'submission-pending' }
type GetActiveTaskSurfaceResult =
| {
active: true
callId: CallId
surfaceId: TaskSurfaceId
model: TaskSurfaceModelV1
pending: TaskSurfacePendingSubmission | null
}
| { active: false; reason: 'not-open' }
interface DismissTaskSurfaceRequest {
sessionId: SessionId
surfaceId: TaskSurfaceId
dismissalId: TaskSurfaceDismissalId
}
type DismissTaskSurfaceResult =
| { dismissed: true; eventSeq: number }
| { dismissed: false; reason: 'not-open' | 'stale' | 'submission-pending' }
```
Host 解析出 `show_task_surface` 的确切成功调用实例，依据其已持久化模型重新校验提交值，并通过普通会话队列接纳响应。该响应成为一条用户角色消息，并使用可合并扩展的消息来源：
```ts ignore-check
interface TaskSurfaceCorrelation {
version: 1
submissionId: TaskSurfaceSubmissionId
callId: CallId
surfaceId: TaskSurfaceId
values: Record<string, JsonValue>
}
interface TaskSurfaceUserMessageSource {
kind: 'user'
rpcId: RpcId
taskSurface: TaskSurfaceCorrelation
}
```
`session/queue` 线上的条目已经携带完整 `Message`。客户端投影会显式扩展以保留其来源，不再丢失关联信息：
```ts ignore-check
interface QueuedMessage {
id: InboxItemId
messageId: MessageId
placement: 'queued' | 'steering'
source: MessageSource
content: readonly ContentBlock[]
preview: string
text: string | null
}
```
浏览器安全的领域包拥有 `TaskSurfaceId`、提交和关闭 ID、`TaskSurfaceCorrelation`，以及待处理提交的形态。ApiProxy 拥有传输扩展，负责将关联信息与 `rpcId` 组合。保留 `kind: 'user'` 可维持普通用户消息气泡和提示词语义，额外字段则提供持久关联信息。消息内容是由产品格式化的可读摘要，包括面板标题、标签和提交值，以及可选备注。模型接收相同的文本。结构化来源不是第二条隐藏指令。
产品外壳负责收起和关闭。收起属于本地视图状态，不会发送任何内容。没有待处理提交时，`taskSurface.dismiss({ sessionId, surfaceId, dismissalId })` 会追加一个 `task-surface/dismissed` 会话事件，但不启动轮次；该精确事件会关闭投影，并更新 Dock 和 transcript 行。重试会复用 `dismissalId` 并返回原始结果，不会再追加事件。提交处于 `queued` 或 `claiming` 阶段时，关闭操作会被禁用，Host 也会以 `submission-pending` 拒绝这类请求。
客户端边界上的提交具有事务性。接纳成功会返回处于 `queued` 阶段的确切 `messageId`；在 `queued` 和 `claiming` 两个阶段中，Dock 会禁用所有变更，并且只有匹配的用户消息持久化后，才会清除已持久化的草稿。若请求被拒绝，则保留值供用户继续编辑，并显示返回的原因。双击和传输重试会复用 `submissionId` 并返回第一次调用的结果；只要第一次提交仍在处理中，另一个提交 ID 就会收到 `submission-pending`。对于一个已接受的 Surface，Host 只会接纳一条用户消息。
Task Surface 服务将已接受提交的协调状态记录为 `pending.phase: 'queued'`，客户端则可通过仍在队列中的行所保留的 `source` 关联它。当 Agent 从队列取出该调用实例进行普通提示词接纳时，服务会先同步把同一份待处理记录改为 `claiming`，然后 ApiProxy 才发布不再包含已认领行的普通队列快照。服务会在异步接纳和重新连接期间一直保留这份进程内认领状态，直到匹配的持久 `user/message` 发布，或 Agent 报告终态丢弃。
匹配的 `user/message` 会关闭持久投影并清除认领状态。在持久化之前发生拒绝、取消或 dispose（资源释放）时，系统会报告丢弃、清除认领状态，并让 Surface 保持打开。Dock 绝不会把队列行消失解读为其中任一结果，而会重新读取 `getActive`：`pending.phase: 'claiming'` 会维持禁用状态，`pending: null` 会恢复草稿，`not-open` 会关闭 Dock。`getActive` 会把由日志推导的活动调用实例与这唯一一份进程内待处理记录合并。该记录属于协调状态，不是第二个持久权威来源；Host 重启后，未提交的认领状态不复存在，日志中仍然打开的 Surface 会恢复为可编辑状态。
对于带有 Task Surface 关联信息的行，`session.updateQueue` 会拒绝 `edit` 和 `steer`。编辑会让格式化内容与消息来源所携带的结构化值脱节，而 steering（中途引导）会持久化一条不符合提交生命周期的 `steering/message`。该行仍在队列中时允许 `remove`；它会报告丢弃并恢复为打开的 Surface。行被认领后即已离开通用队列，队列变更会返回 `queue-item-not-found`。Task Surface 服务会持有一份 single-flight 待处理记录，直至提交或丢弃。
## 生命周期与恢复
会话日志是真源。现有[会话投影系统](../architecture/2026-07-27-session-projection-and-command-log.md)中的一个小型 `taskSurface` 单元会折叠成功调用的 Surface 结果元数据和后续用户消息来源，得到以下状态：
```ts ignore-check
interface TaskSurfaceProjection {
active: { callId: CallId; surfaceId: TaskSurfaceId } | null
}
```
一个会话最多只能有一个打开的 Task Surface。成功的结果会打开它；匹配的 Task Surface 用户消息或关闭事件会将其关闭。后续的普通用户消息也会将其关闭，这是一条显式的绕过路径；在以上任一事件关闭活动调用实例前，再次调用 `show_task_surface` 都会失败。回退和 fork 会通过折叠相应日志推导出活动调用实例；瞬态队列阶段不会被复制，也不会有独立的 Surface 数据库参与其中。
完整模型仍存放在对应的 `tool/result.meta` 中；投影只携带活动身份。`TaskSurfaceDock` 独立于历史行存在，并会响应该身份。`taskSurface.getActive({ sessionId, surfaceId })` 会从会话日志中读取确切调用实例，重新校验其元数据，合并 Task Surface 服务的待处理协调记录，并返回 `{ callId, surfaceId, model, pending }`。调用实例不存在或已经关闭时返回 `not-open`。因此，即使结果位于历史尾段之外，刷新和重新连接仍能恢复可操作的 Surface 及其同进程待处理阶段，而无需把模型复制到每一个投影基线中。
Web 插件将未提交值保存在一个有界、按会话持久化的 slot store 中，并以 `surfaceId` 为 key；这些值永远不会进入会话日志、提示词或长期记忆。已提交值存放在接纳的用户消息中，因此即使浏览器草稿丢失，也不会抹去结论。
## 包边界与依赖
该能力在职责变化处拆分为多个包：
| 包 | 职责 |
|---|---|
| `packages/core/agent` 和 `packages/core/agent-loop` | 为已认领的下一轮 inbox 调用实例提供通用终态结果，让 Host 观察方无需使用 Task Surface 专用类型，即可区分持久接纳和丢弃 |
| `packages/task-surface/task-surface` | 浏览器安全的模型、带品牌类型的 ID、关联和待处理类型、解析器、限制、提交校验器／格式化器、会话事件扩展、投影单元，以及 Host 服务契约 |
| `packages/task-surface/tool-task-surface` | `show_task_surface`、规范输出、呈现元数据、通用 render intent、活动 Surface 检查和 `concludeTurn()` 行为 |
| `packages/client/runtime` | 通用排队消息 `source` 投影和会话作用域的活动投影访问 |
| `packages/client/ui-primitives` | 与 Task Surface 无关的 `MarkdownText.remoteImages` 策略，包括 `alt-only` 图片分支和 URL 策略测试 |
| `packages/client/ui-task-surface` | 静态且可操作的 `TaskSurfaceDock`、带 key 的只读 transcript 行、消费 Task Surface 模型并以 `alt-only` 模式使用 `MarkdownText` 的声明式 Web 渲染器、按会话划分的草稿 store，以及提交客户端 |
| `packages/host/apiproxy` | 类型化的活动 Surface 读取／提交／关闭传输、用户消息来源扩展与传递、队列操作限制，以及认领和终态结果的路由；将校验、待处理协调和接纳委托给 Task Surface 服务 |
`ui-task-surface` 依赖浏览器安全的 Task Surface 领域包、客户端连接与运行时、locale、`ui-conversation` 所声明的 slot 契约、用于注册的 `ui-slots`，以及 `ui-primitives`；`ui-primitives` 不反向依赖 Task Surface。ApiProxy 依赖 Task Surface 服务契约和通用 AgentLoop 终态结果。核心 Agent 包不导入 Task Surface 类型。
该实现依赖现有的消息日志、规范工具输出、带标签的 render intent、会话投影、按会话作用域声明的 slot store 和 slot 生命周期，不依赖在运行时创建客户端插件。Generated Client Plugin 工作流可以使用 Task Surface 展示审阅表单，但两个协议都不拥有或激活另一个协议。
## 交付阶段
1. 实现模型／解析器、`MarkdownText` 模型 URL 策略、投影单元、`show_task_surface`、呈现元数据、只读 Web 行、静态 `TaskSurfaceDock`、活动 Surface 读取，以及带只读块的通用回退。
2. 增加字段、持久化草稿、经 Host 校验的提交／关闭、带品牌类型的关联信息、客户端排队来源传递、Task Surface `queued`/`claiming` 协调、已认领调用实例的终态报告、队列操作限制，以及可见用户消息接纳。
3. 只增加有实际任务依据，并且拥有至少两个消费方或明确通用回退的组件类型。一个单独的显式用户操作可以启动生成式插件编写工作流，但只会创建候选项，绝不会直接推广代码。
## 考虑过的替代方案
**增加产品专用触发方式和面板。**不予采用，因为每种新任务形态都会把 agent 行为与已发布的产品组件耦合。产品代码应当定义一套接纳的组件词汇和放置策略；agent 则显式地从中选择。
**从工具调用中渲染任意 HTML、CSS 或 JavaScript。**不予采用，因为这会把临时交互变成可执行的客户端插件代码，却不具备代码所需的构建、预览、评估、批准或回滚生命周期。
**使用大型表单扩展 `userInteraction.ask()`。**本契约不采用这种做法。`ask()` 是一种阻塞式请求／响应操作，适用于正在运行的工具必须先获得简短答案才能继续执行的情况。Task Surface 会结束当前轮次，可以在刷新后继续保持打开，并把结果提交为下一条可见用户消息。
**每次调用都注册一个动态 `conversation.view`。**不予采用，因为视图账本是全局的，而其渲染作用域按会话划分；同时，临时任务身份会变成注册身份。一个静态的会话作用域 Dock 负责交互，一个静态带 key 的行概述已记录的调用实例；两个注册项都不使用调用实例身份。
**只在规范工具值中保留模型。**不予采用，因为规范值不会持久化。回放要求将规范化模型写入 `presentationMeta`。
**将面板存入长期记忆。**不予采用，因为布局和草稿状态不是可复用事实。现有记忆策略可以保留用户提交的结论。
## 验收标准
- 在 `native` 或 `both` 工具模式下，真实模型可以调用一个稳定的 `show_task_surface` schema；调用结束当前轮次；具备相应能力的 Web 客户端在实时运行和回放后都能渲染同一份规范化模型；仅支持 `code` 的模式不会向模型公布该工具。
- 静态 `TaskSurfaceDock` 是唯一的编辑器，即使活动结果位于已加载历史窗口之外也仍可操作；带 key 的 toolview 始终是 transcript 的只读摘要和回放。composer 接管会隐藏仍处于挂载状态的 Dock、保留其草稿，并在接管释放后重新显示同一个所有者。
- 每个 `submissionId` 的提交操作恰好生成一条可见用户消息，通过普通队列接纳开始下一轮，并在保留 `source.kind: 'user'` 的同时维持带品牌类型的确切调用实例关联；关闭操作记录一条日志事件，且不启动轮次。
- 客户端排队行保留已关联的消息来源。`getActive` 可在同一进程的重新连接前后公开 `queued` 或 `claiming`；提交会关闭投影，显式丢弃则会清除待处理状态并让 Surface 保持打开。队列行消失本身不会改变任何 UI 状态。系统会拒绝编辑和 steering，且移除操作只能在认领前成功。
- 刷新、重新连接、会话切换、fork 和回退都生成日志所决定的生命周期状态；`getActive` 可以恢复历史尾段之外的模型和待处理阶段，任何面板、待处理状态或草稿都不会泄漏到其他会话。
- 不受支持的版本、格式错误的元数据以及客户端能力缺失时，系统回退到带普通消息绕过路径的可读工具结果内容；嵌套调用以及已有另一个活动 Surface 时发起的调用都无法打开 Surface，并以失败结束。
- 线上的 schema 会校验 ID 字符串，领域 API 始终公开带品牌类型的 ID。模型解析器会在面板可交互前强制校验带标签的布局形态、字段值，以及配置的字节数和数量限制。浏览器测试证明：图片语法会变成替代文本，原始 HTML 和嵌入式媒体不会渲染，而且在用户显式操作前不会请求模型提供的 URL。
- 组件测试覆盖纯键盘操作、焦点恢复、无障碍名称、窄屏布局、两种主题，以及中英文产品界面。
- 无密钥浏览器组合测试覆盖显示、Dock 与只读行的职责归属、窗口外恢复、编辑、接纳被拒后的重试、从 `queued` 到 `claiming` 的转换、丢弃、没有可编辑空档的持久交接、禁止的队列操作、关闭、重新连接和双重提交幂等性。
- 前缀快照表明：无论任务特定模型如何变化，都只存在一个稳定的工具定义；只有调用参数和后续用户结论发生变化。
- 卸载 Web 插件时，其所属 Fiber 会对 Dock、工具行和草稿 store 执行 dispose，但不会改变持久 transcript。
## 风险
第一批组件可能小到无法满足实际任务，也可能大到足以演变成一个粗糙的应用框架。是否新增组件应由使用证据决定；v1 不提供表达式语言或网络行为。
Task Surface 的 Markdown 策略舍弃行内图片、媒体和自动链接预览。普通链接仍有用，但只有用户显式操作后，才可以导航或发起请求。
即使设置了字节限制，大型表格和 Markdown 仍可能生成开销较高的 DOM。渲染器必须按需虚拟化或截断内容，同时保留可读回退和明确计数。
填写字段较多时，由产品格式化的提交消息可能过长。格式化器需要使用确定性的紧凑格式，保留每一个提交值，同时避免重复完整显示模型。
在完成持久交接之前一直持有进程内认领状态，会新增一项终态不变量。每条接纳退出路径都必须产生匹配的 `user/message` 或显式丢弃，否则重新连接可能会让 Dock 永久处于禁用状态。
浏览器本地持久化的草稿可能保留敏感的未提交文本。store 需要遵守规定的字节上限、使用按会话划分的 key、在提交成功后显式清除，并采用与现有会话草稿相同的存储策略。
Dock 和 transcript 行以不同角色展示同一个调用实例。将工具行保持为只读，并让 Dock 成为唯一的变更所有者，可以避免草稿冲突，但代价是 Surface 活动期间会出现第二份简洁表示。
apps/web/tests/shipped-composition.e2e.ts
+1
View File
@@ -30,6 +30,7 @@ const EXPECTED_TOOLS = [
'edit',
'exit_plan_mode',
'get_goal',
'interrupt_agent',
'list_agents',
'ralph',
'read',
apps/web/tests/snapshots/subagent-interrupt/offline-composer.expected.md
+22
View File
@@ -0,0 +1,22 @@
- banner:
- navigation "Session hierarchy":
- button "Ask a research subagent to"
- text: /
- button "event-sourcing researcher" [disabled]
- tablist:
- tab "Chat" [selected]
- tab "Trajectory"
- text: Explain event sourcing in one sentence. {{clock}}
- button "Copy":
- img
- button "Context injection @deepseek-ai/dsh-system-prompt":
- img
- img
- text: Context injection @deepseek-ai/dsh-system-prompt
- paragraph: partial
- status: Deep diving...
- textbox "Parent session offline; sending is unavailable but you can still stop the run" [disabled]
- button "Commands" [disabled]:
- img
- 'button "Access mode, current: Workspace Write" [disabled]': Workspace Write
- button "Stop generating"
apps/web/tests/subagent-interrupt-ui.e2e.ts
+271
View File
@@ -0,0 +1,271 @@
// Web e2e scenario: the composer's primary action interrupts a running
// continuable child. The child holds its model turn open through a replay
// hang entry; the browser proves the single primary Send/Stop toggle, the
// parent-offline disabled-input-with-Stop composer, the subagent.interrupt
// (never session.cancel) transport, the parked follow-up, and the FIFO resume
// on a waking send.
//
// Replay-binding note: only the PRIMARY script can hang, and scripts bind by
// first-call order, so the child issues the composition's first model call
// (claiming the overridden primary) and the parent's one UI prompt — needed
// so the non-blank parent renders its header catalog — binds to a derived
// child fixture afterwards.
import { existsSync } from 'node:fs'
import { mkdtemp, readFile, rm, writeFile } from 'node:fs/promises'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import { fileURLToPath } from 'node:url'
import type { Browser, Page } from 'playwright'
import { chromium } from 'playwright'
import { afterAll, beforeAll, describe, expect, it, onTestFailed } from 'vitest'
import type { SessionEvent, SessionId } from '@deepseek-ai/dsh-session'
import type {} from '@deepseek-ai/dsh-agent'
import {
acknowledgeReloadConnectionLoss, assertFixtureInventory, captureStableAria, compareOrRefreshGolden,
launchWebScaffold, watchConsole, webSnapshotMode, type WebScaffold,
} from './scaffold.ts'
import { connectFreshWorkspace, newEnglishPage, saveFailureShot } from './support.ts'
const BASE_FIXTURE = fileURLToPath(new URL('./snapshots/live-interactions/session.jsonl', import.meta.url))
const SNAPSHOT_DIR = fileURLToPath(new URL('./snapshots/subagent-interrupt', import.meta.url))
const OFFLINE_COMPOSER_EXPECTED = join(SNAPSHOT_DIR, 'offline-composer.expected.md')
const MODE = webSnapshotMode()
const LABEL = 'event-sourcing researcher'
const INITIAL = 'Explain event sourcing in one sentence.'
const FOLLOWUP = 'Now give the same explanation to a human reader.'
const WAKING = 'And add one concrete example.'
const PARKED_ANSWER = 'parked follow-up answer'
const WAKING_ANSWER = 'waking answer'
/** Poll a synchronous condition (hook-safe; expect.poll is test-body only). */
async function waitFor(predicate: () => boolean, what: string, timeoutMs = 30_000): Promise<void> {
const deadline = Date.now() + timeoutMs
while (!predicate()) {
if (Date.now() >= deadline) throw new Error(`timed out waiting for ${what}`)
await new Promise<void>(resolve => setTimeout(resolve, 10))
}
}
/** One text-only scripted model completion (no tool calls: real tools are mounted). */
function textCompletion(text: string): object {
return {
kind: 'chunks',
chunks: [
{ type: 'block-start', index: 0, blockType: 'text' },
{ type: 'text-delta', index: 0, text },
{ type: 'block-end', index: 0, block: { type: 'text', text } },
{ type: 'usage', usage: { inputTokens: 20, outputTokens: 8 } },
{ type: 'finish', reason: { kind: 'stop' } },
],
}
}
describe.skipIf(MODE === 'record')('web e2e: composer interrupt for a running continuable child', () => {
let scaffold: WebScaffold
let browser: Browser
let page: Page
let sidecarRoot: string
let childId: SessionId
let tripwire: ReturnType<typeof watchConsole>
const apiCalls: string[] = []
beforeAll(async () => {
sidecarRoot = await mkdtemp(join(tmpdir(), 'dsh-web-subagent-interrupt-ui-'))
const readyFile = join(sidecarRoot, 'hang-ready')
// The child claims this whole-script replacement: held turn 1, then the
// parked follow-up and waking turns.
await writeFile(join(sidecarRoot, 'replay.override.json'), JSON.stringify([
{ kind: 'hang', readyFile },
textCompletion(PARKED_ANSWER),
textCompletion(WAKING_ANSWER),
]))
await writeFile(
join(sidecarRoot, 'session.jsonl'),
'{"type":"session","version":0,"id":"primary","createdAt":0}\n',
)
// The parent's one prompted turn replays this recorded single text-only
// call (binding is positional, not lineage-aware).
const parentTurnPath = join(sidecarRoot, 'parent-turn.jsonl')
const base = await readFile(BASE_FIXTURE, 'utf8')
const [header, ...events] = base.trimEnd().split('\n')
if (header === undefined) throw new Error('base replay fixture has no header')
await writeFile(parentTurnPath, [
header
.replace('"id":"{{sessionId}}"', '"id":"recorded-parent-turn"')
.replace(/"createdAt":\d+/, '"createdAt":1784998084442'),
...events,
'',
].join('\n'))
scaffold = await launchWebScaffold({
replayFixture: join(sidecarRoot, 'session.jsonl'),
replayOverride: join(sidecarRoot, 'replay.override.json'),
replayChildFixtures: [parentTurnPath],
})
browser = await chromium.launch()
page = await newEnglishPage(browser)
page.on('request', (request) => {
const path = new URL(request.url()).pathname
if (path.startsWith('/api/')) apiCalls.push(path)
})
tripwire = watchConsole(page)
await page.goto(scaffold.baseUrl, { waitUntil: 'load' })
await page.waitForSelector('[class*="frame"]', { timeout: 30_000 })
await connectFreshWorkspace(page, scaffold.workspaceCwd)
const parent = scaffold.ctx.agents.roots()[0]
if (parent === undefined) throw new Error('fresh workspace did not publish its parent Agent')
// The child's first model call claims the primary override and holds.
const started = await scaffold.ctx.subagents.startContinuable({
provider: 'spawn',
label: LABEL,
signal: new AbortController().signal,
request: { prompt: [{ type: 'text', text: INITIAL }], parent },
})
childId = started.childId
await waitFor(() => existsSync(readyFile), 'the held child turn to open')
// One prompted parent turn makes the parent non-blank so the session
// header (and its subagent catalog action) renders.
const parentSettled = scaffold.whenTurnSettled()
const parentInput = page.locator('textarea:enabled').first()
await parentInput.fill('Ask a research subagent to explain event sourcing.')
await parentInput.press('Enter')
expect(await parentSettled).toBe(parent.id)
// Reload onto the restart baseline (the proven route to a freshly
// discovered catalog), with the child still live and running host-side.
const warningStart = tripwire.warnings.length
await page.reload({ waitUntil: 'load' })
await page.waitForSelector('[class*="frame"]', { timeout: 30_000 })
await page.getByRole('button', { name: /1 subagent/ }).waitFor({ timeout: 15_000 })
acknowledgeReloadConnectionLoss(tripwire, warningStart)
expect(scaffold.ctx.agents.get(childId)?.status).toBe('running')
}, 120_000)
afterAll(async () => {
const failures: unknown[] = []
await browser?.close().catch((error: unknown) => failures.push(error))
await scaffold?.close().catch((error: unknown) => failures.push(error))
if (sidecarRoot !== undefined) {
await rm(sidecarRoot, { recursive: true, force: true })
.catch((error: unknown) => failures.push(error))
}
if (failures.length === 1) throw failures[0]
if (failures.length > 1) throw new AggregateError(failures, 'subagent interrupt UI teardown failed')
})
it('locks input but keeps the same primary Stop when the parent is offline', async () => {
onTestFailed(() => saveFailureShot(page, 'web-e2e-subagent-interrupt-offline'))
// Simulate a parent that went offline: the catalog delivers
// parentAvailable: false while the child Activation stays live (the
// interrupt RPC itself needs no live parent — PR 1's host coverage).
const pattern = '**/api/subagent.list'
await page.route(pattern, async (route) => {
const response = await route.fetch()
const body = await response.json() as {
result: { ok: true; value: { parentAvailable: boolean } } | { ok: false }
}
if (body.result.ok) body.result.value.parentAvailable = false
await route.fulfill({ response, json: body })
})
try {
await page.getByRole('button', { name: /1 subagent/ }).click()
await page.getByRole('treeitem', { name: new RegExp(LABEL) }).click()
const input = page.getByRole('textbox', {
name: 'Parent session offline; sending is unavailable but you can still stop the run',
})
await input.waitFor({ timeout: 15_000 })
expect(await input.isDisabled()).toBe(true)
// Still exactly one primary action, and it is an enabled Stop.
const stop = page.getByRole('button', { name: 'Stop generating' })
expect(await stop.count()).toBe(1)
expect(await stop.isEnabled()).toBe(true)
expect(await page.getByRole('button', { name: 'Send message' }).count()).toBe(0)
await compareOrRefreshGolden(
OFFLINE_COMPOSER_EXPECTED,
await captureStableAria(page, '[class*="centerCol"]', scaffold.workspaceCwd),
MODE,
)
} finally {
await page.unroute(pattern)
}
}, 60_000)
it('interrupts through subagent.interrupt, parks the follow-up, and resumes it FIFO', async () => {
onTestFailed(() => saveFailureShot(page, 'web-e2e-subagent-interrupt-flow'))
// Reselect the child with the truthful catalog: parent available again.
await page.getByRole('navigation', { name: 'Session hierarchy' })
.getByRole('button').first().click()
await page.getByRole('button', { name: /1 subagent/ }).click()
await page.getByRole('treeitem', { name: new RegExp(LABEL) }).click()
const input = page.getByRole('textbox', { name: 'Message the agent' })
await input.waitFor({ timeout: 15_000 })
expect(await input.isDisabled()).toBe(false)
// Queue a follow-up while the turn is open; the primary stays Stop.
const promptResponse = page.waitForResponse(response =>
new URL(response.url()).pathname === '/api/subagent.prompt')
await input.fill(FOLLOWUP)
await input.press('Enter')
expect(((await (await promptResponse).json()) as { result: { ok: boolean } }).result)
.toMatchObject({ ok: true })
const aborted = new Promise<void>((resolve, reject) => {
const timer = setTimeout(() => {
off()
reject(new Error('interrupt did not reach an aborted turn/end'))
}, 30_000)
const off = scaffold.ctx.on('session/event', (session: { id: SessionId }, event: SessionEvent) => {
if (session.id !== childId || event.type !== 'turn/end') return
clearTimeout(timer)
off()
if (event.data.reason.kind === 'aborted') resolve()
else reject(new Error(`expected an aborted turn/end, got ${event.data.reason.kind}`))
})
})
const stop = page.getByRole('button', { name: 'Stop generating' })
expect(await stop.count()).toBe(1)
const interruptResponse = page.waitForResponse(response =>
new URL(response.url()).pathname === '/api/subagent.interrupt')
await stop.click()
expect(((await (await interruptResponse).json()) as {
result: { ok: boolean; value?: { accepted: boolean } }
}).result).toMatchObject({ ok: true, value: { accepted: true } })
// The addressed child stops through its own RPC, never the generic one.
expect(apiCalls.filter(path => path === '/api/session.cancel')).toEqual([])
await aborted
// Parked: the Activation stays resident and idle with the retained
// follow-up; the primary returns to Send without a new turn starting.
await expect.poll(() => scaffold.ctx.agents.get(childId)?.status, { timeout: 15_000 }).toBe('idle')
const child = scaffold.ctx.agents.get(childId)
expect(child).toBeDefined()
expect(child!.inbox.nextTurn).toHaveLength(1)
expect(child!.session.events.filter(event => event.type === 'turn/start')).toHaveLength(1)
await page.getByRole('button', { name: 'Send message' }).waitFor({ timeout: 15_000 })
// Only the waking send resumes the parked queue, FIFO, to settlement.
await input.fill(WAKING)
await input.press('Enter')
await expect.poll(() => page.getByText(PARKED_ANSWER, { exact: true }).count(), { timeout: 30_000 }).toBe(1)
await expect.poll(() => page.getByText(WAKING_ANSWER, { exact: true }).count(), { timeout: 30_000 }).toBe(1)
await expect.poll(() => scaffold.ctx.agents.get(childId), { timeout: 60_000 }).toBeUndefined()
const loaded = await scaffold.ctx.sessionPersistence.load(childId)
const userTexts = loaded.events.flatMap(event => event.type === 'user/message'
&& event.data.source.kind === 'user'
? event.data.content.flatMap(block => block.type === 'text' ? [block.text] : [])
: [])
expect(userTexts).toEqual([INITIAL, FOLLOWUP, WAKING])
const turnEndKinds = loaded.events
.filter(event => event.type === 'turn/end')
.map(event => event.data.reason.kind)
expect(turnEndKinds).toEqual(['aborted', 'completed', 'completed'])
expect(tripwire.pageErrors).toEqual([])
}, 120_000)
it('keeps its snapshot inventory closed', async () => {
await assertFixtureInventory(SNAPSHOT_DIR, ['offline-composer.expected.md'])
})
})
apps/web/tests/subagent-interrupt.e2e.ts
+176
View File
@@ -0,0 +1,176 @@
// Web e2e scenario (browserless): the subagent.interrupt RPC against the real
// composition. A live continuable child holds its model turn open through a
// replay hang entry; plain HTTP queues a follow-up, interrupts the turn, and
// proves from the real session state that the turn aborted, the follow-up
// parked without auto-starting a new turn, and a later waking send resumed the
// preserved FIFO order. No browser: the RPC surface is the product surface
// under test, and PR-stacked UI coverage owns the composer interaction.
import { existsSync } from 'node:fs'
import { mkdtemp, rm, writeFile } from 'node:fs/promises'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import { afterAll, beforeAll, describe, expect, it } from 'vitest'
import { SessionId as sessionId, type SessionId } from '@deepseek-ai/dsh-session'
import type {} from '@deepseek-ai/dsh-agent'
import { launchWebScaffold, webSnapshotMode, type WebScaffold } from './scaffold.ts'
const MODE = webSnapshotMode()
const INITIAL = 'Explain event sourcing in one sentence.'
const FOLLOWUP = 'Now give the same explanation to a human reader.'
const WAKING = 'And add one concrete example.'
type RpcResult<T> = { ok: true; value: T } | { ok: false; error: { code: string; message: string } }
/** POST one unary RPC through the real HTTP carrier and unwrap its result. */
async function rpc<T>(baseUrl: string, method: string, payload: unknown): Promise<RpcResult<T>> {
const response = await fetch(`${baseUrl}/api/${method}`, {
method: 'POST',
headers: { 'content-type': 'application/json' },
body: JSON.stringify({
type: 'client-request',
rpcId: `interrupt-e2e-${method}-${crypto.randomUUID()}`,
method,
payload,
}),
})
if (!response.ok) throw new Error(`${method} failed over HTTP ${response.status}: ${await response.text()}`)
return (await response.json() as { result: RpcResult<T> }).result
}
/** Poll a synchronous condition (hook-safe; expect.poll is test-body only). */
async function waitFor(predicate: () => boolean, what: string, timeoutMs = 30_000): Promise<void> {
const deadline = Date.now() + timeoutMs
while (!predicate()) {
if (Date.now() >= deadline) throw new Error(`timed out waiting for ${what}`)
await new Promise<void>(resolve => setTimeout(resolve, 10))
}
}
/** One text-only scripted model completion (no tool calls: real tools are mounted). */
function textCompletion(text: string): object {
return {
kind: 'chunks',
chunks: [
{ type: 'block-start', index: 0, blockType: 'text' },
{ type: 'text-delta', index: 0, text },
{ type: 'block-end', index: 0, block: { type: 'text', text } },
{ type: 'usage', usage: { inputTokens: 20, outputTokens: 8 } },
{ type: 'finish', reason: { kind: 'stop' } },
],
}
}
describe.skipIf(MODE === 'record')('web e2e: subagent.interrupt over the real composition', () => {
let scaffold: WebScaffold
let sidecarRoot: string
let readyFile: string
let parentId: SessionId
let childId: SessionId
beforeAll(async () => {
sidecarRoot = await mkdtemp(join(tmpdir(), 'dsh-web-subagent-interrupt-'))
readyFile = join(sidecarRoot, 'hang-ready')
// Whole-script replacement: the child's three model calls are the hang
// (turn 1, interrupted), the parked follow-up's turn, and the waking turn.
// The parent never runs a turn, so the child claims this primary script.
await writeFile(join(sidecarRoot, 'replay.override.json'), JSON.stringify([
{ kind: 'hang', readyFile },
textCompletion('resumed response one'),
textCompletion('resumed response two'),
]))
// Header-only primary fixture: the bare-array override replaces the
// derived script entirely; the path only anchors replay installation.
await writeFile(
join(sidecarRoot, 'session.jsonl'),
'{"type":"session","version":0,"id":"primary","createdAt":0}\n',
)
scaffold = await launchWebScaffold({
replayFixture: join(sidecarRoot, 'session.jsonl'),
replayOverride: join(sidecarRoot, 'replay.override.json'),
})
// A live parent Agent through the real API; no workspace or browser.
const created = await rpc<{ sessionId: string }>(scaffold.baseUrl, 'session.create', {
cwd: scaffold.workspaceCwd,
})
if (!created.ok) throw new Error(`session.create failed: ${created.error.code}`)
parentId = sessionId(created.value.sessionId)
const parent = scaffold.ctx.agents.get(parentId)
if (parent === undefined) throw new Error('created parent session did not publish a live Agent')
const started = await scaffold.ctx.subagents.startContinuable({
provider: 'spawn',
label: 'event-sourcing researcher',
signal: new AbortController().signal,
request: { prompt: [{ type: 'text', text: INITIAL }], parent },
})
childId = started.childId
// The hang entry writes readyFile after its prefix chunks, immediately
// before waiting for cancellation: the deterministic "turn is open" gate.
await waitFor(() => existsSync(readyFile), 'the held child turn to open')
}, 120_000)
afterAll(async () => {
const failures: unknown[] = []
await scaffold?.close().catch((error: unknown) => failures.push(error))
await rm(sidecarRoot, { recursive: true, force: true }).catch((error: unknown) => failures.push(error))
if (failures.length === 1) throw failures[0]
if (failures.length > 1) throw new AggregateError(failures, 'subagent interrupt teardown failed')
})
it('parks a queued follow-up on interrupt and resumes it FIFO on a waking send', async () => {
// Queue the follow-up while the turn is still open, then interrupt.
const queued = await rpc<{ messageId: string }>(scaffold.baseUrl, 'subagent.prompt', {
parentSessionId: parentId,
childSessionId: childId,
mode: 'continuable',
content: [{ type: 'text', text: FOLLOWUP }],
})
expect(queued).toMatchObject({ ok: true })
const settled = scaffold.whenTurnSettled()
const interrupted = await rpc<{ accepted: true }>(scaffold.baseUrl, 'subagent.interrupt', {
parentSessionId: parentId,
childSessionId: childId,
mode: 'continuable',
})
expect(interrupted).toMatchObject({ ok: true, value: { accepted: true } })
// accepted acknowledges the admitted cancel, not quiescence: wait for the
// aborted turn/end (the composition's first turn/end) before asserting.
expect(await settled).toBe(childId)
// Parked, not resumed: the Activation stays resident with an idle driver,
// the follow-up is retained, and no second turn opened.
const child = scaffold.ctx.agents.get(childId)
expect(child).toBeDefined()
expect(child!.status).toBe('idle')
expect(child!.inbox.nextTurn).toHaveLength(1)
expect(child!.session.events.filter(event => event.type === 'turn/start')).toHaveLength(1)
const lastEnd = child!.session.events.filter(event => event.type === 'turn/end').at(-1)
expect((lastEnd)?.data.reason.kind).toBe('aborted')
// Only an explicit waking send resumes the parked queue, FIFO, then the
// child runs both turns to completion and settles.
const waking = await rpc<{ messageId: string }>(scaffold.baseUrl, 'subagent.prompt', {
parentSessionId: parentId,
childSessionId: childId,
mode: 'continuable',
content: [{ type: 'text', text: WAKING }],
})
expect(waking).toMatchObject({ ok: true })
await expect.poll(() => scaffold.ctx.agents.get(childId), { timeout: 60_000 }).toBeUndefined()
const loaded = await scaffold.ctx.sessionPersistence.load(childId)
// Human-origin messages only: the real composition also injects
// runtime-context snapshots as non-user-source messages.
const userTexts = loaded.events.flatMap(event => event.type === 'user/message'
&& event.data.source.kind === 'user'
? event.data.content.flatMap(block => block.type === 'text' ? [block.text] : [])
: [])
expect(userTexts).toEqual([INITIAL, FOLLOWUP, WAKING])
const turnEndKinds = loaded.events
.filter(event => event.type === 'turn/end')
.map(event => (event).data.reason.kind)
expect(turnEndKinds).toEqual(['aborted', 'completed', 'completed'])
}, 120_000)
})
apps/web/tsconfig.json
+2
View File
@@ -67,6 +67,8 @@
"tests/produced-file-mentions.e2e.ts",
"tests/goal-bar.e2e.ts",
"tests/subagent-conversation.e2e.ts",
"tests/subagent-interrupt.e2e.ts",
"tests/subagent-interrupt-ui.e2e.ts",
"tests/sidebar-subagent-activity.e2e.ts",
"tests/bash-abort-row.e2e.ts",
"tests/skill-tool-row.e2e.ts",
docs/cordis-catalog/events.md
+4 -4
View File
@@ -699,7 +699,7 @@ A published child settled. Scope-filtered dispatch uses the same delegating pare
Types: [Scoped](../core-data-structures/scope.md) · [SubagentService](../core-data-structures/subagent.md)
Source: [`packages/subagent/subagent/src/index.ts:160`](../../packages/subagent/subagent/src/index.ts)
Source: [`packages/subagent/subagent/src/index.ts:162`](../../packages/subagent/subagent/src/index.ts)
### `subagent/provider-added` — emit
@@ -716,7 +716,7 @@ A provider became resolvable in the registry.
Types: [SubagentProvider](../core-data-structures/subagent.md)
Source: [`packages/subagent/subagent/src/index.ts:134`](../../packages/subagent/subagent/src/index.ts)
Source: [`packages/subagent/subagent/src/index.ts:136`](../../packages/subagent/subagent/src/index.ts)
### `subagent/provider-removed` — emit
@@ -731,7 +731,7 @@ A provider left the registry. Accepted runs remain holder-owned.
'subagent/provider-removed'(name: string): void
```
Source: [`packages/subagent/subagent/src/index.ts:140`](../../packages/subagent/subagent/src/index.ts)
Source: [`packages/subagent/subagent/src/index.ts:142`](../../packages/subagent/subagent/src/index.ts)
### `subagent/start` — emit
@@ -753,7 +753,7 @@ A provider established a published child. For in-process providers, `ctx.agents.
Types: [Scoped](../core-data-structures/scope.md) · [SubagentService](../core-data-structures/subagent.md)
Source: [`packages/subagent/subagent/src/index.ts:151`](../../packages/subagent/subagent/src/index.ts)
Source: [`packages/subagent/subagent/src/index.ts:153`](../../packages/subagent/subagent/src/index.ts)
## `system-prompt/*`
docs/cordis-catalog/services.md
+36 -2
View File
@@ -2071,6 +2071,23 @@ async startContinuable(spec: ContinuableStartSpec): Promise<ContinuableStart>
*/
async followup( parent: Agent, childId: SessionId, content: ContentBlock[], options: SubagentFollowupOptions, ): Promise<MessageId>
/**
* Interrupt one live continuable child's current turn under a human parent
* address or an exact live ancestor Agent. Fire-and-return: the cancel
* signal is issued before this returns, but the target may keep running
* until it observes the signal. Unclaimed pending inbox work, the Activation,
* and published descendants are preserved; claimed work is not requeued.
* Once the interrupted driver is idle, a waking send resumes the parked FIFO
* queue. An absent target — including a one-shot or unknown id —
* is an accepted no-op, as is a manager-less composition, which cannot own a
* live Activation.
* @param targetSessionId - the durable child session id to interrupt.
* @param authority - the human parent address or exact live ancestor Agent.
* @throws {SubagentError} `UNAUTHORIZED` when the authority does not own the
* live target.
*/
interrupt(targetSessionId: SessionId, authority: SubagentInterruptAuthority): void
/**
* Deliver selected content from one live continuable child to its durable
* direct parent. The child is the authority credential; callers cannot name a
@@ -2136,6 +2153,23 @@ async drainContinuableDescendants(parents: readonly Agent[]): Promise<void>
*/
listChildren(parentSessionId: SessionId, signal?: AbortSignal): Promise<SubagentListEntry[]>
/**
* Enumerate the root's complete session-backed subagent tree in stable
* pre-order from one live-preferred corpus, without loading or resuming an
* Agent. Ordinary sessions and one-shot children remain traversal nodes so
* continuable descendants below them are discovered; each returned entry
* adds its durable `parentId` and root-relative `depth`. Identity resolution,
* diagnostics, optional persistence, and cancellation follow the same
* projection-backed contract as {@link listChildren}.
* @param rootSessionId - session whose complete descendant tree is listed.
* @param signal - caller-owned cancellation forwarded to persistence reads
* and observed around every read await.
* @returns children and per-candidate diagnostics with tree position, in
* stable pre-order.
* @throws {@link SubagentError} under the same conditions as {@link listChildren}.
*/
listDescendants(rootSessionId: SessionId, signal?: AbortSignal): Promise<SubagentDescendantListEntry[]>
/**
* Register a provider under its name. Registration is effect-scoped and HMR
* safe; removing a provider blocks new starts but does not revoke runs that
@@ -2171,9 +2205,9 @@ list(): string[]
async start(name: string, request: SubagentStartRequest): Promise<SubagentRun>
```
Types: [Agent](../core-data-structures/core.md) · [ContentBlock](../core-data-structures/core.md) · [ContinuableSetupContribution](../core-data-structures/subagent.md) · [ContinuableStart](../core-data-structures/subagent.md) · [ContinuableStartSpec](../core-data-structures/subagent.md) · [MessageId](../core-data-structures/core.md) · [SessionId](../core-data-structures/core.md) · [SubagentFollowupOptions](../core-data-structures/subagent.md) · [SubagentListEntry](../core-data-structures/subagent.md) · [SubagentProvider](../core-data-structures/subagent.md) · [SubagentReportOptions](../core-data-structures/subagent.md) · [SubagentRun](../core-data-structures/subagent.md) · [SubagentStartRequest](../core-data-structures/subagent.md)
Types: [Agent](../core-data-structures/core.md) · [ContentBlock](../core-data-structures/core.md) · [ContinuableSetupContribution](../core-data-structures/subagent.md) · [ContinuableStart](../core-data-structures/subagent.md) · [ContinuableStartSpec](../core-data-structures/subagent.md) · [MessageId](../core-data-structures/core.md) · [SessionId](../core-data-structures/core.md) · [SubagentDescendantListEntry](../core-data-structures/subagent.md) · [SubagentFollowupOptions](../core-data-structures/subagent.md) · [SubagentInterruptAuthority](../core-data-structures/subagent.md) · [SubagentListEntry](../core-data-structures/subagent.md) · [SubagentProvider](../core-data-structures/subagent.md) · [SubagentReportOptions](../core-data-structures/subagent.md) · [SubagentRun](../core-data-structures/subagent.md) · [SubagentStartRequest](../core-data-structures/subagent.md)
Source: [`packages/subagent/subagent/src/index.ts:165`](../../packages/subagent/subagent/src/index.ts)
Source: [`packages/subagent/subagent/src/index.ts:167`](../../packages/subagent/subagent/src/index.ts)
## `ctx.subprocess` — `SubprocessService` (abstract seam)
docs/core-data-structures/subagent.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write docs/core-data-structures/subagent.md
subagent.md: 4d7552eb5749284d90e31e62d8ac02e3d7a21b1d
subagent.zh.md: ebcb1ce445c164352109d6613028c7a36c10d6d4
subagent.md: b26a12d1d50305d86d7ada29cac83474009d81ce
subagent.zh.md: 6c4c64ff22050b73699a97093acd0032668fde3d
docs/core-data-structures/subagent.md
+34 -4
View File
@@ -4,7 +4,7 @@ English | [中文](subagent.zh.md)
The subagent seam — an agent delegating work to a child agent. Like [bash](bash.md) it is **one optional capability**, not part of the agent-loop spine, so its vocabulary lives here rather than in [core.md](core.md). But it differs from every other seam on one axis: **multiple provider implementations coexist** in one context, registered by name (`ctx.subagents`), where bash allows only one executor. The registry shape mirrors the [LLM adapter registry](llm-streaming.md), not the single-service bash executor.
Interface: [dsh-subagent](../../packages/subagent/subagent) (`ctx.subagents` + the vocabulary below). Implementations are sibling packages (`dsh-subagent-spawn`, `-fork`, `-acp`, `-codex`, `-claude-code`, `-dsh-sdk`); the model-facing consumers are [dsh-tool-subagent](../../packages/subagent/tool-subagent) (per-provider delegation), [dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control) (the optional global `send_message` and `list_agents` controls), and [dsh-tool-subagent-report](../../packages/subagent/tool-subagent-report) (the optional child-scoped `report` return channel). The same `ctx.subagents` service owns continuable-child orchestration through an internal activation manager and read-only direct-child discovery straight from the session store and optional session persistence. Product-provider rationale lives in [the Codex and Claude Code Agent Note](../../.agents/notes/implemented/feature/2026-08-04-claude-code-and-codex-subagent-backends.md); common-seam rationale lives in [the subagent Agent Note](../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md), [the continuable subagents Agent Note](../../.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md), [the report-tool Agent Note](../../.agents/notes/implemented/feature/2026-07-30-continuable-subagent-report-tool.md), [the durable catalog Agent Note](../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md), [the list-identity-projection Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md), and [the merged-service Agent Note](../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md).
Interface: [dsh-subagent](../../packages/subagent/subagent) (`ctx.subagents` + the vocabulary below). Implementations are sibling packages (`dsh-subagent-spawn`, `-fork`, `-acp`, `-codex`, `-claude-code`, `-dsh-sdk`); the model-facing consumers are [dsh-tool-subagent](../../packages/subagent/tool-subagent) (per-provider delegation), [dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control) (the optional global `send_message`, `interrupt_agent`, and `list_agents` controls), and [dsh-tool-subagent-report](../../packages/subagent/tool-subagent-report) (the optional child-scoped `report` return channel). The same `ctx.subagents` service owns continuable-child orchestration through an internal activation manager and read-only child and descendant discovery straight from the session store and optional session persistence. Product-provider rationale lives in [the Codex and Claude Code Agent Note](../../.agents/notes/implemented/feature/2026-08-04-claude-code-and-codex-subagent-backends.md); common-seam rationale lives in [the subagent Agent Note](../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md), [the continuable subagents Agent Note](../../.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md), [the report-tool Agent Note](../../.agents/notes/implemented/feature/2026-07-30-continuable-subagent-report-tool.md), [the durable catalog Agent Note](../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md), [the list-identity-projection Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md), and [the merged-service Agent Note](../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md).
Sources: [`packages/subagent/subagent/src/types.ts`](../../packages/subagent/subagent/src/types.ts), [`packages/subagent/subagent/src/index.ts`](../../packages/subagent/subagent/src/index.ts), and [`packages/subagent/subagent/src/continuation.ts`](../../packages/subagent/subagent/src/continuation.ts)
@@ -139,7 +139,20 @@ The Agent inbox is the only queue. Every continuation message becomes one `Agent
Follow-up authority comes from an exact live Agent tool context. The authenticated Agent must be the durable child's direct parent recorded in `SessionHeader.parentSession`. `MessageSource` and `senderSessionId` are durable provenance after admission and grant no authority; the optional model-facing tool uses `CoordinatorMessageSource`.
For both operations the caller signal owns lookup, materialization, and admission only until inbox acceptance. Afterwards the manager owns the Activation independently: later caller cancellation neither cancels the accepted turn nor disposes the child, and the seam exposes no public subagent cancellation or steering operation.
For both operations the caller signal owns lookup, materialization, and admission only until inbox acceptance. Afterwards the manager owns the Activation independently: later caller cancellation neither cancels the accepted turn nor disposes the child, and the seam exposes no steering operation.
`SubagentService.interrupt(targetSessionId, authority)` is the one public stop: it authorizes synchronously, issues `Agent.cancel(cause, { keepInbox: true })` on the live target, and returns without awaiting quiescence. The Activation, its unclaimed pending inbox work, and published descendants are untouched; work already claimed into the interrupted turn is not requeued. Once the interrupted driver is idle, a waking send resumes the parked FIFO queue. An absent target — unknown, one-shot, or already settled — and a manager-less composition are accepted no-ops. For a live target, a mismatched parent address or caller outside its live ancestry rejects with `UNAUTHORIZED`; stale ancestor objects and self-targeting ancestor requests reject before target lookup.
```ts type-equiv
/**
* Authority under which one interrupt request is admitted. `user` carries the
* durable direct-parent address a human client presented; `ancestor` carries
* the exact live Agent object whose recorded lineage must contain the caller.
*/
type SubagentInterruptAuthority =
| { readonly kind: 'user'; readonly parentSessionId: SessionId }
| { readonly kind: 'ancestor'; readonly agent: Agent }
```
Every Activation owns its `AgentHandle` and an `ownedChildren: Set<SessionId>`; because one Session has at most one live Activation, the child Session id identifies the live child without another runtime-incarnation reference. Starting a child or submitting parent-originated work registers the child in a continuation-managed parent's set before the child can run, and that parent cannot settle while the set is non-empty. A top-level or other non-continuation Agent has no Activation and stays outside the waiting graph. Child release happens only after the child Agent is quiescent, every child of that child is disposed, the best-effort final session flush settles, and the child's `AgentHandle` completes disposal.
@@ -250,9 +263,26 @@ The descriptor (`SubagentDescriptorData` in [descriptor.ts](../../packages/subag
A local one-shot provider appends the descriptor inside the child's initial turn before its first request. The continuation manager appends the descriptor after any provider-supplied lineage and before the initial prompt is admitted; `header.seedLength` remains the fork-lineage boundary: resume-time descriptor authority reads the child's own suffix, while the list-serving identity projection folds `subagent/descriptor` last-wins so the child's own descriptor overrides a fork-seeded ancestor's. The event is log-only: no `surfaceOp`, never in model history, and retained across compaction by the append-only log. Malformed current-version descriptors are corrupt; unsupported versions cannot be classified by this runtime.
## Durable enumeration: `listChildren()` and `SubagentListEntry`
## Durable enumeration: `listChildren()`, `listDescendants()`, and their entries
`SubagentService.listChildren(parentSessionId)` enumerates the parent's direct session-backed subagents from the live-preferred merge of `ctx.sessions.list()` and optional `ctx.sessionPersistence.list()` — no query seam, and no Agent is loaded or resumed. Candidates are the direct children whose durable header carries `origin: 'subagent'`; the marker classifies enumeration and coarse generic-route denial but cannot establish a valid descriptor, resumability, or authorization — the projection fold owns identity, and the Activation contract owns resume. Each row's `mode`/`label` is the registered `subagent` projection unit's value, served through a three-rung ladder: the registry's watermark cache for a live child (zero log reads); the optional projection checkpoint cache for a cold one (`cachedSnapshot` — an identity passing the own-suffix seq gate is final, because an own descriptor is immutable once appended); otherwise one `persistence.inspect()` reading folded through the registry (bounded concurrency, recomputed per listing). The cache is a pure optional accelerator: absent, serving the `null` sentinel or missing the key, failing the seq gate, or faulting, it falls silently through to the authoritative refold. The fold is `subagent/descriptor` last-wins with no failure channel: the child's own descriptor overrides a fork-seeded ancestor's, and a malformed or unknown-version payload folds to a serializable `null` sentinel, treated as no value. The result is one `SubagentListEntry[]` in `createdAt`-then-id order: a served identity yields a `child` entry with `mode: 'one-shot' | 'continuable'` and `activity: 'running' | 'inactive'`; continuable entries always carry `label`, while one-shot entries carry it only when the start caller supplied presentation metadata. A settled candidate whose fold served no identity yields a `corrupt` diagnostic — missing, malformed, and unknown-version descriptors deliberately undistinguished, with `unsupported` kept in the type for consumers already routing on it but no longer produced; a running candidate without an identity is omitted (the creation window before its descriptor lands); a failed cold inspection yields one `unavailable` diagnostic retried on the next listing, so one damaged sibling cannot hide healthy children. `hasChildren` marks a direct descendant with durable subagent origin, read from the same merged material. Activity snapshots only whether the logical record is live in `ctx.sessions`, not outcome or resumability. Absent persistence, enumeration is live-only rather than an error — a cold child cannot be resumed then either. `listChildren()` throws `SubagentError` with code `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` when the `ctx.sessionProjections` registry is absent and `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` when the session store is, both checked before any read so a deployment with zero children still fails deterministically; the list tool requires `ctx.subagents` and `ctx.agents` at plugin load. A service consumer such as a UI can display both modes and choose an unlabeled one-shot fallback, while the model-facing `list_agents` adapter (the separately loadable `/list-agents` plugin of [dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control)) keeps only continuable entries and refines status through the live Agent registry's `running`/`idle`/`complete` vocabulary. Listing does not consult the continuation manager's Activation map, Agent registry, or provider availability; `send_message` remains the authoritative delivery-time operation, and a listed running continuable child may still reject delivery as an ownership conflict. The read-path rationale lives in [the list-identity-projection Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md).
`SubagentService.listDescendants(rootSessionId)` applies the same live-preferred corpus and projection-backed interpretation to the root's complete descendant tree in stable pre-order. Ordinary sessions and one-shot children remain traversal nodes, so continuable descendants below them are discovered; only `origin: 'subagent'` candidates produce rows. Each returned child or diagnostic adds its position from the enumerated durable header, while a cold inspection revalidates that complete lifecycle before serving identity:
```ts type-equiv
/**
* One entry of a descendant listing: the interpreted subagent facts plus its
* position in the complete session tree. `parentId` is the durable direct
* parent from the enumerated header, and `depth` counts edges from the root.
*/
type SubagentDescendantListEntry = SubagentListEntry & {
/** Durable direct parent of this candidate in the enumerated tree. */
readonly parentId: SessionId
/** Edge distance from the requested root; direct children are `1`. */
readonly depth: number
}
```
`SubagentService.listChildren(parentSessionId)` enumerates the parent's direct session-backed subagents from the live-preferred merge of `ctx.sessions.list()` and optional `ctx.sessionPersistence.list()` — no query seam, and no Agent is loaded or resumed. Candidates are the direct children whose durable header carries `origin: 'subagent'`; the marker classifies enumeration and coarse generic-route denial but cannot establish a valid descriptor, resumability, or authorization — the projection fold owns identity, and the Activation contract owns resume. Each row's `mode`/`label` is the registered `subagent` projection unit's value, served through a three-rung ladder: the registry's watermark cache for a live child (zero log reads); the optional projection checkpoint cache for a cold one (`cachedSnapshot` — an identity passing the own-suffix seq gate is final, because an own descriptor is immutable once appended); otherwise one `persistence.inspect()` reading folded through the registry (bounded concurrency, recomputed per listing). The cache is a pure optional accelerator: absent, serving the `null` sentinel or missing the key, failing the seq gate, or faulting, it falls silently through to the authoritative refold. The fold is `subagent/descriptor` last-wins with no failure channel: the child's own descriptor overrides a fork-seeded ancestor's, and a malformed or unknown-version payload folds to a serializable `null` sentinel, treated as no value. The result is one `SubagentListEntry[]` in `createdAt`-then-id order: a served identity yields a `child` entry with `mode: 'one-shot' | 'continuable'` and `activity: 'running' | 'inactive'`; continuable entries always carry `label`, while one-shot entries carry it only when the start caller supplied presentation metadata. A settled candidate whose fold served no identity yields a `corrupt` diagnostic — missing, malformed, and unknown-version descriptors deliberately undistinguished, with `unsupported` kept in the type for consumers already routing on it but no longer produced; a running candidate without an identity is omitted (the creation window before its descriptor lands); a failed cold inspection yields one `unavailable` diagnostic retried on the next listing, so one damaged sibling cannot hide healthy children. `hasChildren` marks a direct descendant with durable subagent origin, read from the same merged material. Activity snapshots only whether the logical record is live in `ctx.sessions`, not outcome or resumability. Absent persistence, enumeration is live-only rather than an error — a cold child cannot be resumed then either. `listChildren()` throws `SubagentError` with code `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` when the `ctx.sessionProjections` registry is absent and `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` when the session store is, both checked before any read so a deployment with zero children still fails deterministically; the list tool requires `ctx.subagents` at plugin load. A service consumer such as a UI can display both modes and choose an unlabeled one-shot fallback, while the model-facing `list_agents` adapter (the separately loadable `/list-agents` plugin of [dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control)) keeps only continuable entries and maps activity to its existing `running`/`complete` vocabulary. Listing does not consult the continuation manager's Activation map, Agent registry, or provider availability; `send_message` remains the authoritative delivery-time operation, and a listed running continuable child may still reject delivery as an ownership conflict. The read-path rationale lives in [the list-identity-projection Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md).
## The terminal result: `SubagentResult`
docs/core-data-structures/subagent.zh.md
+34 -4
View File
@@ -4,7 +4,7 @@
subagent seam：一个 agent（智能体）将工作委派给子 agent。与 [bash](bash.md) 一样，它是**一项可选能力**，不属于 agent loop（智能体循环）主干，因此其词汇定义在此而非 [core.md](core.md) 中。但它在一个维度上与其他所有 seam 不同：**同一上下文中可共存多个提供方实现**，按名称注册（`ctx.subagents`），而 bash 只允许一个执行器。注册表的形状参照 [LLM（大语言模型）适配器注册表](llm-streaming.md)，而非单服务的 bash 执行器。
接口：[dsh-subagent](../../packages/subagent/subagent)（`ctx.subagents` + 下文词汇）。实现为六个兄弟包（package）：`dsh-subagent-spawn`、`-fork`、`-acp`、`-codex`、`-claude-code`、`-dsh-sdk`；面向模型的消费方包括 [dsh-tool-subagent](../../packages/subagent/tool-subagent)（按提供方委派）、[dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control)（可选的全局 `send_message` 与 `list_agents` 控制工具）和 [dsh-tool-subagent-report](../../packages/subagent/tool-subagent-report)（可选的 child 作用域 `report` 返回通道）。同一个 `ctx.subagents` 服务通过内部激活管理器负责可继续子 agent 编排，并直接从会话存储与可选的会话持久化负责只读的直接 child 发现。产品提供方设计理由见 [Codex 与 Claude Code Agent Note（agent 决策记录）](../../.agents/notes/implemented/feature/2026-08-04-claude-code-and-codex-subagent-backends.md)；通用 seam 的设计理由见 [subagent Agent Note](../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md)、[可继续 subagent Agent Note](../../.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md)、[report 工具 Agent Note](../../.agents/notes/implemented/feature/2026-07-30-continuable-subagent-report-tool.md)、[持久化目录 Agent Note](../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md)、[列表身份投影 Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md)和[服务合并 Agent Note](../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md)。
接口：[dsh-subagent](../../packages/subagent/subagent)（`ctx.subagents` + 下文词汇）。实现为六个兄弟包（package）：`dsh-subagent-spawn`、`-fork`、`-acp`、`-codex`、`-claude-code`、`-dsh-sdk`；面向模型的消费方包括 [dsh-tool-subagent](../../packages/subagent/tool-subagent)（按提供方委派）、[dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control)（可选的全局 `send_message`、`interrupt_agent` 与 `list_agents` 控制工具）和 [dsh-tool-subagent-report](../../packages/subagent/tool-subagent-report)（可选的 child 作用域 `report` 返回通道）。同一个 `ctx.subagents` 服务通过内部激活管理器负责可继续子 agent 编排，并直接从会话存储与可选的会话持久化负责只读的 child 与后代发现。产品提供方设计理由见 [Codex 与 Claude Code Agent Note（agent 决策记录）](../../.agents/notes/implemented/feature/2026-08-04-claude-code-and-codex-subagent-backends.md)；通用 seam 的设计理由见 [subagent Agent Note](../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md)、[可继续 subagent Agent Note](../../.agents/notes/implemented/feature/2026-07-28-continuable-subagent-conversations.md)、[report 工具 Agent Note](../../.agents/notes/implemented/feature/2026-07-30-continuable-subagent-report-tool.md)、[持久化目录 Agent Note](../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md)、[列表身份投影 Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md)和[服务合并 Agent Note](../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md)。
源码：[`packages/subagent/subagent/src/types.ts`](../../packages/subagent/subagent/src/types.ts)、[`packages/subagent/subagent/src/index.ts`](../../packages/subagent/subagent/src/index.ts)和 [`packages/subagent/subagent/src/continuation.ts`](../../packages/subagent/subagent/src/continuation.ts)
@@ -139,7 +139,20 @@ Agent 收件箱是唯一的队列。每条继续执行消息都会成为一个 `
后续操作的权限来自确切的在线 Agent 工具上下文。已认证的 Agent 必须是持久化子 agent 在 `SessionHeader.parentSession` 中记录的直接父级。`MessageSource` 与 `senderSessionId` 在准入之后是持久的来源凭据，不授予任何权限；可选的面向模型工具使用 `CoordinatorMessageSource`。
对于这两种操作，调用方 signal 仅在收件箱接受之前掌管查找、物化与准入。此后管理器独立掌管该 Activation：之后的调用方取消既不会取消已接受的轮次，也不会 dispose 子 agent，并且该 seam 不对外暴露任何 subagent 取消或 steering（中途引导）操作。
对于这两种操作，调用方 signal 仅在收件箱接受之前掌管查找、物化与准入。此后管理器独立掌管该 Activation：之后的调用方取消既不会取消已接受的轮次，也不会 dispose 子 agent，并且该 seam 不对外暴露任何 steering（中途引导）操作。
`SubagentService.interrupt(targetSessionId, authority)` 是唯一的公开停止操作：它同步完成鉴权，对在线目标发出 `Agent.cancel(cause, { keepInbox: true })`，然后不等待静止即返回。Activation、其尚未领取的待处理 inbox 工作与已发布的后代均不受影响；已被领取进入中断轮次的工作不会重新入队。被中断的 driver 进入 idle 后，一次唤醒发送会恢复被暂停的 FIFO 队列。不存在的目标——未知、一次性或已结算——以及未绑定管理器的组合是被接受的 no-op。对在线目标，错误的 parent 地址或不在其在线祖先链中的调用方会以 `UNAUTHORIZED` 拒绝；过期的 ancestor 对象和指向自身的 ancestor 请求会在查找目标前拒绝。
```ts type-equiv
/**
* Authority under which one interrupt request is admitted. `user` carries the
* durable direct-parent address a human client presented; `ancestor` carries
* the exact live Agent object whose recorded lineage must contain the caller.
*/
type SubagentInterruptAuthority =
| { readonly kind: 'user'; readonly parentSessionId: SessionId }
| { readonly kind: 'ancestor'; readonly agent: Agent }
```
每个 Activation 都拥有自己的 `AgentHandle` 和一个 `ownedChildren: Set<SessionId>`；由于一份会话至多有一个存活 Activation，子会话 id 无需另一个运行时化身引用即可标识存活的子 agent。启动子 agent 或提交源自 parent 的工作，会在子 agent 能够运行之前将其注册到受继续执行管理的父级集合中；只要该集合非空，该父级就无法 settle。顶层或其他非继续执行的 Agent 没有 Activation，处于 waiting 图之外。只有当子 Agent 已停稳、该子 agent 的每个子级都已 dispose、best-effort 的最终会话 flush 结算完毕，且子 agent 的 `AgentHandle` 完成 dispose 之后，才会释放子 agent。
@@ -250,9 +263,26 @@ interface ContinuableCreateSpec {
本地一次性提供方会在子 agent 的初始轮次内、首次请求前追加描述符。继续执行管理器会在任何提供方提供的谱系之后、初始 prompt 获准之前追加描述符；`header.seedLength` 仍是 fork 谱系边界：恢复时的描述符权威读取子 agent 自身的后缀，而供列表使用的身份投影以 last-wins 折叠 `subagent/descriptor`，子 agent 自己的描述符会覆盖 fork seed 中祖先的描述符。该事件只进入日志：不含 `surfaceOp`，绝不进入模型历史，并由仅追加日志跨压缩保留。格式错误的当前版本描述符属于损坏；本运行时无法对不受支持的版本进行分类。
## 持久化枚举：`listChildren()` 与 `SubagentListEntry`
## 持久化枚举：`listChildren()`、`listDescendants()` 与其条目
`SubagentService.listChildren(parentSessionId)` 从 `ctx.sessions.list()` 与可选 `ctx.sessionPersistence.list()` 的实时优先合并中枚举 parent 直接且由会话支撑的 subagent——不经查询 seam，也不会加载或恢复任何 Agent。候选是持久 header 携带 `origin: 'subagent'` 的直接 child；该标记只负责枚举分类与粗粒度的通用路由拒绝，不能证明描述符有效、child 可恢复或操作已获授权——身份由投影折叠负责，恢复由 Activation 契约负责。每行的 `mode`／`label` 是已注册 `subagent` projection unit 的值，经三级阶梯供值：存活 child 由注册表水位缓存供值（零日志读取）；冷 child 先读可选的投影 checkpoint 缓存（`cachedSnapshot`——过 own-suffix seq 门的身份即定值，own descriptor 一经追加不可变）；否则在一次 `persistence.inspect()` 读取上经注册表折叠（有界并发，每次列表重新计算）。该缓存是纯可选加速层：服务缺席、行里是 `null` 哨兵或 key 缺席、seq 门不过、读取出错，都静默落到权威重折。折叠规则是 `subagent/descriptor` last-wins 且没有失败通道：子 agent 自己的描述符覆盖 fork seed 中祖先的描述符，格式错误或版本不认识的载荷折叠为可序列化的 `null` 哨兵，视同无值。结果是按 `createdAt`、再按 id 排序的 `SubagentListEntry[]`：取到身份即生成带有 `mode: 'one-shot' | 'continuable'` 和 `activity: 'running' | 'inactive'` 的 `child` 条目；可继续条目始终携带 `label`，一次性条目则只在启动调用方提供展示元数据时携带该字段。已定局而折叠无身份的候选生成 `corrupt` diagnostic——缺失、格式错误与版本不认识的描述符有意不再细分，`unsupported` 为已按其路由的消费方保留在类型中但不再产出；运行中而无身份的候选被省略（描述符落盘前的创建窗口）；冷检查失败生成一条 `unavailable` diagnostic 并在下次列表自然重试，因此一个损坏的 sibling 不会隐藏健康 child。`hasChildren` 标记存在持久 subagent origin 的直接后代，读取自同一份合并材料。活动状态只表示逻辑记录是否在 `ctx.sessions` 中存活，而不表示结果或可恢复性。缺少持久化时，枚举退化为仅存活枚举而不是报错——此时冷 child 本就无法恢复。缺少 `ctx.sessionProjections` 注册表时，`listChildren()` 抛出携带错误码 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 的 `SubagentError`，缺少会话存储时则抛出 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`，两者都在任何读取之前检查，因此零 child 的部署同样确定失败；列表工具在插件加载时要求 `ctx.subagents` 与 `ctx.agents`。UI 等服务消费方可以展示两种模式，并为无标签的一次性 child 选择回退展示；面向模型的 `list_agents` 适配器（[dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control) 中可单独加载的 `/list-agents` 插件）则只保留可继续条目，并通过在线 Agent 注册表将状态细化为 `running`／`idle`／`complete` 词汇。枚举不会查询继续执行管理器的 Activation map、Agent 注册表或提供方可用性；`send_message` 仍是消息送达时的权威操作，列表中的运行中可继续 child 仍可能因所有权冲突而拒绝投递。读路径的设计理由见[列表身份投影 Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md)。
`SubagentService.listDescendants(rootSessionId)` 将同一份实时优先语料与基于投影的解释应用到根的完整后代树，并按稳定 pre-order 输出。普通会话和一次性 child 仍作为遍历节点，因此其下的可继续后代仍可发现；只有 `origin: 'subagent'` 的候选会生成条目。每个返回的 child 或 diagnostic 都从枚举所得的持久 header 附加树位置；冷检查在提供身份前还会重新校验完整生命周期：
```ts type-equiv
/**
* One entry of a descendant listing: the interpreted subagent facts plus its
* position in the complete session tree. `parentId` is the durable direct
* parent from the enumerated header, and `depth` counts edges from the root.
*/
type SubagentDescendantListEntry = SubagentListEntry & {
/** Durable direct parent of this candidate in the enumerated tree. */
readonly parentId: SessionId
/** Edge distance from the requested root; direct children are `1`. */
readonly depth: number
}
```
`SubagentService.listChildren(parentSessionId)` 从 `ctx.sessions.list()` 与可选 `ctx.sessionPersistence.list()` 的实时优先合并中枚举 parent 直接且由会话支撑的 subagent——不经查询 seam，也不会加载或恢复任何 Agent。候选是持久 header 携带 `origin: 'subagent'` 的直接 child；该标记只负责枚举分类与粗粒度的通用路由拒绝，不能证明描述符有效、child 可恢复或操作已获授权——身份由投影折叠负责，恢复由 Activation 契约负责。每行的 `mode`／`label` 是已注册 `subagent` projection unit 的值，经三级阶梯供值：存活 child 由注册表水位缓存供值（零日志读取）；冷 child 先读可选的投影 checkpoint 缓存（`cachedSnapshot`——过 own-suffix seq 门的身份即定值，own descriptor 一经追加不可变）；否则在一次 `persistence.inspect()` 读取上经注册表折叠（有界并发，每次列表重新计算）。该缓存是纯可选加速层：服务缺席、行里是 `null` 哨兵或 key 缺席、seq 门不过、读取出错，都静默落到权威重折。折叠规则是 `subagent/descriptor` last-wins 且没有失败通道：子 agent 自己的描述符覆盖 fork seed 中祖先的描述符，格式错误或版本不认识的载荷折叠为可序列化的 `null` 哨兵，视同无值。结果是按 `createdAt`、再按 id 排序的 `SubagentListEntry[]`：取到身份即生成带有 `mode: 'one-shot' | 'continuable'` 和 `activity: 'running' | 'inactive'` 的 `child` 条目；可继续条目始终携带 `label`，一次性条目则只在启动调用方提供展示元数据时携带该字段。已定局而折叠无身份的候选生成 `corrupt` diagnostic——缺失、格式错误与版本不认识的描述符有意不再细分，`unsupported` 为已按其路由的消费方保留在类型中但不再产出；运行中而无身份的候选被省略（描述符落盘前的创建窗口）；冷检查失败生成一条 `unavailable` diagnostic 并在下次列表自然重试，因此一个损坏的 sibling 不会隐藏健康 child。`hasChildren` 标记存在持久 subagent origin 的直接后代，读取自同一份合并材料。活动状态只表示逻辑记录是否在 `ctx.sessions` 中存活，而不表示结果或可恢复性。缺少持久化时，枚举退化为仅存活枚举而不是报错——此时冷 child 本就无法恢复。缺少 `ctx.sessionProjections` 注册表时，`listChildren()` 抛出携带错误码 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 的 `SubagentError`，缺少会话存储时则抛出 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`，两者都在任何读取之前检查，因此零 child 的部署同样确定失败；列表工具在插件加载时只要求 `ctx.subagents`。UI 等服务消费方可以展示两种模式，并为无标签的一次性 child 选择回退展示；面向模型的 `list_agents` 适配器（[dsh-tool-subagent-control](../../packages/subagent/tool-subagent-control) 中可单独加载的 `/list-agents` 插件）则只保留可继续条目，并将活动状态映射到现有的 `running`／`complete` 词汇。枚举不会查询继续执行管理器的 Activation map、Agent 注册表或提供方可用性；`send_message` 仍是消息送达时的权威操作，列表中的运行中可继续 child 仍可能因所有权冲突而拒绝投递。读路径的设计理由见[列表身份投影 Agent Note](../../.agents/notes/implemented/architecture/2026-08-06-subagent-list-identity-projection.md)。
## 终态结果：`SubagentResult`
docs/event-producer-consumer.md
+4 -4
View File
@@ -37,10 +37,10 @@ This matrix shows which packages dispatch each harness-owned event and which pac
| `settings/document-updated` | `emit` | [`packages/settings/settings/src/index.ts:170`](../packages/settings/settings/src/index.ts) | [`settings`](../packages/settings/settings) (`events.dispatch`) | `apiproxy` |
| `settings/updated` | `emit` | [`packages/settings/settings/src/index.ts:157`](../packages/settings/settings/src/index.ts) | [`settings`](../packages/settings/settings) (`events.dispatch`) | [`settings`](../packages/settings/settings) |
| `skills/change` | `emit` | [`packages/skill/skill/src/index.ts:283`](../packages/skill/skill/src/index.ts) | [`skill`](../packages/skill/skill) (`events.dispatch`) | - |
| `subagent/end` | `emit` | [`packages/subagent/subagent/src/index.ts:160`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude`](../packages/hooks/hooks-claude), [`jsonrpc`](../packages/ui/jsonrpc), [`subagent`](../packages/subagent/subagent) |
| `subagent/provider-added` | `emit` | [`packages/subagent/subagent/src/index.ts:134`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`emit`) | [`subagent`](../packages/subagent/subagent), [`tool-subagent`](../packages/subagent/tool-subagent) |
| `subagent/provider-removed` | `emit` | [`packages/subagent/subagent/src/index.ts:140`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`subagent`](../packages/subagent/subagent), [`tool-subagent`](../packages/subagent/tool-subagent) |
| `subagent/start` | `emit` | [`packages/subagent/subagent/src/index.ts:151`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude`](../packages/hooks/hooks-claude), [`subagent`](../packages/subagent/subagent) |
| `subagent/end` | `emit` | [`packages/subagent/subagent/src/index.ts:162`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude`](../packages/hooks/hooks-claude), [`jsonrpc`](../packages/ui/jsonrpc), [`subagent`](../packages/subagent/subagent) |
| `subagent/provider-added` | `emit` | [`packages/subagent/subagent/src/index.ts:136`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`emit`) | [`subagent`](../packages/subagent/subagent), [`tool-subagent`](../packages/subagent/tool-subagent) |
| `subagent/provider-removed` | `emit` | [`packages/subagent/subagent/src/index.ts:142`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`subagent`](../packages/subagent/subagent), [`tool-subagent`](../packages/subagent/tool-subagent) |
| `subagent/start` | `emit` | [`packages/subagent/subagent/src/index.ts:153`](../packages/subagent/subagent/src/index.ts) | [`subagent`](../packages/subagent/subagent) (`events.dispatch`) | [`hooks-claude`](../packages/hooks/hooks-claude), [`subagent`](../packages/subagent/subagent) |
| `system-prompt/assemble` | `waterfall` | [`packages/core/system-prompt/src/index.ts:29`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`waterfall`) | [`agent`](../packages/core/agent), [`system-prompt`](../packages/core/system-prompt) |
| `system-prompt/change` | `emit` | [`packages/core/system-prompt/src/index.ts:35`](../packages/core/system-prompt/src/index.ts) | [`system-prompt`](../packages/core/system-prompt) (`emit`) | - |
| `telemetry/record` | `waterfall` | [`packages/telemetry/session-telemetry/src/index.ts:43`](../packages/telemetry/session-telemetry/src/index.ts) | [`session-telemetry`](../packages/telemetry/session-telemetry) (`waterfall`) | - |
docs/tool-catalog.md
+35 -5
View File
@@ -32,7 +32,7 @@ This table connects model-visible tool names to the plugin package and service s
| `@deepseek-ai/dsh-tool-skill` | `skill` | `ctx.tools`, `ctx.agents`, `ctx.skills` | `tool/call`, `tool/result`, `user/message replacement catalogs via agent.inject()` | - | - |
| `@deepseek-ai/dsh-tool-session-query` | `session_event_read`, `session_event_search`, `session_event_trace`, `session_search`, `session_trace` | `ctx.tools`, `ctx.systemPrompt`, `ctx.sessionQuery`, `a calling Agent for workspace authority` | `tool/call`, `tool/result` | - | The five read-only tools hide provider cursors and authorize every result from the immutable calling agent session. The package is opt-in; compositions that need enforced deadlines or bounded inline output also mount the generic timeout or spill policies. |
| `@deepseek-ai/dsh-tool-subagent` | `subagent` | `ctx.tools`, `ctx.subagents` | `tool/call`, `tool/result`, `child session events through the chosen provider` | `subagent`, `subagent_fork` | The registered tool name is the load-time `toolName` config (default `subagent`); the schema above is that default. The shipped example agents load this package once per subagent backend, so the model additionally sees `subagent_fork` (bound to the fork backend) with an identical schema — see `packages/bundle/base/cordis.patch.yml` and `examples/acp-agent/cordis.yml`. |
| `@deepseek-ai/dsh-tool-subagent-control` | `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.sessionProjections (list_agents catalog rows)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows are served through the sessionProjections registry). |
| `@deepseek-ai/dsh-tool-subagent-control` | `interrupt_agent`, `list_agents`, `send_message` | `ctx.tools`, `ctx.subagents`, `ctx.agents and ctx.sessionProjections (list_agents only)` | `tool/call`, `tool/result`, `child session events through ctx.subagents` | - | The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries). |
| `@deepseek-ai/dsh-tool-subagent-report` | `report` | `ctx.subagents`, `a live continuable in-process child Agent` | `tool/call`, `tool/result`, `a user-role message in the direct parent session` | - | Registered per continuable in-process child rather than globally, so this schema is visible only inside such a child and survives its global `toolFilter`. The parent-facing `send_message` tool is installed independently. |
| `@deepseek-ai/dsh-tool-tasks` | `task_kill`, `task_list`, `task_output` | `ctx.tools`, `ctx.tasks`, `ctx.systemPrompt` | `tool/call`, `tool/result`, `user/message via agent.inject() for background completion notices` | - | The kind-agnostic background-task control surface: background bash commands, PTY sends, and subagents are read, listed, and killed through the same three tools. Loading the plugin attaches the control surface that arms producers' `ctx.tasks.start()`. |
| `@deepseek-ai/dsh-tool-todo` | `todo_write` | `ctx.tools`, `owning Agent session` | `tool/call`, `todo/write`, `tool/result` | - | todo_write is session-owned state; UIs render the latest todo/write event as a checklist. `allowParallelInProgress` is required with no default, so the catalog states its choice: `true`, whose description invites several `in_progress` items. A deployment choosing `false` receives the same tool with a description asking for exactly one active task. |
@@ -1193,14 +1193,44 @@ The registered tool name is the load-time `toolName` config (default `subagent`)
## `@deepseek-ai/dsh-tool-subagent-control`
### `list_agents`
### `interrupt_agent`
List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.
Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.
```json
{
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
```
Source: [`packages/subagent/tool-subagent-control/src/index.ts`](../packages/subagent/tool-subagent-control/src/index.ts)
### `list_agents`
List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.
```json
{
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
```
@@ -1232,7 +1262,7 @@ Send a message to a background subagent by its subagent id, continuing the same
Source: [`packages/subagent/tool-subagent-control/src/index.ts`](../packages/subagent/tool-subagent-control/src/index.ts)
The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows are served through the sessionProjections registry).
The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries).
## `@deepseek-ai/dsh-tool-subagent-report`
examples/acp-agent/tests/snapshots/advanced-toolchain/system-prompt.expected.md
+18 -3
View File
@@ -94,8 +94,16 @@ interface ToolArgsMap {
} & Record<string, JsonValue>;
/** Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal. */
get_goal: Record<string, JsonValue>;
/** List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. */
list_agents: Record<string, JsonValue>;
/** Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op. */
interrupt_agent: {
/** The agent id of the running agent to interrupt. */
agent_id: string;
} & Record<string, JsonValue>;
/** List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only. */
list_agents: {
/** children (default) lists direct children only; descendants walks the complete tree below you. */
scope?: "children" | "descendants";
} & Record<string, JsonValue>;
/** Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools. */
ralph: {
/** The immutable completion objective for every fresh Ralph round. */
@@ -304,15 +312,22 @@ interface ToolOutputMap {
};
activation: "armed" | "disarmed";
};
interrupt_agent: {
accepted: boolean;
};
list_agents: ({
kind: "child";
id: string;
label: string;
status: "running" | "complete";
status: "running" | "idle" | "complete";
parent?: string;
depth?: number;
} | {
kind: "diagnostic";
id: string;
reason: "corrupt" | "unsupported" | "unavailable";
parent?: string;
depth?: number;
})[];
ralph: {
runId: string;
examples/acp-agent/tests/snapshots/advanced-toolchain/tool-schemas.expected.json
+28 -3
View File
@@ -173,11 +173,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/both-mode-turn/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/code-mode-turn/system-prompt.expected.md
+18 -3
View File
@@ -77,8 +77,16 @@ interface ToolArgsMap {
} & Record<string, JsonValue>;
/** Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal. */
get_goal: Record<string, JsonValue>;
/** List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. */
list_agents: Record<string, JsonValue>;
/** Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op. */
interrupt_agent: {
/** The agent id of the running agent to interrupt. */
agent_id: string;
} & Record<string, JsonValue>;
/** List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only. */
list_agents: {
/** children (default) lists direct children only; descendants walks the complete tree below you. */
scope?: "children" | "descendants";
} & Record<string, JsonValue>;
/** Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools. */
ralph: {
/** The immutable completion objective for every fresh Ralph round. */
@@ -275,15 +283,22 @@ interface ToolOutputMap {
};
activation: "armed" | "disarmed";
};
interrupt_agent: {
accepted: boolean;
};
list_agents: ({
kind: "child";
id: string;
label: string;
status: "running" | "complete";
status: "running" | "idle" | "complete";
parent?: string;
depth?: number;
} | {
kind: "diagnostic";
id: string;
reason: "corrupt" | "unsupported" | "unavailable";
parent?: string;
depth?: number;
})[];
ralph: {
runId: string;
examples/acp-agent/tests/snapshots/lsp-definition/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/pty-tools/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/session-query-spill/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/subagent-child-question-rejection/tool-schemas.expected.json
+28 -3
View File
@@ -179,11 +179,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/subagent-continuable/tool-schemas.1.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/subagent-list-agents/tool-schemas.1.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/subagent-report/tool-schemas.1.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/text-turn/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
examples/acp-agent/tests/snapshots/web-fetch/tool-schemas.expected.json
+28 -3
View File
@@ -116,11 +116,36 @@
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status is a snapshot of the stored record: running means the subagent session is currently live in this process, complete means it exists only in storage and a `send_message` starts a new turn on the same conversation. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped.",
"name": "interrupt_agent",
"description": "Request cancellation of a background agent's current turn by its agent id. The target may be your direct child or a deeper agent created under you. Only the current turn stops: messages already queued for the agent stay parked until a later send_message, agents it started keep running, and the agent itself stays available for follow-ups. This call returns as soon as the stop request is accepted, so the target may keep running briefly; interrupting an agent that already finished is an accepted no-op.",
"parameters": {
"type": "object",
"properties": {}
"properties": {
"agent_id": {
"type": "string",
"description": "The agent id of the running agent to interrupt."
}
},
"required": [
"agent_id"
]
}
},
{
"name": "list_agents",
"description": "List your continuable background subagents by durable id and label. Status comes from the live registry: running means the agent is working right now, idle means it is loaded but between turns (it may be waiting on agents it started), and complete means it exists only in storage — a direct child remains a `send_message` candidate in every status. The snapshot is not a delivery promise — `send_message` performs the authoritative check and may still fail. Children that could not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are candidates for `interrupt_agent` only.",
"parameters": {
"type": "object",
"properties": {
"scope": {
"type": "string",
"description": "children (default) lists direct children only; descendants walks the complete tree below you.",
"enum": [
"children",
"descendants"
]
}
}
}
},
{
packages/client/connection/src/client/fixture.ts
+2
View File
@@ -2203,6 +2203,7 @@ function createFixtureWorld(options: FixtureOptions): FixtureWorld {
prompt: request => Promise.resolve(ok(request, {
messageId: `fixture-message-${request.payload.childSessionId}` as never,
})),
interrupt: request => Promise.resolve(ok(request, { accepted: true as const })),
},
host: {
describe: request => ok(request, { version: '0.0.0-fixture', cwd: '/tmp/fixture', attachedSessions }),
@@ -2748,6 +2749,7 @@ export class FixtureApiClient extends AbstractApiClient {
case 'subagent.list': return this.api.subagents.list(request)
case 'subagent.history': return this.api.subagents.history(request)
case 'subagent.prompt': return this.api.subagents.prompt(request, signal)
case 'subagent.interrupt': return this.api.subagents.interrupt(request)
case 'host.describe': return this.api.host.describe(request)
case 'host.pickDirectory': return this.api.host.pickDirectory(request, new AbortController().signal)
case 'host.listDirectory': return this.api.host.listDirectory(request, new AbortController().signal)
packages/client/connection/tests/fake-api.ts
+3
View File
@@ -126,6 +126,9 @@ export class FakeApiClient implements IApiClient {
prompt: (payload: unknown) => this.record('subagent.prompt', payload, Promise.resolve(ok({
messageId: 'fake-message' as never,
}))),
interrupt: (payload: unknown) => this.record('subagent.interrupt', payload, Promise.resolve(ok({
accepted: true as const,
}))),
}
readonly host: IApiClient['host'] = {
packages/client/runtime/src/client/sessions/session.ts
+11 -4
View File
@@ -281,17 +281,22 @@ export class Session implements SessionFace {
/**
* Stop the active turn while the Host preserves pending inbox work; failures
* land in promptError (same error-strip display slot).
* land in promptError (same error-strip display slot). A continuable
* subagent address routes through `subagent.interrupt`, whose durable
* parent-address authority works without a live parent Agent; a one-shot
* address stays uncancellable (the UI offers no stop action, so this arm is
* defensive).
* @returns the cancel result.
*/
async cancel(): Promise<RpcResult<{ accepted: true }>> {
if (this.address !== undefined) {
const address = this.address
if (address !== undefined && address.mode === 'one-shot') {
const result: RpcResult<{ accepted: true }> = {
ok: false,
error: {
code: 'subagent-delivery-unavailable',
message: 'subagent activation cancellation is unavailable',
details: { childSessionId: this.address.childSessionId },
details: { childSessionId: address.childSessionId },
},
}
this.promptError = { op: 'stop', error: result.error }
@@ -300,7 +305,9 @@ export class Session implements SessionFace {
}
let result: RpcResult<{ accepted: true }>
try {
result = (await this.api.sessions.cancel({ sessionId: this.sessionId })).result
result = address !== undefined
? (await this.api.subagents.interrupt(address)).result
: (await this.api.sessions.cancel({ sessionId: this.sessionId })).result
} catch (error) {
result = transportError(error)
}
packages/client/runtime/tests/fake-api.ts
+4
View File
@@ -140,10 +140,14 @@ export class FakeApiClient implements IApiClient {
onSubagentPrompt: (payload: unknown) => Promise<RpcResponse<{ messageId: never }>>
= () => Promise.resolve(ok({ messageId: 'fake-message' as never }))
onSubagentInterrupt: (payload: unknown) => Promise<RpcResponse<{ accepted: true }>>
= () => Promise.resolve(ok({ accepted: true as const }))
readonly subagents: IApiClient['subagents'] = {
list: (payload: unknown) => this.record('subagent.list', payload, this.onSubagentList(payload)),
history: (payload: unknown) => this.record('subagent.history', payload, this.onSubagentHistory(payload)),
prompt: (payload: unknown) => this.record('subagent.prompt', payload, this.onSubagentPrompt(payload)),
interrupt: (payload: unknown) => this.record('subagent.interrupt', payload, this.onSubagentInterrupt(payload)),
}
readonly host: IApiClient['host'] = {
packages/client/runtime/tests/session.spec.ts
+28 -2
View File
@@ -639,7 +639,7 @@ describe('paging', () => {
})
describe('prompt and cancel errors', () => {
it('routes an addressed child through non-activating history and continuation prompt only', async () => {
it('routes an addressed child through non-activating history, continuation prompt, and interrupt only', async () => {
const api = new FakeApiClient()
const session = new Session(SID, api, {
address: { parentSessionId: PARENT, childSessionId: SID, mode: 'continuable' },
@@ -650,7 +650,7 @@ describe('prompt and cancel errors', () => {
const cancelled = await session.cancel()
expect(prompted).toEqual({ ok: true, value: { accepted: true } })
expect(cancelled).toMatchObject({ ok: false, error: { code: 'subagent-delivery-unavailable' } })
expect(cancelled).toEqual({ ok: true, value: { accepted: true } })
expect(api.callsOf('subagent.history')).toEqual([
{ parentSessionId: PARENT, childSessionId: SID, mode: 'continuable', maxMessages: 50 },
])
@@ -660,15 +660,37 @@ describe('prompt and cancel errors', () => {
content: [{ type: 'text', text: '继续' }],
},
])
expect(api.callsOf('subagent.interrupt')).toEqual([
{ parentSessionId: PARENT, childSessionId: SID, mode: 'continuable' },
])
expect(api.callsOf('session.history')).toEqual([])
expect(api.callsOf('session.prompt')).toEqual([])
expect(api.callsOf('session.cancel')).toEqual([])
// A successful interrupt leaves no stop error behind.
expect(session.getSnapshot().promptError).toBeNull()
expect(session.getSnapshot().subagent).toEqual({
address: { parentSessionId: PARENT, childSessionId: SID, mode: 'continuable' },
parentAvailable: true,
})
})
it('lands an interrupt business failure in promptError with op=stop', async () => {
const api = new FakeApiClient()
api.onSubagentInterrupt = () => Promise.resolve(err({
code: 'subagent-unauthorized', message: 'nope', details: { childSessionId: SID },
}) as never)
const session = new Session(SID, api, {
address: { parentSessionId: PARENT, childSessionId: SID, mode: 'continuable' },
parentAvailable: true,
})
await session.open()
const cancelled = await session.cancel()
expect(cancelled).toMatchObject({ ok: false, error: { code: 'subagent-unauthorized' } })
expect(session.getSnapshot().promptError).toMatchObject({
op: 'stop', error: { code: 'subagent-unauthorized' },
})
})
it('keeps one-shot history readable without exposing prompt or cancel transport', async () => {
const api = new FakeApiClient()
const session = new Session(SID, api, {
@@ -676,12 +698,16 @@ describe('prompt and cancel errors', () => {
})
await session.open()
const prompted = await session.prompt([{ type: 'text', text: '继续' }], 'queue')
const cancelled = await session.cancel()
expect(prompted).toMatchObject({ ok: false, error: { code: 'subagent-not-resumable' } })
expect(cancelled).toMatchObject({ ok: false, error: { code: 'subagent-delivery-unavailable' } })
expect(api.callsOf('subagent.history')).toEqual([
{ parentSessionId: PARENT, childSessionId: SID, mode: 'one-shot', maxMessages: 50 },
])
expect(api.callsOf('subagent.prompt')).toEqual([])
expect(api.callsOf('subagent.interrupt')).toEqual([])
expect(api.callsOf('session.cancel')).toEqual([])
})
it('sends content through session.prompt; composerPhase steps blank → engaging synchronously at send entry', async () => {
packages/client/ui-conversation/src/client/locales.ts
+2
View File
@@ -17,6 +17,7 @@ export const zh = {
'placeholder.plan': PLAN_NEXT_ACTION_ZH,
'placeholder.default': '给智能体发消息',
'placeholder.unavailable': '会话不可用',
'placeholder.parentOffline': '父会话已离线，无法继续发送；仍可停止当前运行',
'placeholder.hero': '描述你想要构建的内容',
'placeholder.workspace': '选择一个工作区开始',
'input.commands': '命令',
@@ -159,6 +160,7 @@ export const en = {
'placeholder.plan': PLAN_NEXT_ACTION_EN,
'placeholder.default': 'Message the agent',
'placeholder.unavailable': 'Session unavailable',
'placeholder.parentOffline': 'Parent session offline; sending is unavailable but you can still stop the run',
'placeholder.hero': 'Describe what you want to build',
'placeholder.workspace': 'Choose a workspace to start',
'input.commands': 'Commands',
packages/client/ui-conversation/src/client/skeleton/InputBar.tsx
+16 -9
View File
@@ -83,11 +83,15 @@ export function InputBar({
// (undefined = capability absent → the chip renders nothing).
const permissions = useProjection('permissions')
// A continuable child without its live parent cannot accept human input,
// but its primary Stop below stays available while it runs.
const continuable = subagent?.address.mode === 'continuable'
const parentOffline = continuable && !subagent.parentAvailable
// Queue cut 1: running input stays free; locked = session removed, the
// inert no-workspace state, or the machine faces absent (no session). The
// transient machine locks (adjudicating pending / submitting) render
// read-only — the draft stays visible and focused, keystrokes drop.
const disabled = removed || inert || !live || blocked !== undefined
// inert no-workspace state, the machine faces absent (no session), or a
// parent-offline continuable child. An owner block also disables input;
// adjudicating and submitting render read-only so the draft stays visible.
const disabled = removed || inert || !live || blocked !== undefined || parentOffline
const locked = disabled
// The model seat is the ONE control a block leaves live: every block this
// contract has is cleared by choosing a model, so locking it too would leave
@@ -350,8 +354,9 @@ export function InputBar({
if (el !== null) toggleCommandMenu?.(selectionOf(el))
}
const ordinary = subagent === null
const stopping = running && ordinary
// Ordinary sessions and continuable children stop through the same primary
// action; one-shot children stay send-and-stop-free (read-only takeover).
const stopping = running && (subagent === null || continuable)
const primaryLabel = stopping ? t('input.stop') : t('input.send')
const onPrimary = (): void => {
if (stopping) {
@@ -478,9 +483,11 @@ export function InputBar({
disabled={locked}
readOnly={machineBusy}
data-phase={input?.phase ?? 'inert'}
placeholder={placeholder ?? (disabled
? t('placeholder.unavailable')
: planActive ? t('placeholder.plan') : t('placeholder.default'))}
placeholder={placeholder ?? (parentOffline
? t('placeholder.parentOffline')
: disabled
? t('placeholder.unavailable')
: planActive ? t('placeholder.plan') : t('placeholder.default'))}
rows={2}
onChange={onChange}
onKeyDown={onKeyDown}
packages/client/ui-conversation/tests/input-bar.spec.tsx
+39 -7
View File
@@ -143,7 +143,8 @@ function bench(over?: BenchOptions) {
}
const view = render(<InputBar {...props} />)
const textarea = view.container.querySelector('textarea')!
const stopping = over?.running === true && over.subagent === undefined
const stopping = over?.running === true
&& (over.subagent === undefined || over.subagent.address.mode === 'continuable')
const button = view.container.querySelector<HTMLButtonElement>(
`button[aria-label="${stopping ? '停止生成' : '发送消息'}"]`,
)!
@@ -250,8 +251,8 @@ describe('running and lock semantics (queue cut 1)', () => {
expect(ctrl.sink).toHaveBeenCalledWith('also queue', 'queue')
})
it('running subagent primary admits a follow-up instead of exposing Stop', () => {
const { button, sink, stop } = bench({
it('running continuable subagent turns the same primary into Stop while typing stays free', () => {
const { button, textarea, sink, stop } = bench({
running: true,
draft: '后续消息',
subagent: {
@@ -263,23 +264,54 @@ describe('running and lock semantics (queue cut 1)', () => {
parentAvailable: true,
},
})
expect(button.getAttribute('aria-label')).toBe('发送消息')
// One primary action only: Send switched to Stop, no side-by-side interrupt.
expect(button.getAttribute('aria-label')).toBe('停止生成')
expect(textarea.disabled).toBe(false)
fireEvent.click(button)
expect(stop).toHaveBeenCalledTimes(1)
expect(sink).not.toHaveBeenCalled()
// Enter still queues the follow-up while the primary offers Stop.
fireEvent.keyDown(textarea, { key: 'Enter' })
expect(sink).toHaveBeenCalledWith('后续消息', 'queue')
expect(stop).not.toHaveBeenCalled()
})
const empty = bench({
it('parent-offline running continuable locks the input but keeps the same primary Stop usable', () => {
const { button, textarea, stop, view } = bench({
running: true,
draft: '',
subagent: {
address: {
parentSessionId: 'parent' as SessionId,
childSessionId: SID,
mode: 'continuable',
},
parentAvailable: false,
},
})
expect(textarea.disabled).toBe(true)
expect(textarea.placeholder).toBe('父会话已离线，无法继续发送；仍可停止当前运行')
expect((view.getByLabelText('命令') as HTMLButtonElement).disabled).toBe(true)
expect(button.getAttribute('aria-label')).toBe('停止生成')
expect(button.disabled).toBe(false)
fireEvent.click(button)
expect(stop).toHaveBeenCalledTimes(1)
})
it('running one-shot subagent never exposes Stop', () => {
const { button, stop } = bench({
running: true,
draft: '不可停止',
subagent: {
address: {
parentSessionId: 'parent' as SessionId,
childSessionId: SID,
mode: 'one-shot',
},
parentAvailable: true,
},
})
expect(empty.button.disabled).toBe(true)
expect(button.getAttribute('aria-label')).toBe('发送消息')
expect(stop).not.toHaveBeenCalled()
})
it('keeps both running subagent Enter gestures on Queue transport', () => {
packages/client/ui-subagent/README.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/client/ui-subagent/README.md
README.md: cb210b219a8c66985eb4e1370468372eed9614b4
README.zh.md: 857e92d05a7ed2d0df9398acc9698db13b0c6eb2
README.md: de55963bbe356e18a4e2b653be823cc5015c1bb7
README.zh.md: ce4244db094bb8b602b854289ec000477bf45f0b
packages/client/ui-subagent/README.md
+2 -2
View File
@@ -6,7 +6,7 @@ Web subagent feature owner: contributes the lazily expandable catalog tree to `c
The header action reads `subagentsByParent` and session summaries through the standard `useSessions` hook. After a non-empty direct catalog arrives, its trigger counts the complete subagent-only descendant lineage, stops at ordinary forks, and shows ongoing activity when any counted descendant is running. The compact tree remains direct-catalog authoritative: continuable and one-shot rows display mode plus `running`/`inactive` activity and an optional log-backed title, while the trailing column stacks total durable provider usage above active-turn duration. Token totals sum the four disjoint `tokenUsage` buckets. Visual duration stays exact to the second below one day, then uses at most two adjacent units—days/hours, approximate months/days, or approximate years/months—while hover and the accessible name retain the exact day/hour/minute/second value. Duration sums completed `subagentTiming` turns, advances once per second only for an open turn on a running child, and freezes after the child becomes inactive; an interrupted open turn is bounded by its same-cut `active.through`, never by newer session metadata. An unlabeled one-shot row falls back to its session id, while corrupt, unsupported, or unavailable rows remain readable but disabled. Each healthy row's `hasChildren` hint determines disclosure before interaction, so known leaves never show an arrow; a catalog level reserves the disclosure column only when at least one healthy row is a branch, allowing branchless levels to start at the leading status marker. Expanding a branch immediately reserves one disabled loading row per known direct descendant, then lazily replaces them with that child's authoritative catalog. Every visible branch is reported to the runtime so membership frames cause a debounced refresh only where the tree is being consumed. Selecting any depth calls `SessionsService.openSubagent()` with the row's exact `{parentSessionId, childSessionId, mode}` address. Component-local state owns tree visibility, expanded branches, keyboard focus, and the running-duration clock. ArrowRight/ArrowLeft expand and collapse branches; ArrowUp/ArrowDown, Home, End, and Escape navigate or close the tree; closing returns focus to the trigger. Styling uses tokens only.
A one-shot child always elects a read-only composer that identifies the transcript as a completed execution record. A continuable child does so only when its exact parent is unavailable, with copy explaining the recovery path. A continuable child with a live parent keeps the ordinary input chrome, whose Session routes through `subagent.prompt`; running input remains Send because every follow-up joins the child's FIFO inbox, and addressed sessions never expose Stop. This package never receives host context or calls a model-facing tool. The catalog and composer behavior are specified by the [Web subagent conversations Agent Note](../../../.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md).
A one-shot child always elects a read-only composer that identifies the transcript as a completed execution record. A continuable child does so only when its exact parent is unavailable and the child is not running, with copy explaining the recovery path; while such a child still runs, the selector yields to the ordinary composer, whose input is disabled but whose primary Stop stays usable, and the takeover returns once it stops. A continuable child with a live parent keeps the ordinary input chrome, whose Session routes prompts through `subagent.prompt`: typing stays free while the child runs because every follow-up joins the child's FIFO inbox, and the single primary action follows the ordinary convention — Send while idle, Stop while running, routed through `subagent.interrupt`. This package never receives host context or calls a model-facing tool. The catalog and composer behavior are specified by the [Web subagent conversations Agent Note](../../../.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md) and the [current-turn interrupt Agent Note](../../../.agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.md).
Subagent-origin Session rows are omitted from the ordinary sidebar, so the parent header catalog is their navigation entry point. Ordinary forks remain in the sidebar.
@@ -30,5 +30,5 @@ Append-only. This package never edits earlier request tokens.
## Known Limitations and Deferred Work
- **The catalog has no durable outcome** — activity and timing do not distinguish completion, failure, or cancellation, and the UI exposes neither Activation identity nor an authority-safe cancel button.
- **The catalog has no durable outcome** — activity and timing do not distinguish completion, failure, or cancellation, and the UI exposes no Activation identity; stopping is limited to the composer's current-turn Stop for a running continuable child.
- **`@` references remain display-title text** — duplicate or renamed labels are ambiguous, so they intentionally do not acquire continuation semantics.
packages/client/ui-subagent/README.zh.md
+2 -2
View File
@@ -6,7 +6,7 @@ Web subagent 功能 owner：向 `conversation.session.header.actions` 贡献可
页头操作通过标准 `useSessions` 钩子读取 `subagentsByParent` 与会话摘要。非空直接目录到达后，其触发器会统计仅含 subagent 的完整后代谱系，在普通 fork 处停止，并在任一计入统计的后代处于 `running` 时显示活动仍在进行。紧凑树仍以直接目录为权威依据：可继续和 one-shot 行会显示 mode、`running`／`inactive` 活动状态和由日志支撑的可选 title，尾随列则在上行显示提供方的持久化 token 用量总计，在下行显示活跃轮次耗时。token 用量总计为四个互不重叠的 `tokenUsage` 桶之和。视觉耗时在不足一天时精确到秒，达到一天后则最多使用两个相邻单位——天／小时、近似月份／天或近似年份／月份——而悬停信息与无障碍名称会保留精确的天／小时／分钟／秒数值。耗时会累加已完成的 `subagentTiming` 轮次，仅在运行中 child 存在未结束轮次时每秒递增一次，并在 child 变为 inactive 后冻结；被中断的未结束轮次以其同一切面的 `active.through` 为上界，绝不使用更新的会话元数据。没有 label 的 one-shot 行会回退到其会话 id，而损坏、不受支持或不可用的行仍保持可读但禁用。每个健康行的 `hasChildren` 提示会在交互前决定是否显示展开控件，因此已知叶子节点从不显示箭头；每层目录仅在其中至少一个健康行是分支时才预留展开列，使完全不含分支的层级能从最前面的状态标记开始。展开分支时，会立即为每个已知直接后代预留一行禁用的加载行，随后再用该 child 的权威目录懒加载结果替换这些占位行。每个可见分支都会上报给运行时，使成员帧只在树正被消费的位置触发去抖动刷新。选择任意深度的条目都会使用该行的确切地址 `{parentSessionId, childSessionId, mode}` 调用 `SessionsService.openSubagent()`。组件局部状态负责树的可见性、已展开分支、键盘焦点与运行中耗时时钟。ArrowRight／ArrowLeft 展开和折叠分支；ArrowUp／ArrowDown、Home、End 与 Escape 用于导航或关闭树；关闭后焦点返回触发器。样式只使用 token。
one-shot child 始终选用只读编辑器，并将 transcript（文本记录）说明为已完成的执行记录。可继续 child 仅在其确切 parent 不可用时选用只读编辑器，并以文案说明恢复路径。确切 parent 存活时，可继续 child 保留普通输入 chrome，其会话会通过 `subagent.prompt` 路由；child 运行期间，输入操作仍为 Send，因为每条后续消息都会进入 child 的 FIFO inbox，且已寻址会话绝不公开 Stop。本包绝不接收宿主上下文，也不调用面向模型的工具。目录与编辑器行为由 [Web subagent 对话 Agent Note](../../../.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md) 规定。
one-shot child 始终选用只读编辑器，并将 transcript（文本记录）说明为已完成的执行记录。可继续 child 仅在其确切 parent 不可用且 child 未在运行时选用只读编辑器，并以文案说明恢复路径；此类 child 仍在运行期间，selector 会让位给普通编辑器——其输入区被禁用，但 primary Stop 保持可用，停止后只读替代恢复。确切 parent 存活时，可继续 child 保留普通输入 chrome，其会话通过 `subagent.prompt` 路由提示词：child 运行期间输入保持自由，因为每条后续消息都会进入 child 的 FIFO inbox，而唯一的 primary 操作遵循普通约定——空闲时为 Send，运行时为 Stop，经由 `subagent.interrupt` 路由。本包绝不接收宿主上下文，也不调用面向模型的工具。目录与编辑器行为由 [Web subagent 对话 Agent Note](../../../.agents/notes/implemented/feature/2026-07-27-web-subagent-conversations.md) 与[当前轮次中断 Agent Note](../../../.agents/notes/implemented/feature/2026-08-06-continuable-subagent-interrupt.md) 规定。
普通侧边栏会省略带 subagent origin 的会话行，因此 parent 页头目录是它们的导航入口。普通 fork 仍保留在侧边栏中。
@@ -30,5 +30,5 @@ one-shot child 始终选用只读编辑器，并将 transcript（文本记录）
## 已知限制与暂缓事项
- **目录没有持久化结果**：活动状态与计时无法区分完成、失败或取消，且 UI 既不公开 Activation 身份，也不公开符合授权边界的取消按钮。
- **目录没有持久化结果**：活动状态与计时无法区分完成、失败或取消，且 UI 不公开 Activation 身份；停止能力仅限编辑器上针对运行中可继续 child 的当前轮次 Stop。
- **`@` 引用仍是显示标题文本**：重复或改名后的 label 会有歧义，因此它们刻意不获得继续执行语义。
packages/client/ui-subagent/src/client/index.ts
+5 -1
View File
@@ -43,7 +43,11 @@ function selectReadOnlySubagent(owner: ComposerChainProps): SubagentReadOnlyMatc
const subagent = owner.session?.subagent
if (subagent === undefined || subagent === null) return null
if (subagent.address.mode === 'one-shot') return { reason: 'one-shot' }
return subagent.parentAvailable ? null : { reason: 'parent-unavailable' }
if (subagent.parentAvailable) return null
// A RUNNING parent-offline continuable child keeps the default composer:
// its input is disabled there, but the same primary Stop stays available so
// the child can be interrupted. Once it stops, this takeover returns.
return owner.session?.running === true ? null : { reason: 'parent-unavailable' }
}
/**
packages/client/ui-subagent/tests/browser-plugin.spec.ts
+8 -1
View File
@@ -161,19 +161,26 @@ describe('apply', () => {
const select = composerEntry.select as (owner: ComposerChainProps) => SubagentReadOnlyMatch | null
const owner = (
subagent: ConversationSnapshot['subagent'] | undefined,
running = false,
): ComposerChainProps => ({
interactions: [],
session: subagent === undefined
? undefined
: ({ subagent } as unknown as ConversationSnapshot),
: ({ subagent, running } as unknown as ConversationSnapshot),
})
expect(select(owner(undefined))).toBeNull()
expect(select(owner(null))).toBeNull()
expect(select(owner({ address: { ...address, mode: 'one-shot' }, parentAvailable: true })))
.toEqual({ reason: 'one-shot' })
// One-shot stays read-only even while running: it has no stop action.
expect(select(owner({ address: { ...address, mode: 'one-shot' }, parentAvailable: true }, true)))
.toEqual({ reason: 'one-shot' })
expect(select(owner({ address, parentAvailable: true }))).toBeNull()
expect(select(owner({ address, parentAvailable: false })))
.toEqual({ reason: 'parent-unavailable' })
// A RUNNING parent-offline continuable yields the default composer, whose
// disabled input still carries the primary Stop; stopped, it takes back over.
expect(select(owner({ address, parentAvailable: false }, true))).toBeNull()
})
})
packages/cordis/tool-cordis/src/api-catalog.ts
+16
View File
@@ -928,6 +928,10 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
signature: 'async followup( parent: Agent, childId: SessionId, content: ContentBlock[], options: SubagentFollowupOptions, ): Promise<MessageId>',
jsDoc: '/**\n * Deliver one later message to a continuable child as its next FIFO turn. A\n * resident child\'s Agent inbox accepts it directly (waking a `waiting`\n * Activation), while an absent one is cold-resumed from its persisted\n * Session. The Agent inbox is the only queue, so every accepted message has\n * one observable order.\n * @param parent - the exact live direct parent authorizing this delivery.\n * @param childId - durable child session id.\n * @param content - user-role content to deliver.\n * @param options - durable provenance and caller cancellation, which stops the\n * operation only before inbox acceptance.\n * @returns the accepted message\'s inbox id.\n * @throws when continuation services are unavailable, parent authority is\n * rejected, or the message was not admitted.\n */',
},
{
signature: 'interrupt(targetSessionId: SessionId, authority: SubagentInterruptAuthority): void',
jsDoc: '/**\n * Interrupt one live continuable child\'s current turn under a human parent\n * address or an exact live ancestor Agent. Fire-and-return: the cancel\n * signal is issued before this returns, but the target may keep running\n * until it observes the signal. Unclaimed pending inbox work, the Activation,\n * and published descendants are preserved; claimed work is not requeued.\n * Once the interrupted driver is idle, a waking send resumes the parked FIFO\n * queue. An absent target — including a one-shot or unknown id —\n * is an accepted no-op, as is a manager-less composition, which cannot own a\n * live Activation.\n * @param targetSessionId - the durable child session id to interrupt.\n * @param authority - the human parent address or exact live ancestor Agent.\n * @throws {SubagentError} `UNAUTHORIZED` when the authority does not own the\n * live target.\n */',
},
{
signature: 'async reportFrom( child: Agent, content: ContentBlock[], options: SubagentReportOptions, ): Promise<MessageId>',
jsDoc: '/**\n * Deliver selected content from one live continuable child to its durable\n * direct parent. The child is the authority credential; callers cannot name a\n * recipient. Reporting does not conclude the child\'s turn or Activation.\n * @param child - exact live reporting child.\n * @param content - selected model-facing content.\n * @param options - parent scheduling and pre-acceptance cancellation.\n * @returns the stable identity of the parent-accepted message.\n * @throws when continuation services are unavailable, sender authorization\n * fails, or the direct parent is not live.\n */',
@@ -944,6 +948,10 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
signature: 'listChildren(parentSessionId: SessionId, signal?: AbortSignal): Promise<SubagentListEntry[]>',
jsDoc: '/**\n * Enumerate the parent\'s direct session-backed subagents without loading or\n * resuming an Agent and without any query seam: the listing merges the live\n * session store with optional session persistence (live-preferred) and\n * serves each child\'s durable mode/label from the registered `subagent`\n * projection unit down a three-rung ladder — the registry\'s watermark\n * snapshot for a live child; for a cold one, a durable projection-cache\n * row when the optional cache serves an own-suffix identity (its `seq`\n * gate proves the value postdates the fork seed, where a child\'s own\n * descriptor is immutable once appended), else one persistence inspection\n * folded through the registry. The\n * projection fold is the single classification authority; per-child\n * diagnostics relay a fold that served no identity or a failed inspection,\n * never a list-time descriptor parse. Absent persistence, enumeration is\n * live-only (a cold child cannot be resumed then either, so its absence is\n * capability absence, not an error). This service consults no Agent\n * registrations, Activations, or providers.\n *\n * Every persistence read receives `signal`, and the listing rechecks\n * cancellation around each of those awaits. Read rejections that settle\n * after an abort become a stable `SubagentError` with code `CANCELLED`.\n * @param parentSessionId - parent session whose direct children are listed.\n * @param signal - caller-owned cancellation forwarded to persistence reads\n * and observed around every read await.\n * @returns children and per-child diagnostics ordered by `createdAt`, then id.\n * @throws {@link SubagentError} when the projection registry or the session\n * store is not mounted, or the caller cancels the listing.\n */',
},
{
signature: 'listDescendants(rootSessionId: SessionId, signal?: AbortSignal): Promise<SubagentDescendantListEntry[]>',
jsDoc: '/**\n * Enumerate the root\'s complete session-backed subagent tree in stable\n * pre-order from one live-preferred corpus, without loading or resuming an\n * Agent. Ordinary sessions and one-shot children remain traversal nodes so\n * continuable descendants below them are discovered; each returned entry\n * adds its durable `parentId` and root-relative `depth`. Identity resolution,\n * diagnostics, optional persistence, and cancellation follow the same\n * projection-backed contract as {@link listChildren}.\n * @param rootSessionId - session whose complete descendant tree is listed.\n * @param signal - caller-owned cancellation forwarded to persistence reads\n * and observed around every read await.\n * @returns children and per-candidate diagnostics with tree position, in\n * stable pre-order.\n * @throws {@link SubagentError} under the same conditions as {@link listChildren}.\n */',
},
{
signature: 'registerProvider(provider: SubagentProvider): () => void',
jsDoc: '/**\n * Register a provider under its name. Registration is effect-scoped and HMR\n * safe; removing a provider blocks new starts but does not revoke runs that\n * were already returned to their holders.\n * @param provider - the trusted provider implementation.\n * @returns the exact Cordis effect disposer.\n */',
@@ -2779,6 +2787,10 @@ export const TYPE_API: readonly TypeApiEntry[] = [
name: 'SubagentCapabilities',
declaration: 'export interface SubagentCapabilities {\n readonly outputSchema: boolean;\n readonly depthLimit: boolean;\n readonly toolFilter: boolean;\n readonly persona: boolean;\n}',
},
{
name: 'SubagentDescendantListEntry',
declaration: 'export type SubagentDescendantListEntry = SubagentListEntry & {\n readonly parentId: SessionId;\n readonly depth: number;\n};',
},
{
name: 'SubagentDescriptorData',
declaration: 'export type SubagentDescriptorData = OneShotSubagentDescriptorData | ContinuableSubagentDescriptorData;',
@@ -2787,6 +2799,10 @@ export const TYPE_API: readonly TypeApiEntry[] = [
name: 'SubagentFollowupOptions',
declaration: 'export interface SubagentFollowupOptions {\n readonly source: MessageSource;\n readonly signal: AbortSignal;\n}',
},
{
name: 'SubagentInterruptAuthority',
declaration: 'export type SubagentInterruptAuthority = {\n readonly kind: \'user\';\n readonly parentSessionId: SessionId;\n} | {\n readonly kind: \'ancestor\';\n readonly agent: Agent;\n};',
},
{
name: 'SubagentListEntry',
declaration: 'export type SubagentListEntry = {\n readonly kind: \'child\';\n readonly id: SessionId;\n readonly activity: \'running\' | \'inactive\';\n readonly hasChildren: boolean;\n} & ({\n readonly mode: \'one-shot\';\n readonly label?: string;\n} | {\n readonly mode: \'continuable\';\n readonly label: string;\n}) | {\n readonly kind: \'diagnostic\';\n readonly id: SessionId;\n readonly reason: \'corrupt\' | \'unsupported\' | \'unavailable\';\n};',
packages/core/tools/tests/gen-tool-catalog.spec.ts
+2 -1
View File
@@ -23,7 +23,7 @@ describe('gen-tool-catalog collectToolCatalog', () => {
it('boots every shipped tool package and harvests its model-facing schemas', async () => {
const catalog = await collectToolCatalog()
const names = catalog.flatMap(entry => entry.schemas.map(s => s.name)).sort()
expect(names).toEqual(['ask_user_question', 'bash', 'bash', 'cordis_inspect', 'cordis_mount', 'cordis_unmount', 'create_goal', 'edit', 'exit_plan_mode', 'get_goal', 'glob', 'grep', 'list_agents', 'lsp', 'pwsh', 'ralph', 'read', 'report', 'run_code', 'send_message', 'session_event_read', 'session_event_search', 'session_event_trace', 'session_search', 'session_trace', 'skill', 'str_replace_editor', 'subagent', 'task_kill', 'task_list', 'task_output', 'terminal_close', 'terminal_list', 'terminal_open', 'terminal_read', 'terminal_send', 'terminal_signal', 'todo_write', 'update_goal', 'web_fetch', 'web_search', 'workflow', 'write'])
expect(names).toEqual(['ask_user_question', 'bash', 'bash', 'cordis_inspect', 'cordis_mount', 'cordis_unmount', 'create_goal', 'edit', 'exit_plan_mode', 'get_goal', 'glob', 'grep', 'interrupt_agent', 'list_agents', 'lsp', 'pwsh', 'ralph', 'read', 'report', 'run_code', 'send_message', 'session_event_read', 'session_event_search', 'session_event_trace', 'session_search', 'session_trace', 'skill', 'str_replace_editor', 'subagent', 'task_kill', 'task_list', 'task_output', 'terminal_close', 'terminal_list', 'terminal_open', 'terminal_read', 'terminal_send', 'terminal_signal', 'todo_write', 'update_goal', 'web_fetch', 'web_search', 'workflow', 'write'])
// Every tool carries a JSON-Schema `parameters` object (what the model sees).
for (const entry of catalog) {
for (const schema of entry.schemas) {
@@ -49,6 +49,7 @@ describe('gen-tool-catalog collectToolCatalog', () => {
expect(bash?.sources.bash).toBe('packages/bash/tool-bash/src/index.ts')
const control = catalog.find(entry => entry.pkg === '@deepseek-ai/dsh-tool-subagent-control')
expect(control?.sources).toEqual({
interrupt_agent: 'packages/subagent/tool-subagent-control/src/index.ts',
list_agents: 'packages/subagent/tool-subagent-control/src/list-agents.ts',
send_message: 'packages/subagent/tool-subagent-control/src/index.ts',
})
packages/host/apiproxy/src/api-proxy.ts
+25
View File
@@ -2013,6 +2013,31 @@ export function createApiProxy(ctx: Context, defaults: ApiProxyDefaults): ApiPro
return subagentPromptError(request, error, signal)
}
},
// Deliberately no catalog, history, persistence, or parent Agent lookup:
// the core primitive alone authorizes the durable address against the
// live Activation, which is what keeps a live child interruptible while
// its parent Agent is offline. Absent targets are accepted no-ops there.
interrupt(request) {
const { parentSessionId, childSessionId } = request.payload
try {
ctx.subagents.interrupt(childSessionId, { kind: 'user', parentSessionId })
} catch (error: unknown) {
if (error instanceof SubagentError && error.code === 'UNAUTHORIZED') {
return Promise.resolve(err(request, {
code: 'subagent-unauthorized',
message: 'subagent does not belong to this parent',
details: { childSessionId },
}))
}
return Promise.resolve(err(request, {
code: 'internal',
message: 'subagent interrupt failed',
details: {},
}))
}
return Promise.resolve(ok(request, { accepted: true as const }))
},
},
workspace: {
packages/host/apiproxy/src/api/index.ts
+2 -1
View File
@@ -42,7 +42,8 @@ export type {
} from './sessions.ts'
export type { DirectoryEntry, DirectoryListing, HostApi } from './host.ts'
export type {
SubagentAddress, SubagentCatalog, SubagentListEntry, SubagentPromptReceipt, SubagentsApi,
SubagentAddress, SubagentCatalog, SubagentInterruptReceipt, SubagentListEntry,
SubagentPromptReceipt, SubagentsApi,
} from './subagents.ts'
export type { WorkspaceApi, WorkspaceId, WorkspaceView } from './workspace.ts'
export type { CommandsApi, CommandDescriptor } from './commands.ts'
packages/host/apiproxy/src/api/rpc-map.ts
+1
View File
@@ -36,6 +36,7 @@ export interface RpcMethodMap {
'subagent.list': SubagentsApi['list']
'subagent.history': SubagentsApi['history']
'subagent.prompt': SubagentsApi['prompt']
'subagent.interrupt': SubagentsApi['interrupt']
'host.describe': HostApi['describe']
'host.pickDirectory': HostApi['pickDirectory']
'host.listDirectory': HostApi['listDirectory']
packages/host/apiproxy/src/api/subagents.schema.ts
+12
View File
@@ -69,6 +69,18 @@ export const subagentPromptRequestSchema = z.object({
content: z.array(contentBlockSchema),
}) as unknown as z.ZodType<RequestPayload<'subagent.prompt'>>
/** subagent.interrupt request payload. */
export const subagentInterruptRequestSchema = z.object({
parentSessionId: sessionIdSchema,
childSessionId: sessionIdSchema,
mode: z.literal('continuable'),
}) satisfies z.ZodType<Wire<RequestPayload<'subagent.interrupt'>>>
/** subagent.interrupt response value. */
export const subagentInterruptValueSchema = z.object({
accepted: z.literal(true),
}) satisfies z.ZodType<Wire<ResponseValue<'subagent.interrupt'>>>
const messageIdSchema = z.string() as unknown as z.ZodType<MessageId>
/** subagent.prompt response value. */
packages/host/apiproxy/src/api/subagents.ts
+18
View File
@@ -40,6 +40,11 @@ export interface SubagentPromptReceipt {
messageId: MessageId
}
/** Uniform acknowledgement that one interrupt request was admitted. */
export interface SubagentInterruptReceipt {
accepted: true
}
/** Durable parent/child address that selects subagent transport in the client. */
export type SubagentAddress =
& {
@@ -94,4 +99,17 @@ export interface SubagentsApi {
>,
signal: AbortSignal,
): Promise<RpcResponse<SubagentPromptReceipt>>
/**
* Interrupts a live continuable child's current turn under the address's
* durable direct-parent authority, without requiring a live parent Agent,
* consulting the catalog, or resuming anything. Fire-and-return: `accepted`
* acknowledges the admitted cancel signal, not target quiescence, so the
* child may remain visibly running briefly. Unclaimed queued follow-ups are
* kept and parked; an absent, idle, or already-completed target is likewise
* `accepted`.
*/
interrupt(
request: RpcRequest<Extract<SubagentAddress, { mode: 'continuable' }>>,
): Promise<RpcResponse<SubagentInterruptReceipt>>
}
packages/host/apiproxy/src/fetch/client.ts
+4
View File
@@ -58,6 +58,7 @@ import {
import { llmDiscoverModelsValueSchema, llmModelsValueSchema, llmProvidersValueSchema } from '../api/llm.schema.ts'
import {
subagentHistoryValueSchema,
subagentInterruptValueSchema,
subagentListValueSchema,
subagentPromptValueSchema,
} from '../api/subagents.schema.ts'
@@ -96,6 +97,7 @@ export interface IApiClient {
list(payload: RequestPayload<'subagent.list'>, signal?: AbortSignal): Promise<RpcResponse<ResponseValue<'subagent.list'>>>
history(payload: RequestPayload<'subagent.history'>, signal?: AbortSignal): Promise<RpcResponse<ResponseValue<'subagent.history'>>>
prompt(payload: RequestPayload<'subagent.prompt'>, signal?: AbortSignal): Promise<RpcResponse<ResponseValue<'subagent.prompt'>>>
interrupt(payload: RequestPayload<'subagent.interrupt'>, signal?: AbortSignal): Promise<RpcResponse<ResponseValue<'subagent.interrupt'>>>
}
host: {
describe(payload: RequestPayload<'host.describe'>, signal?: AbortSignal): Promise<RpcResponse<ResponseValue<'host.describe'>>>
@@ -171,6 +173,7 @@ const UNARY_VALUE_SCHEMAS: { [K in keyof RpcMethodMap]: z.ZodType<Wire<ResponseV
'subagent.list': subagentListValueSchema,
'subagent.history': subagentHistoryValueSchema,
'subagent.prompt': subagentPromptValueSchema,
'subagent.interrupt': subagentInterruptValueSchema,
'host.describe': hostDescribeValueSchema,
'host.pickDirectory': hostPickDirectoryValueSchema,
'host.listDirectory': hostListDirectoryValueSchema,
@@ -407,6 +410,7 @@ export abstract class AbstractApiClient implements IApiClient {
list: (payload, signal) => this.callUnary('subagent.list', payload, signal),
history: (payload, signal) => this.callUnary('subagent.history', payload, signal),
prompt: (payload, signal) => this.callUnary('subagent.prompt', payload, signal),
interrupt: (payload, signal) => this.callUnary('subagent.interrupt', payload, signal),
}
readonly host: IApiClient['host'] = {
packages/host/apiproxy/src/fetch/handler.ts
+2
View File
@@ -60,6 +60,7 @@ import {
import { llmDiscoverModelsRequestSchema, llmModelsRequestSchema, llmProvidersRequestSchema } from '../api/llm.schema.ts'
import {
subagentHistoryRequestSchema,
subagentInterruptRequestSchema,
subagentListRequestSchema,
subagentPromptRequestSchema,
} from '../api/subagents.schema.ts'
@@ -95,6 +96,7 @@ const UNARY_ROUTES: UnaryRoutes = {
'subagent.list': { schema: subagentListRequestSchema, invoke: (api, r, signal) => api.subagents.list(r, signal) },
'subagent.history': { schema: subagentHistoryRequestSchema, invoke: (api, r, signal) => api.subagents.history(r, signal) },
'subagent.prompt': { schema: subagentPromptRequestSchema, invoke: (api, r, signal) => api.subagents.prompt(r, signal) },
'subagent.interrupt': { schema: subagentInterruptRequestSchema, invoke: (api, r) => api.subagents.interrupt(r) },
'host.describe': { schema: hostDescribeRequestSchema, invoke: (api, r) => api.host.describe(r) },
'host.pickDirectory': { schema: hostPickDirectoryRequestSchema, invoke: (api, r, signal) => api.host.pickDirectory(r, signal) },
'host.listDirectory': { schema: hostListDirectoryRequestSchema, invoke: (api, r, signal) => api.host.listDirectory(r, signal) },
packages/host/apiproxy/tests/api-proxy-subagents.spec.ts
+53 -2
View File
@@ -19,6 +19,7 @@ function bench(options: {
childStatus?: 'idle' | 'running'
entries?: object[]
followupError?: Error
interruptError?: Error
listError?: Error
/** Persistence forgets the child entirely (the vanished-mid-read race). */
storedChild?: false
@@ -53,6 +54,12 @@ function bench(options: {
) => options.followupError === undefined
? Promise.resolve('message-1')
: Promise.reject(options.followupError))
const interrupt = vi.fn((
_targetSessionId: SessionId,
_authority: { kind: 'user'; parentSessionId: SessionId },
) => {
if (options.interruptError !== undefined) throw options.interruptError
})
const childHeader = {
version: 0, id: CHILD, createdAt: 1, cwd: '/proj', parentSession: options.historyParent ?? PARENT,
} satisfies SessionHeader
@@ -72,7 +79,7 @@ function bench(options: {
})
const ctx = new Context()
ctx.provide('agents', { get: getAgent })
ctx.provide('subagents', { listChildren, followup })
ctx.provide('subagents', { listChildren, followup, interrupt })
ctx.provide('sessions', {
get: (id: SessionId) => options.liveChild === true && id === CHILD
? { id: CHILD, header: childHeader, events: childEvents }
@@ -90,7 +97,7 @@ function bench(options: {
const api = createApiProxy(ctx, {
defaultTarget: () => ({ provider: 'p', model: 'm' }), cwd: '/tmp', workspaceRoot: '/tmp',
})
return { api, getAgent, listChildren, inspect, snapshot, restore, followup, parent }
return { api, getAgent, listChildren, inspect, snapshot, restore, followup, interrupt, parent }
}
describe('subagent gateway', () => {
@@ -309,4 +316,48 @@ describe('subagent gateway', () => {
error: { code: 'internal', message: 'subagent prompt failed' },
})
})
it('interrupts through the core primitive alone while the parent Agent is offline', async () => {
const { api, interrupt, getAgent, listChildren, inspect } = bench({ parentLive: false })
const response = await api.subagents.interrupt(request({
parentSessionId: PARENT, childSessionId: CHILD, mode: 'continuable' as const,
}))
expect(response.rpcId).toBe('subagent-rpc')
expect(response.result).toEqual({ ok: true, value: { accepted: true } })
expect(interrupt).toHaveBeenCalledExactlyOnceWith(CHILD, { kind: 'user', parentSessionId: PARENT })
// No parent-registry, catalog, or history dependency: this is what keeps a
// live child interruptible after its parent Agent went offline.
expect(getAgent).not.toHaveBeenCalled()
expect(listChildren).not.toHaveBeenCalled()
expect(inspect).not.toHaveBeenCalled()
})
it('maps interrupt authorization rejection without touching other services', async () => {
const { api, listChildren } = bench({
interruptError: new SubagentError('secret lineage', 'UNAUTHORIZED'),
})
const response = await api.subagents.interrupt(request({
parentSessionId: PARENT, childSessionId: CHILD, mode: 'continuable' as const,
}))
expect(response.result).toEqual({
ok: false,
error: {
code: 'subagent-unauthorized',
message: 'subagent does not belong to this parent',
details: { childSessionId: CHILD },
},
})
expect(listChildren).not.toHaveBeenCalled()
})
it('hides unexpected interrupt failures behind the internal code', async () => {
const { api } = bench({ interruptError: new Error('secret activation state') })
const response = await api.subagents.interrupt(request({
parentSessionId: PARENT, childSessionId: CHILD, mode: 'continuable' as const,
}))
expect(response.result).toEqual({
ok: false,
error: { code: 'internal', message: 'subagent interrupt failed', details: {} },
})
})
})
packages/host/apiproxy/tests/client-handler.spec.ts
+27
View File
@@ -63,6 +63,7 @@ function scriptedApi(overrides: {
list: r => ok(r, { entries: [], parentAvailable: false }),
history: r => ok(r, { events: [], hasMore: false }),
prompt: r => ok(r, { messageId: 'message-1' as never }),
interrupt: r => ok(r, { accepted: true as const }),
...overrides.subagents,
},
host: {
@@ -248,6 +249,32 @@ describe('unary round trip', () => {
}
})
it('round-trips subagent.interrupt and rejects a one-shot or incomplete address', async () => {
const interrupt = vi.fn((r: RpcRequest<unknown>) => ok(r, { accepted: true as const }))
const api = scriptedApi({ subagents: { interrupt } })
const c = client(api)
const accepted = await c.subagents.interrupt({
parentSessionId: sid('parent'), childSessionId: sid('child'), mode: 'continuable',
})
expect(accepted.result).toEqual({ ok: true, value: { accepted: true } })
expect(interrupt).toHaveBeenCalledTimes(1)
// The wire schema owns the mode fence: a one-shot address never reaches the impl.
const oneShot = await c.subagents.interrupt({
parentSessionId: sid('parent'), childSessionId: sid('child'), mode: 'one-shot',
} as never)
expect(oneShot.result.ok).toBe(false)
if (!oneShot.result.ok) expect(oneShot.result.error.code).toBe('bad-request')
const incomplete = await c.subagents.interrupt({
parentSessionId: sid('parent'), mode: 'continuable',
} as never)
expect(incomplete.result.ok).toBe(false)
if (!incomplete.result.ok) expect(incomplete.result.error.code).toBe('bad-request')
expect(interrupt).toHaveBeenCalledTimes(1)
})
it('rejects a method/path mismatch as bad-request', async () => {
const handler = toFetchHandler(scriptedApi())
const body = { type: 'client-request', rpcId: 'r1', method: 'session.create', payload: {} }
packages/host/apiproxy/tests/fetch-carrier.spec.ts
+8
View File
@@ -128,6 +128,9 @@ function fakeApi(overrides: Partial<{ muxFrames: MuxFrame[]; hostFrames: HostFra
result: { ok: true, value: { messageId: 'message-1' as never } },
}
},
async interrupt(request) {
return { rpcId: request.rpcId, result: { ok: true, value: { accepted: true as const } } }
},
},
host: {
async describe(request) {
@@ -433,6 +436,11 @@ describe('unary round trip (handler ⇄ client, no network)', () => {
mode: 'continuable',
content: [],
})).result).toEqual({ ok: true, value: { messageId: 'message-1' } })
expect((await c.subagents.interrupt({
parentSessionId: 'parent' as never,
childSessionId: 'child' as never,
mode: 'continuable',
})).result).toEqual({ ok: true, value: { accepted: true } })
})
it('keeps caller and connection aborts on command.execute', async () => {
packages/subagent/subagent/README.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/subagent/subagent/README.md
README.md: 9d2e38c8730f7b7f26e690aa878a4466fa7c2829
README.zh.md: 341c18617af4d040ec44814fac1ec4502d9b8902
README.md: cd25bace16c91e8b44331dc8e5eab0987607e83a
README.zh.md: 05b1e4dd4c74b6d62df8b0a310534fc9476f3ade
packages/subagent/subagent/README.md
+6 -3
View File
@@ -18,10 +18,12 @@ The [subagent family overview](../README.md) maps implementations and model-faci
| `start(name, request)` | Validate an ordinary caller request, resolve its detached `one-shot` descriptor, then await the provider until a real one-shot child is published. Fulfillment returns a holder-owned `SubagentRun`; rejection means the provider has already cleaned every unpublished startup resource, while post-publication turn or infrastructure faults settle through the run. Continuable children never enter through this operation. |
| `startContinuable(spec)` | Establish one durable continuable child and deliver its initial prompt. Resolves with `{ childId, messageId }` when the child's inbox accepts that prompt, without waiting for the turn to start or for the message to reach the Session log; any earlier failure rejects with no ids and rolls the child back entirely. Requires `ctx.agents`, session persistence, and a provider with the `prepareContinuable` capability. |
| `followup(parent, childId, content, { source, signal })` | Deliver one later message from the exact live direct parent as the child's next FIFO turn, matching `Agent.followup()` terminology, and return the accepted `MessageId`. A resident child's inbox accepts it directly (waking a waiting Activation); an absent one cold-resumes from its persisted Session. Requires `ctx.agents`; cold resume also requires session persistence. |
| `interrupt(targetSessionId, authority)` | Interrupt one live continuable child's current turn under a human durable parent address (`{ kind: 'user', parentSessionId }`) or an exact live ancestor Agent (`{ kind: 'ancestor', agent }`). Admission is synchronous and the effect asynchronous: it issues `Agent.cancel(cause, { keepInbox: true })` and returns without waiting for the target to observe the signal. Unclaimed pending inbox work, the Activation, and published descendants are preserved; work already claimed into the interrupted turn is not requeued. An absent target is an accepted no-op; a wrong parent address or a stale, self-targeting, or non-ancestor caller rejects with `UNAUTHORIZED`. |
| `reportFrom(child, content, { delivery, signal })` | Deliver one selected message from the exact live continuable child to its exact live direct parent and return the accepted stable `MessageId`. Quiet delivery injects context; waking delivery submits one later parent turn. |
| `registerContinuableSetup(contribution)` | Compose an optional deployment capability into each continuable child's unpublished scope, with immediate revocation from resident children. |
| `drainContinuableDescendants(parents)` | Close admission below exact live host-owned parent Agents, stop only their visible continuable descendants, await materializations admitted below those roots through publication or rollback, then release the selected forests child-first. The cutoff lasts until each exact parent leaves the registry; unrelated parent forests and manager-wide admission remain live. |
| `listChildren(parentSessionId, signal?)` | List direct session-backed subagents with their `one-shot`/`continuable` mode, `running`/`inactive` activity, origin-classified one-level `hasChildren` hint, and per-child diagnostics, ordered by `createdAt` then id, without loading or resuming them. Reads the live session store and optional session persistence directly (live-only enumeration when persistence is absent) and requires the mounted `sessionProjections` registry; it does not require `ctx.agents`, the continuation manager, or any query service. |
| `listDescendants(rootSessionId, signal?)` | Flatten the root's complete session tree in stable pre-order from the same live-preferred corpus, adding each subagent entry's durable `parentId` and root-relative `depth`. Ordinary sessions and one-shot children remain traversal nodes so continuable descendants below them are discovered. Identity, diagnostics, dependencies, and cancellation follow `listChildren()`. |
`SubagentStartRequest.label` is an optional short durable display label for a session-backed one-shot child. Model-facing delegation supplies its existing `description`; lower-level callers need not invent presentation metadata. Continuable starts always carry their own required label. `signal` is required and is the canonical cancellation channel for a one-shot `start`. An abort before publication makes `start()` reject after rollback; an abort after publication cancels the returned run's remaining turn work without hiding its id. The request may also select a model, require structured output, cap delegation depth, restrict child tools, or set a child persona. For a continuable start or follow-up, the caller signal owns lookup, materialization, and admission only until inbox acceptance; afterward the manager owns the Activation independently, so later caller cancellation neither cancels the accepted turn nor disposes the child.
@@ -76,7 +78,7 @@ Run events are scoped to the delegating parent. Every listener is independently
Provider additions and removals also emit `subagent/provider-added` and `subagent/provider-removed`. Consumers such as the model-facing tool use those events because Cordis may load sibling plugins concurrently; configuration order does not prove registration order.
Continuable children do not create `SubagentRun` or Tasks. The continuation manager directly owns one process-local Activation and retained `AgentHandle` per resident child Session, uses the Agent inbox as the only FIFO, and cold-resumes from the durable descriptor. Exact live direct-parent identity authorizes parent-to-child delivery. Exact live child identity authorizes reports; the manager derives the recipient from durable `parentSession`, and `MessageSource` remains provenance rather than authority.
Continuable children do not create `SubagentRun` or Tasks. The continuation manager directly owns one process-local Activation and retained `AgentHandle` per resident child Session, uses the Agent inbox as the only FIFO, and cold-resumes from the durable descriptor. Exact live direct-parent identity authorizes parent-to-child delivery. Exact live child identity authorizes reports; the manager derives the recipient from durable `parentSession`, and `MessageSource` remains provenance rather than authority. Interrupt authority is deliberately wider than delivery authority: a human presents the durable direct-parent address so a live child stays stoppable while its parent Agent is offline, and any exact live ancestor recorded in the Activation's materialization lineage may stop its descendant, because stopping a turn is idempotent and delivers no content.
When `ctx.sessionProjections` is available, the service registers two projection units. `subagentTiming` resets at each descriptor so a fork seed's ancestor work cannot enter the child's total, then accumulates `turn/start` → `turn/end` active time and retains same-cut `active.since` and `active.through` bounds for an open turn; while that turn remains open, `active.through` follows the latest folded event, giving an inactive consumer a conservative crash bound without mixing in newer session metadata. `subagent` folds the durable identity — mode plus creation label — from `subagent/descriptor` events with the same last-wins reset discipline, so a fork seed's ancestor descriptor stands only until the child's own overrides it; a malformed or unrecognized-version payload folds to the serializable `null` sentinel — indistinguishable from a log with no descriptor, and surviving every JSON push frame so a consumer replaces a stale identity instead of keeping it — and never throws.
@@ -84,7 +86,7 @@ When `ctx.sessionProjections` is available, the service registers two projection
## Collection model
The model-facing tool collects synchronously by default: it awaits the child result and disposes the run before returning. One-shot background delegation registers a plain Task in the tool, whose generic status, collection, and cancellation tools own later interaction, and persists its model-supplied `description` as the optional display label. Continuable background delegation calls `ctx.subagents.startContinuable()` and returns only the durable child id; the child owns its own turns from inbox acceptance, so there is no Task, no result promise, and no public subagent cancellation — a caller sends later work with the `send_message` follow-up tool, and the durable child Session remains the source of the child's detailed output. The continuation manager exists only while `ctx.agents` is available, and session persistence is resolved per continuation operation. Independently, `listChildren()` enumerates the live-preferred merge of the live session store and optional session persistence — live-only when persistence is absent, since a cold child cannot be resumed then either — and serves each child's durable mode/label from the registered `subagent` projection unit: the registry's watermark snapshot for a live child; for a cold one, a durable projection-cache row when it serves an own-suffix identity — its `seq` gate proves the value postdates the fork seed, where a child's own descriptor is immutable once appended — else one bounded-concurrency persistence inspection folded through the registry, whose result must still name the enumerated lifecycle (a re-published id degrades to a `corrupt` diagnostic). A throwing cache read renders no verdict — the cache is derived data — and silently falls through to that authoritative re-fold. The projection fold is the single classification authority; listing parses no descriptor itself. A served identity produces a child row; a settled candidate whose fold served no identity is a `corrupt` diagnostic, a failed inspection is a transient `unavailable` retried on the next listing, and a running candidate without an identity yet is omitted (the creation window before its descriptor is appended). It never consults the continuation manager, Agent registrations, Activations, or providers. Each child row derives its read-time `hasChildren` hint from merged headers carrying durable `origin: 'subagent'`; it does not read descendant event logs, and the descriptor-backed child catalog remains authoritative when expanded. Service consumers such as a UI can retain both modes and choose a fallback for an unlabeled one-shot child; the model-facing `list_agents` tool projects only `continuable` entries and maps service activity to its existing `running`/`complete` vocabulary. The listing forwards the caller's signal to every persistence read, checks cancellation around each of those awaits, and reports every observed abort as `SubagentError` code `CANCELLED`; an unmounted projection registry fails loud with `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE`, and a missing session store with `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`. See the [background subagent tasks Agent Note](../../../.agents/notes/implemented/feature/2026-07-08-background-subagent-tasks.md), the [continuable background subagents Agent Note](../../../.agents/notes/implemented/feature/2026-07-21-continuable-background-subagents.md), the [durable catalog Agent Note](../../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md), the [merged-service Agent Note](../../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md), the [capability-seam Agent Note](../../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md), and `src/types.ts` for the complete contracts.
The model-facing tool collects synchronously by default: it awaits the child result and disposes the run before returning. One-shot background delegation registers a plain Task in the tool, whose generic status, collection, and cancellation tools own later interaction, and persists its model-supplied `description` as the optional display label. Continuable background delegation calls `ctx.subagents.startContinuable()` and returns only the durable child id; the child owns its own turns from inbox acceptance, so there is no Task and no result promise — a caller sends later work with the `send_message` follow-up tool, and `interrupt()` stops only the current turn without disposing the child, and the durable child Session remains the source of the child's detailed output. The continuation manager exists only while `ctx.agents` is available, and session persistence is resolved per continuation operation. Independently, `listChildren()` enumerates the live-preferred merge of the live session store and optional session persistence — live-only when persistence is absent, since a cold child cannot be resumed then either — and serves each child's durable mode/label from the registered `subagent` projection unit: the registry's watermark snapshot for a live child; for a cold one, a durable projection-cache row when it serves an own-suffix identity — its `seq` gate proves the value postdates the fork seed, where a child's own descriptor is immutable once appended — else one bounded-concurrency persistence inspection folded through the registry, whose result must still name the enumerated lifecycle (a re-published id degrades to a `corrupt` diagnostic). A throwing cache read renders no verdict — the cache is derived data — and silently falls through to that authoritative re-fold. The projection fold is the single classification authority; listing parses no descriptor itself. A served identity produces a child row; a settled candidate whose fold served no identity is a `corrupt` diagnostic, a failed inspection is a transient `unavailable` retried on the next listing, and a running candidate without an identity yet is omitted (the creation window before its descriptor is appended). It never consults the continuation manager, Agent registrations, Activations, or providers. Each child row derives its read-time `hasChildren` hint from merged headers carrying durable `origin: 'subagent'`; it does not read descendant event logs, and the descriptor-backed child catalog remains authoritative when expanded. Service consumers such as a UI can retain both modes and choose a fallback for an unlabeled one-shot child; the model-facing `list_agents` tool projects only `continuable` entries and refines status through the live Agent registry (`running`/`idle`/`complete`) and walks `listDescendants()` for its `descendants` scope. The listing forwards the caller's signal to every persistence read, checks cancellation around each of those awaits, and reports every observed abort as `SubagentError` code `CANCELLED`; an unmounted projection registry fails loud with `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE`, and a missing session store with `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE`. See the [background subagent tasks Agent Note](../../../.agents/notes/implemented/feature/2026-07-08-background-subagent-tasks.md), the [continuable background subagents Agent Note](../../../.agents/notes/implemented/feature/2026-07-21-continuable-background-subagents.md), the [durable catalog Agent Note](../../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md), the [merged-service Agent Note](../../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md), the [capability-seam Agent Note](../../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md), and `src/types.ts` for the complete contracts.
Continuable Activations await a best-effort final session flush without treating listener participation as durability confirmation. One-shot runs retain best-effort session checkpointing, so a completed one-shot child is discoverable after disposal only when its session actually reached persistence; the service does not invent a catalog entry from Task history when that checkpoint is absent.
@@ -99,8 +101,9 @@ No direct invalidation; the named consumers own any request-prefix changes.
## Known Limitations and Deferred Work
- **ACP children remain one-shot and are not trace-enumerable** — an ACP run has no local child session in the parent's session corpus. An ACP `prepareContinuable` requires persisting the remote session id in provider-specific descriptor data and a per-child continuation advertisement, since ACP `loadSession` support is negotiated per child rather than established by the method's presence. Remote providers also require a separate Activation ownership contract with equivalent authenticated control and child-first quiescence before they support continuable children.
- **No host-user continuation** — `followup()` requires the exact live direct parent. A future host adapter needs a concrete authenticated interaction before the seam gains a separate user capability.
- **No host-user continuation** — `followup()` requires the exact live direct parent. Only `interrupt()` accepts a durable parent-address user authority, because stopping a turn is idempotent and delivers no content; a future host adapter needs a concrete authenticated interaction before the seam gains a user delivery capability.
- **No current-turn steering** — continuable messages and waking reports enqueue later turns; neither redirects an open turn.
- **Wake gap during cancellation convergence** — a waking follow-up accepted after the interrupt signal is issued but before the active driver becomes idle remains queued until another waking send. Issue #1838 owns the agent-loop wake latch, which also affects ordinary session cancellation.
- **Process-local residency** — the Activation inbox and ownership graph do not coordinate two harness processes; concurrent access to one persistence store still requires a durable mailbox and cross-process lease protocol.
- **No replay of accepted-but-unlogged messages** — only messages written to the child Session log are reconstructable with their admitted provenance. A crash may lose an accepted initial prompt or follow-up that never reached the log; a later authorized message can cold-resume the child, but the lost message is not replayed automatically.
- **No durable report mailbox** — reports require a live direct parent and provide acceptance identity rather than exactly-once delivery or a read receipt.
packages/subagent/subagent/README.zh.md
+6 -3
View File
@@ -18,10 +18,12 @@ subagent seam 允许一个 agent（智能体）通过具名提供方把工作委
| `start(name, request)` | 校验普通调用方请求，解析其分离的 `one-shot` 描述符，然后等待提供方，直到真实的一次性子 agent 发布。兑现时返回由持有方拥有的 `SubagentRun`；拒绝表示提供方已清理所有未发布的启动资源，而发布后的轮次或基础设施故障会通过该 run 结算。可继续子 agent 绝不通过此操作进入。 |
| `startContinuable(spec)` | 建立一个持久化可继续子 agent，并投递其初始提示词。子 agent 的 inbox 接受该提示词时，兑现为 `{ childId, messageId }`，无需等待轮次开始或消息写入 Session 日志；此前任何失败都会以无 id 拒绝，并完全回滚该子 agent。要求 `ctx.agents`、会话持久化以及具备 `prepareContinuable` 能力的提供方。 |
| `followup(parent, childId, content, { source, signal })` | 将来自确切在线直接父级的一条后续消息作为子 agent 的下一个 FIFO 轮次投递，术语与 `Agent.followup()` 一致，并返回被接受的 `MessageId`。驻留中的子 agent 由其 inbox 直接接受（唤醒处于 waiting 的 Activation）；不驻留的则从其持久化 Session 冷恢复。要求 `ctx.agents`；冷恢复还要求会话持久化。 |
| `interrupt(targetSessionId, authority)` | 以人类持久化 parent 地址（`{ kind: 'user', parentSessionId }`）或确切在线 ancestor Agent（`{ kind: 'ancestor', agent }`）为授权，中断一个在线可继续 child 的当前轮次。准入同步完成、生效异步进行：它发出 `Agent.cancel(cause, { keepInbox: true })` 后立即返回，不等待目标观察到信号。尚未领取的待处理 inbox 工作、Activation 与已发布的后代均保持不变；已被领取进入中断轮次的工作不会重新入队。目标不存在时接受为 no-op；错误的 parent 地址以及过期、指向自身或非祖先调用方以 `UNAUTHORIZED` 拒绝。 |
| `reportFrom(child, content, { delivery, signal })` | 从确切在线可继续 child 向其确切在线直接 parent 投递一条选中消息，并返回已接受的稳定 `MessageId`。静默投递会注入上下文；唤醒投递会提交一个后续 parent 轮次。 |
| `registerContinuableSetup(contribution)` | 把一项可选部署能力组合到每个可继续 child 尚未发布的作用域中，并支持从驻留 child 立即撤销。 |
| `drainContinuableDescendants(parents)` | 在由 host 确切拥有的在线 parent Agent 之下关闭准入，只停止其可见的可继续后代，等待在这些根之下已获准的物化过程完成发布或回滚，再按 child-first 顺序释放所选森林。该截止状态会持续到每个确切 parent 离开注册表；无关的 parent 森林和管理器全局准入保持在线。 |
| `listChildren(parentSessionId, signal?)` | 按 `createdAt` 再按 id 的顺序列出由会话支撑的直接 subagent，包括其 `one-shot`／`continuable` 模式、`running`／`inactive` 活动状态、基于 origin 分类的一层 `hasChildren` 提示与逐 child diagnostic，且不会加载或恢复它们。直接读取在线会话存储与可选的会话持久化（持久化缺席时仅枚举在线 child），并要求已挂载 `sessionProjections` 注册表；不要求 `ctx.agents`、继续执行管理器或任何查询服务。 |
| `listDescendants(rootSessionId, signal?)` | 从同一份实时优先语料按稳定 pre-order 展平根的完整会话树，并为每个 subagent 条目附加持久 `parentId` 与相对根的 `depth`。普通会话与一次性 child 仍作为遍历节点，因此其下的可继续后代仍可发现。身份、diagnostic、依赖与取消契约均沿用 `listChildren()`。 |
`SubagentStartRequest.label` 是由会话支撑的一次性 child 所使用的可选简短持久化显示标签。面向模型的委派会提供其已有的 `description`；底层调用方无需凭空构造展示元数据。可继续启动始终携带自身的必填标签。`signal` 是必填项，也是一次性 `start` 的规范取消通道。发布前中止会使 `start()` 在回滚后拒绝；发布后中止会取消已返回 run 的剩余轮次工作，但不会隐藏其 id。请求还可以选择模型、要求结构化输出、限制委派深度、约束子 agent 工具或设置子 agent persona。对于可继续启动或后续操作，调用方信号只在 inbox 接受之前掌管查找、物化和准入；此后由管理器独立拥有 Activation，因此调用方后续取消既不会取消已接受的轮次，也不会 dispose（资源释放）子 agent。
@@ -76,7 +78,7 @@ subagent seam 允许一个 agent（智能体）通过具名提供方把工作委
提供方新增和移除还会发出 `subagent/provider-added` 与 `subagent/provider-removed`。面向模型的工具等消费方使用这些事件，因为 Cordis 可能并发加载同级插件；配置顺序不能证明注册顺序。
可继续子级不会创建 `SubagentRun` 或 Task。继续执行管理器为每个驻留子 Session 直接拥有一个仅存在于当前进程的 Activation 和一个留存的 `AgentHandle`，使用 Agent inbox 作为唯一 FIFO，并从持久化描述符冷恢复。父到子投递由确切在线的直接父级身份授权。上报则由确切在线的子级身份授权；管理器根据持久化的 `parentSession` 推导接收方，`MessageSource` 仍只表示来源，不表示权限。
可继续子级不会创建 `SubagentRun` 或 Task。继续执行管理器为每个驻留子 Session 直接拥有一个仅存在于当前进程的 Activation 和一个留存的 `AgentHandle`，使用 Agent inbox 作为唯一 FIFO，并从持久化描述符冷恢复。父到子投递由确切在线的直接父级身份授权。上报则由确切在线的子级身份授权；管理器根据持久化的 `parentSession` 推导接收方，`MessageSource` 仍只表示来源，不表示权限。中断权限被刻意设计得比投递权限更宽：人类出示持久化直接 parent 地址，因此即使 parent Agent 离线，在线 child 仍可被停止；Activation 物化时记录的任何确切在线 ancestor 也可以停止其后代，因为停止一个轮次是幂等的，且不投递任何内容。
当 `ctx.sessionProjections` 可用时，服务会注册两个投影单元。`subagentTiming` 会在每个描述符处重置，使 fork 种子中的祖先工作不会计入 child 总量，随后累加 `turn/start` → `turn/end` 活跃时间，并为未结束的轮次保留同一切面的 `active.since` 和 `active.through` 边界；在该轮次保持未结束期间，`active.through` 会跟随最近折叠的事件，从而为 inactive 消费方提供保守的崩溃上界，又不会混入更新的会话元数据。`subagent` 以同样的 last-wins 重置纪律从 `subagent/descriptor` 事件折叠持久化身份——模式与创建标签——因此 fork 种子中的祖先描述符只在 child 自身的描述符覆盖之前有效；畸形或版本不识别的载荷折叠为可序列化的 `null` 哨兵——与没有描述符的日志不可区分，且能完好通过每个 JSON 推送帧，让消费方以之替换掉手中过时的身份而非永久滞留——绝不抛错。
@@ -84,7 +86,7 @@ subagent seam 允许一个 agent（智能体）通过具名提供方把工作委
## 收集模型
面向模型的工具默认同步收集：先等待子 agent 结果，再 dispose 运行，然后才返回。一次性后台委派会在工具中注册普通 Task，其通用状态、收集和取消工具负责后续交互，并将模型提供的 `description` 持久化为可选显示标签。可继续后台委派会调用 `ctx.subagents.startContinuable()`，只返回持久化子 agent id；子 agent 自 inbox 接受起就拥有自己的轮次，因此没有 Task、没有结果 promise，也没有公开的子 agent 取消操作——调用方通过 `send_message` 后续操作工具发送后续工作，而持久化子 agent Session 仍是子 agent 详细输出的来源。只有 `ctx.agents` 可用时，继续执行管理器才会存在，而会话持久化按每项继续执行操作解析。与此独立，`listChildren()` 枚举在线会话存储与可选会话持久化的在线优先合并——持久化缺席时仅枚举在线 child，因为那时冷 child 本就无法恢复——并由已注册的 `subagent` 投影单元供给每个 child 的持久化模式与标签：在线 child 取注册表的水位快照；冷 child 先取可选投影缓存的持久化行，且仅当其 `seq` 门证明该值折叠自 child 自身后缀（fork 种子之后——自有描述符一经追加即不可变）才直接采用，否则经一次有界并发的持久化 inspect 再经注册表折叠，且 inspect 结果必须仍指向枚举时的生命周期（同 id 被重新发布的会话降级为 `corrupt` diagnostic）。缓存读取抛错不产生判决——缓存是派生数据——静默落到该权威重折。投影折叠是唯一的分类权威；列表自身不解析任何描述符。取得身份值即产出 child 行；已定局而折叠未产出身份的候选是 `corrupt` diagnostic，inspect 失败是瞬时的 `unavailable`（下次列表重试），运行中而暂无身份值的候选整行省略（描述符尚未追加的创建窗口）。它不查询继续执行管理器、Agent 注册信息、Activation 或提供方。每个 child 行都会根据合并结果中携带持久化 `origin: 'subagent'` 的 header 派生读取时的 `hasChildren` 提示；它不会读取后代事件日志，展开后仍以描述符支撑的 child 目录为权威依据。UI 等服务消费方可以保留两种模式，并为无标签的一次性 child 选择回退展示；面向模型的 `list_agents` 工具只投影 `continuable` 条目，并将服务活动状态映射到现有的 `running`／`complete` 词汇。列表操作会把调用方的取消信号转发到每次持久化读取，在这些 await 前后检查取消，并将每次检测到的中止报告为 `SubagentError` 错误码 `CANCELLED`；投影注册表未挂载则以 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 响亮失败，会话存储缺失则以 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` 响亮失败。完整契约见[后台 subagent 任务 Agent Note](../../../.agents/notes/implemented/feature/2026-07-08-background-subagent-tasks.md)、[可继续后台 subagent Agent Note](../../../.agents/notes/implemented/feature/2026-07-21-continuable-background-subagents.md)、[持久化目录 Agent Note](../../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md)、[服务合并 Agent Note](../../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md)、[能力 seam Agent Note](../../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md)和 `src/types.ts`。
面向模型的工具默认同步收集：先等待子 agent 结果，再 dispose 运行，然后才返回。一次性后台委派会在工具中注册普通 Task，其通用状态、收集和取消工具负责后续交互，并将模型提供的 `description` 持久化为可选显示标签。可继续后台委派会调用 `ctx.subagents.startContinuable()`，只返回持久化子 agent id；子 agent 自 inbox 接受起就拥有自己的轮次，因此没有 Task、也没有结果 promise——调用方通过 `send_message` 后续操作工具发送后续工作，`interrupt()` 只停止当前轮次而不 dispose 子 agent，而持久化子 agent Session 仍是子 agent 详细输出的来源。只有 `ctx.agents` 可用时，继续执行管理器才会存在，而会话持久化按每项继续执行操作解析。与此独立，`listChildren()` 枚举在线会话存储与可选会话持久化的在线优先合并——持久化缺席时仅枚举在线 child，因为那时冷 child 本就无法恢复——并由已注册的 `subagent` 投影单元供给每个 child 的持久化模式与标签：在线 child 取注册表的水位快照；冷 child 先取可选投影缓存的持久化行，且仅当其 `seq` 门证明该值折叠自 child 自身后缀（fork 种子之后——自有描述符一经追加即不可变）才直接采用，否则经一次有界并发的持久化 inspect 再经注册表折叠，且 inspect 结果必须仍指向枚举时的生命周期（同 id 被重新发布的会话降级为 `corrupt` diagnostic）。缓存读取抛错不产生判决——缓存是派生数据——静默落到该权威重折。投影折叠是唯一的分类权威；列表自身不解析任何描述符。取得身份值即产出 child 行；已定局而折叠未产出身份的候选是 `corrupt` diagnostic，inspect 失败是瞬时的 `unavailable`（下次列表重试），运行中而暂无身份值的候选整行省略（描述符尚未追加的创建窗口）。它不查询继续执行管理器、Agent 注册信息、Activation 或提供方。每个 child 行都会根据合并结果中携带持久化 `origin: 'subagent'` 的 header 派生读取时的 `hasChildren` 提示；它不会读取后代事件日志，展开后仍以描述符支撑的 child 目录为权威依据。UI 等服务消费方可以保留两种模式，并为无标签的一次性 child 选择回退展示；面向模型的 `list_agents` 工具只投影 `continuable` 条目，通过在线 Agent 注册表细化状态（`running`／`idle`／`complete`），并在 `descendants` scope 下遍历 `listDescendants()`。列表操作会把调用方的取消信号转发到每次持久化读取，在这些 await 前后检查取消，并将每次检测到的中止报告为 `SubagentError` 错误码 `CANCELLED`；投影注册表未挂载则以 `SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE` 响亮失败，会话存储缺失则以 `SUBAGENT_CONTROL_SESSION_STORE_UNAVAILABLE` 响亮失败。完整契约见[后台 subagent 任务 Agent Note](../../../.agents/notes/implemented/feature/2026-07-08-background-subagent-tasks.md)、[可继续后台 subagent Agent Note](../../../.agents/notes/implemented/feature/2026-07-21-continuable-background-subagents.md)、[持久化目录 Agent Note](../../../.agents/notes/implemented/feature/2026-07-22-durable-subagent-catalog-and-list-agents.md)、[服务合并 Agent Note](../../../.agents/notes/implemented/simplification/2026-07-26-merge-subagent-control-service.md)、[能力 seam Agent Note](../../../.agents/notes/implemented/feature/2026-06-21-subagent-capability-seam.md)和 `src/types.ts`。
可继续 Activation 会等待 best-effort 的最终会话 flush，但不会把 listener 参与视为持久性确认。一次性运行保留尽力执行的会话检查点，因此已完成的一次性 child 只有在其会话确实进入持久化存储时，才可在 dispose 后继续被发现；如果该检查点缺失，服务不会根据 Task 历史虚构目录条目。
@@ -99,8 +101,9 @@ subagent seam 允许一个 agent（智能体）通过具名提供方把工作委
## 已知限制与暂缓事项
- **ACP 子 agent 仍为一次性，且无法通过追踪枚举**：ACP 运行在 parent 会话语料中没有本地 child 会话。ACP 的 `prepareContinuable` 需要在提供方专用描述符数据中持久化远端会话 id，以及逐子 agent 的继续执行能力声明，因为 ACP 的 `loadSession` 支持按子 agent 协商，而不是通过方法是否存在来确定。远程提供方还需要一份独立的 Activation 所有权契约，具备等效的经认证控制和子先于父的停稳保证，才能支持可继续子 agent。
- **无 host-user 继续执行**：`followup()` 要求确切在线直接父级。未来 host 适配器需要具体的经认证交互，才能让该 seam 获得单独的用户能力。
- **无 host-user 继续执行**：`followup()` 要求确切在线直接父级。只有 `interrupt()` 接受持久化 parent 地址形式的用户授权，因为停止一个轮次是幂等的且不投递任何内容；未来 host 适配器需要具体的经认证交互，才能让该 seam 获得用户投递能力。
- **不对当前轮次进行 steering**：可继续消息和唤醒式 report 会排入后续轮次，均不会重定向正在进行的轮次。
- **取消收敛期间存在唤醒缺口**：中断信号发出后、活动 driver 进入 idle 前被接受的唤醒型 follow-up 会保持排队，直到另一条唤醒发送到达。Issue #1838 负责 agent-loop 的唤醒锁存；普通会话取消也受此影响。
- **驻留仅限进程内**：Activation inbox 与所有权图不会在两个 harness 进程之间协调；对单个持久化存储的并发访问仍然需要持久化邮箱和跨进程租约协议。
- **不重放已接受但未记录的消息**：只有写入子 agent Session 日志的消息才能连同其被接受时的来源一起重建。崩溃可能丢失从未写入日志、已被接受的初始提示词或后续消息；此后一条经授权的消息可以冷恢复该子 agent，但丢失的消息不会自动重放。
- **没有持久化的上报 mailbox**：上报需要实时直接父级，提供的是接受标识，不保证恰好一次投递，也不提供已读回执。
packages/subagent/subagent/src/continuation.ts
+72
View File
@@ -103,6 +103,15 @@ export interface ContinuableStart {
readonly messageId: MessageId
}
/**
* Authority under which one interrupt request is admitted. `user` carries the
* durable direct-parent address a human client presented; `ancestor` carries
* the exact live Agent object whose recorded lineage must contain the caller.
*/
export type SubagentInterruptAuthority =
| { readonly kind: 'user'; readonly parentSessionId: SessionId }
| { readonly kind: 'ancestor'; readonly agent: Agent }
/** Options for following up with one continuable child. */
export interface SubagentFollowupOptions {
/** Durable attribution retained on the delivered message; it grants no authority. */
@@ -412,6 +421,69 @@ export class SubagentContinuationManager {
}
}
/**
* Interrupt one live continuable child's current turn. Admission is
* synchronous and the effect is asynchronous: this authorizes the caller,
* requests `Agent.cancel(cause, { keepInbox: true })` on the target, and
* returns without waiting for the target to observe the signal or reach
* quiescence. The Activation, its handle, accepted unclaimed inbox work, and
* already-published descendants are untouched; work already claimed into the
* interrupted turn is not requeued. Once the interrupted driver is idle, a
* waking send resumes the parked queue.
*
* An absent target is an accepted no-op, which uniformly covers natural
* completion races, repeated requests, one-shot ids, and unknown ids without
* consulting the durable catalog. A target whose disposal transaction is
* already open is likewise an accepted no-op after authorization.
* @param targetSessionId - the durable child session id to interrupt.
* @param authority - the human parent address or exact live ancestor Agent.
* @throws {SubagentError} `UNAUTHORIZED` when the presented authority does
* not own the live target: a stale or self-targeting ancestor caller, a
* parent address that is not the live target's durable direct parent, or
* an ancestor outside the target's recorded live lineage.
*/
interrupt(targetSessionId: SessionId, authority: SubagentInterruptAuthority): void {
if (authority.kind === 'ancestor') {
const caller = authority.agent
// A stale caller is rejected even when the target is absent, so a
// replaced same-id Agent can never probe this manager's state.
if (this.ctx.agents.get(caller.id) !== caller) {
throw new SubagentError(
`interrupting "${targetSessionId}" requires the exact live ancestor agent`,
'UNAUTHORIZED',
)
}
if (caller.id === targetSessionId) {
throw new SubagentError(
`agent "${caller.id}" cannot interrupt itself`,
'UNAUTHORIZED',
)
}
}
const activation = this.activations.get(targetSessionId)
if (activation === undefined) return
if (authority.kind === 'user') {
if (activation.handle.agent.session.header.parentSession !== authority.parentSessionId) {
throw new SubagentError(
`subagent "${targetSessionId}" belongs to another parent session`,
'UNAUTHORIZED',
)
}
} else if (!activation.ancestry.has(authority.agent)) {
throw new SubagentError(
`subagent "${targetSessionId}" is not a live descendant of agent "${authority.agent.id}"`,
'UNAUTHORIZED',
)
}
// Disposal already stopped the target with a whole-Activation teardown;
// a second cancel would be a redundant signal on a closing handle.
if (activation.disposal !== undefined) return
activation.handle.agent.cancel(
authority.kind === 'user' ? { kind: 'user' } : { kind: 'parent' },
{ keepInbox: true },
)
}
/**
* Deliver explicitly selected content from one resident continuable child to
* its durable direct parent. Sender authorization, parent resolution, and
packages/subagent/subagent/src/index.ts
+46 -6
View File
@@ -19,9 +19,9 @@
* resident. Continuable children never become a {@link SubagentRun}: the
* continuation manager holds their `AgentHandle` directly and orders every turn
* through the child's own inbox, so providers contribute only the detached
* creation spec and see no handle, turn, or teardown. Direct-child discovery
* reads the live session store and optional session persistence directly and
* does not require that continuation runtime.
* creation spec and see no handle, turn, or teardown. Child and descendant
* discovery read the live session store and optional session persistence
* directly and do not require that continuation runtime.
*
* Same-process providers are trusted typed collaborators. Requests, provider
* descriptors, results, and lifecycle payloads are borrowed immutable values;
@@ -58,12 +58,13 @@ import type {
ContinuableStart,
ContinuableStartSpec,
SubagentFollowupOptions,
SubagentInterruptAuthority,
SubagentReportOptions,
} from './continuation.ts'
import SubagentActivationSetupRegistry from './activation-setup-registry.ts'
import type { ContinuableSetupContribution } from './activation-setup-registry.ts'
import { listChildren as listSubagentChildren } from './list-children.ts'
import type { SubagentListEntry } from './list-children.ts'
import { listChildren as listSubagentChildren, listDescendants as listSubagentDescendants } from './list-children.ts'
import type { SubagentDescendantListEntry, SubagentListEntry } from './list-children.ts'
import { snapshotSubagentDescriptor } from './descriptor.ts'
import { subagentIdentityProjectionDefinition, subagentTimingProjectionDefinition } from './projection.ts'
@@ -111,12 +112,13 @@ export type {
ContinuableStartSpec,
CoordinatorMessageSource,
SubagentFollowupOptions,
SubagentInterruptAuthority,
SubagentReportDelivery,
SubagentReportMessageSource,
SubagentReportOptions,
} from './continuation.ts'
export type { ContinuableSetupContribution } from './activation-setup-registry.ts'
export type { SubagentListEntry } from './list-children.ts'
export type { SubagentDescendantListEntry, SubagentListEntry } from './list-children.ts'
export type { SubagentRunEndInfo, SubagentRunInfo } from './types.ts'
export type { SubagentIdentityProjection, SubagentTimingProjection } from './projection-types.ts'
@@ -231,6 +233,25 @@ export class SubagentService extends Service {
return this.requireContinuations().followup(parent, childId, content, options)
}
/**
* Interrupt one live continuable child's current turn under a human parent
* address or an exact live ancestor Agent. Fire-and-return: the cancel
* signal is issued before this returns, but the target may keep running
* until it observes the signal. Unclaimed pending inbox work, the Activation,
* and published descendants are preserved; claimed work is not requeued.
* Once the interrupted driver is idle, a waking send resumes the parked FIFO
* queue. An absent target — including a one-shot or unknown id —
* is an accepted no-op, as is a manager-less composition, which cannot own a
* live Activation.
* @param targetSessionId - the durable child session id to interrupt.
* @param authority - the human parent address or exact live ancestor Agent.
* @throws {SubagentError} `UNAUTHORIZED` when the authority does not own the
* live target.
*/
interrupt(targetSessionId: SessionId, authority: SubagentInterruptAuthority): void {
this.continuations?.interrupt(targetSessionId, authority)
}
/**
* Deliver selected content from one live continuable child to its durable
* direct parent. The child is the authority credential; callers cannot name a
@@ -315,6 +336,25 @@ export class SubagentService extends Service {
return listSubagentChildren(this.ctx, parentSessionId, signal)
}
/**
* Enumerate the root's complete session-backed subagent tree in stable
* pre-order from one live-preferred corpus, without loading or resuming an
* Agent. Ordinary sessions and one-shot children remain traversal nodes so
* continuable descendants below them are discovered; each returned entry
* adds its durable `parentId` and root-relative `depth`. Identity resolution,
* diagnostics, optional persistence, and cancellation follow the same
* projection-backed contract as {@link listChildren}.
* @param rootSessionId - session whose complete descendant tree is listed.
* @param signal - caller-owned cancellation forwarded to persistence reads
* and observed around every read await.
* @returns children and per-candidate diagnostics with tree position, in
* stable pre-order.
* @throws {@link SubagentError} under the same conditions as {@link listChildren}.
*/
listDescendants(rootSessionId: SessionId, signal?: AbortSignal): Promise<SubagentDescendantListEntry[]> {
return listSubagentDescendants(this.ctx, rootSessionId, signal)
}
/**
* Register a provider under its name. Registration is effect-scoped and HMR
* safe; removing a provider blocks new starts but does not revoke runs that
packages/subagent/subagent/src/list-children.ts
+140 -20
View File
@@ -1,18 +1,16 @@
/**
* Read-only enumeration of one parent's durable subagent children straight
* from the live session store and optional session persistence — no query
* seam. Candidates are the live-preferred merge of both listings filtered to
* durable `origin: 'subagent'` under the parent; each child's mode/label is
* the registered `subagent` projection unit's value, resolved down a
* three-rung ladder: the registry's watermark cache for a live child, a
* durable projection-cache row when it serves an own-suffix identity (the
* Read-only enumeration of durable subagent children and descendant trees
* straight from the live session store and optional session persistence — no
* query seam. Candidates come from one live-preferred corpus; each child's
* mode/label is the registered `subagent` projection unit's value, resolved
* down a three-rung ladder: the registry's watermark cache for a live child,
* a durable projection-cache row when it serves an own-suffix identity (the
* seq gate), and one persistence inspection folded through the registry
* otherwise, validated against the enumerated lifecycle. The projection
* fold is the single
* classification authority — this module parses no descriptor itself. Absent
* persistence, enumeration is live-only: a cold child is unreachable for
* resume anyway, so its absence is capability absence, not an error. The
* module owns no catalog state and does not consult Activation,
* otherwise, validated against the enumerated lifecycle. The projection fold
* is the single classification authority — this module parses no descriptor
* itself. Absent persistence, enumeration is live-only: a cold child is
* unreachable for resume anyway, so its absence is capability absence, not an
* error. The module owns no catalog state and does not consult Activation,
* Agent-registry, continuation-manager, or provider state.
*
* @module @deepseek-ai/dsh-subagent
@@ -88,6 +86,34 @@ export type SubagentListEntry =
readonly reason: 'corrupt' | 'unsupported' | 'unavailable'
}
/**
* One entry of a descendant listing: the interpreted subagent facts plus its
* position in the complete session tree. `parentId` is the durable direct
* parent from the enumerated header, and `depth` counts edges from the root.
*/
export type SubagentDescendantListEntry = SubagentListEntry & {
/** Durable direct parent of this candidate in the enumerated tree. */
readonly parentId: SessionId
/** Edge distance from the requested root; direct children are `1`. */
readonly depth: number
}
type CorpusRecord = { readonly header: SessionHeader; readonly live: Session | undefined }
interface ListingRuntime {
readonly projections: SessionProjectionRegistry
readonly persistence: SessionPersistence | undefined
readonly cache: SessionProjectionCache | undefined
readonly corpus: ReadonlyMap<SessionId, CorpusRecord>
readonly subagentParents: ReadonlySet<SessionId>
}
interface PositionedCandidate {
readonly record: CorpusRecord
readonly parentId: SessionId
readonly depth: number
}
/**
* Enumerate one parent's origin-classified direct children from the
* live-preferred merge of `ctx.sessions` and optional session persistence,
@@ -110,6 +136,55 @@ export async function listChildren(
parentSessionId: SessionId,
signal?: AbortSignal,
): Promise<SubagentListEntry[]> {
const listing = await prepareListing(ctx, signal)
const candidates = [...listing.corpus.values()]
.filter(record => record.header.parentSession === parentSessionId
&& record.header.origin === 'subagent')
.sort(compareCorpusRecords)
const rows = await resolveCandidateRows(candidates, listing, signal)
return rows.filter((row): row is SubagentListEntry => row !== undefined)
}
/**
* Enumerate every session-backed subagent below one root in stable pre-order.
* Ordinary sessions and one-shot children remain traversal nodes, so a
* continuable child below either is still discovered. Classification uses the
* same projection-backed runtime as {@link listChildren}; no Agent is loaded or
* resumed.
* @see SubagentService.listDescendants for the public cancellation and failure contract.
* @param ctx - context carrying the session store, projection registry, and optional persistence/cache.
* @param rootSessionId - session whose complete descendant tree is listed.
* @param signal - caller-owned cancellation observed around every persistence read.
* @returns interpreted subagents with durable direct-parent and root-relative depth.
* @throws {@link SubagentError} under the same conditions as {@link listChildren}.
*/
export async function listDescendants(
ctx: Context,
rootSessionId: SessionId,
signal?: AbortSignal,
): Promise<SubagentDescendantListEntry[]> {
const listing = await prepareListing(ctx, signal)
const positioned = descendantCandidates(listing.corpus, rootSessionId)
const rows = await resolveCandidateRows(
positioned.map(candidate => candidate.record),
listing,
signal,
)
const entries: SubagentDescendantListEntry[] = []
positioned.forEach((position, index) => {
const row = rows[index]
if (row !== undefined) {
entries.push({ ...row, parentId: position.parentId, depth: position.depth })
}
})
return entries
}
/** Resolve listing services once and build one live-preferred session corpus. */
async function prepareListing(
ctx: Context,
signal: AbortSignal | undefined,
): Promise<ListingRuntime> {
const projections = ctx.get('sessionProjections')
// Checked before any read, even with zero candidates: mode/label are the
// row's strong contract, so a missing fold capability is a deterministic
@@ -150,7 +225,7 @@ export async function listChildren(
}
// Live-preferred merge without header reconciliation: a live record wins
// its id wholesale, exactly as a live-preferred corpus would serve it.
const corpus = new Map<SessionId, { header: SessionHeader; live: Session | undefined }>()
const corpus = new Map<SessionId, CorpusRecord>()
for (const header of persistedHeaders) corpus.set(header.id, { header, live: undefined })
for (const session of sessions.list()) {
corpus.set(session.header.id, { header: session.header, live: session })
@@ -161,12 +236,16 @@ export async function listChildren(
subagentParents.add(record.header.parentSession)
}
}
const candidates = [...corpus.values()]
.filter(record => record.header.parentSession === parentSessionId
&& record.header.origin === 'subagent')
.sort((a, b) => a.header.createdAt - b.header.createdAt
|| a.header.id.localeCompare(b.header.id))
return { projections, persistence, cache, corpus, subagentParents }
}
/** Resolve projection-backed rows for aligned candidates with bounded cold reads. */
async function resolveCandidateRows(
candidates: readonly CorpusRecord[],
listing: ListingRuntime,
signal: AbortSignal | undefined,
): Promise<(SubagentListEntry | undefined)[]> {
const { projections, persistence, cache, subagentParents } = listing
const rows: (SubagentListEntry | undefined)[] = Array.from({ length: candidates.length })
const coldReads: { index: number; header: SessionHeader }[] = []
candidates.forEach((candidate, index) => {
@@ -212,7 +291,48 @@ export async function listChildren(
))
}
assertListingNotCancelled(signal)
return rows.filter((row): row is SubagentListEntry => row !== undefined)
return rows
}
/** Build origin-classified candidates from the complete tree without recursion. */
function descendantCandidates(
corpus: ReadonlyMap<SessionId, CorpusRecord>,
rootSessionId: SessionId,
): PositionedCandidate[] {
const children = new Map<SessionId, CorpusRecord[]>()
for (const record of corpus.values()) {
const parentId = record.header.parentSession
if (parentId === undefined) continue
const siblings = children.get(parentId)
if (siblings === undefined) children.set(parentId, [record])
else siblings.push(record)
}
for (const siblings of children.values()) siblings.sort(compareCorpusRecords)
const positioned: PositionedCandidate[] = []
const stack: PositionedCandidate[] = (children.get(rootSessionId) ?? [])
.map(record => ({ record, parentId: rootSessionId, depth: 1 }))
.reverse()
const visited = new Set<SessionId>([rootSessionId])
while (stack.length > 0) {
// The length guard proves one frame exists.
// oxlint-disable-next-line typescript/no-non-null-assertion
const position = stack.pop()!
const id = position.record.header.id
if (visited.has(id)) continue
visited.add(id)
if (position.record.header.origin === 'subagent') positioned.push(position)
const descendants = children.get(id) ?? []
for (const record of [...descendants].reverse()) {
stack.push({ record, parentId: id, depth: position.depth + 1 })
}
}
return positioned
}
/** Compare siblings by durable creation time, then id. */
function compareCorpusRecords(a: CorpusRecord, b: CorpusRecord): number {
return a.header.createdAt - b.header.createdAt || a.header.id.localeCompare(b.header.id)
}
/**
packages/subagent/subagent/tests/continuation.spec.ts
+220
View File
@@ -1717,3 +1717,223 @@ describe('continuable errors', () => {
expect(ctx.agents.get(started.childId)).toBeUndefined()
})
})
describe('SubagentService.interrupt', () => {
it('aborts the current turn durably, parks accepted follow-ups, and resumes them only on a waking send', async () => {
const releaseFirst = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('first'), gate: releaseFirst.promise },
{ chunks: textResponse('second') },
{ chunks: textResponse('third') },
{ chunks: textResponse('fourth') },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
await followup(ctx, parent, started.childId, message('parked B'))
await followup(ctx, parent, started.childId, message('parked C'))
const cancelSpy = vi.spyOn(child, 'cancel')
ctx.subagents.interrupt(started.childId, { kind: 'user', parentSessionId: parent.id })
expect(cancelSpy).toHaveBeenCalledTimes(1)
expect(cancelSpy).toHaveBeenCalledWith({ kind: 'user' }, { keepInbox: true })
// Cancellation is cooperative: the held model call observes it on release.
releaseFirst.resolve(undefined)
await child.whenIdle()
// Parked, not resumed: no second model request follows the abort, the
// accepted follow-ups stay pending, and the same Activation stays resident.
expect(adapter.requests).toHaveLength(1)
expect(child.inbox.nextTurn).toHaveLength(2)
expect(child.status).toBe('idle')
expect(ctx.agents.get(started.childId)).toBe(child)
// Only an explicit waking send restores the driver; the parked items then
// run before it in the existing FIFO order.
await followup(ctx, parent, started.childId, message('waking D'))
await waitNoActivation(ctx, started.childId)
const loaded = await ctx.sessionPersistence.load(started.childId)
expect(userTexts(loaded.events)).toEqual(['child task', 'parked B', 'parked C', 'waking D'])
const turnEnds = loaded.events
.filter(event => event.type === 'turn/end')
.map(event => (event).data.reason.kind)
expect(turnEnds).toEqual(['aborted', 'completed', 'completed', 'completed'])
})
it('interrupts only the target while its resident descendant keeps running', async () => {
const releaseChild = Promise.withResolvers<undefined>()
const releaseGrandchild = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('child'), gate: releaseChild.promise },
{ chunks: textResponse('grandchild'), gate: releaseGrandchild.promise },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const grandchild = await ctx.subagents.startContinuable(startSpec(child))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
const grandchildAgent = ctx.agents.get(grandchild.childId)!
const childCancel = vi.spyOn(child, 'cancel')
const grandchildCancel = vi.spyOn(grandchildAgent, 'cancel')
ctx.subagents.interrupt(started.childId, { kind: 'user', parentSessionId: parent.id })
expect(childCancel).toHaveBeenCalledTimes(1)
releaseChild.resolve(undefined)
await child.whenIdle()
// The target parks as a waiting owner; the published descendant was never
// signalled and keeps its own turn open.
expect(grandchildCancel).not.toHaveBeenCalled()
expect(ctx.agents.get(started.childId)).toBe(child)
expect(ctx.agents.get(grandchild.childId)).toBe(grandchildAgent)
releaseGrandchild.resolve(undefined)
await waitNoActivation(ctx, grandchild.childId)
await waitNoActivation(ctx, started.childId)
const loaded = await ctx.sessionPersistence.load(grandchild.childId)
const turnEnds = loaded.events
.filter(event => event.type === 'turn/end')
.map(event => (event).data.reason.kind)
expect(turnEnds).toEqual(['completed'])
})
it('authorizes the human address against the live target\'s durable direct parent', async () => {
const hold = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([{ chunks: textResponse('working'), gate: hold.promise }])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const cancelSpy = vi.spyOn(child, 'cancel')
expect(() => { ctx.subagents.interrupt(started.childId, {
kind: 'user',
parentSessionId: SessionId('stranger'),
}) }).toThrow(/belongs to another parent session/)
expect(cancelSpy).not.toHaveBeenCalled()
ctx.subagents.interrupt(started.childId, { kind: 'user', parentSessionId: parent.id })
expect(cancelSpy).toHaveBeenCalledWith({ kind: 'user' }, { keepInbox: true })
hold.resolve(undefined)
await waitNoActivation(ctx, started.childId)
})
it('lets a deep exact live ancestor interrupt its descendant with the parent cause', async () => {
const releaseChild = Promise.withResolvers<undefined>()
const releaseGrandchild = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('child'), gate: releaseChild.promise },
{ chunks: textResponse('grandchild'), gate: releaseGrandchild.promise },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const grandchild = await ctx.subagents.startContinuable(startSpec(child))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
const grandchildAgent = ctx.agents.get(grandchild.childId)!
const childCancel = vi.spyOn(child, 'cancel')
const grandchildCancel = vi.spyOn(grandchildAgent, 'cancel')
// Deep ancestor: the top-level parent interrupts the grandchild.
ctx.subagents.interrupt(grandchild.childId, { kind: 'ancestor', agent: parent })
expect(grandchildCancel).toHaveBeenCalledWith({ kind: 'parent' }, { keepInbox: true })
// Direct ancestor: the same authority kind covers the immediate parent.
ctx.subagents.interrupt(started.childId, { kind: 'ancestor', agent: parent })
expect(childCancel).toHaveBeenCalledWith({ kind: 'parent' }, { keepInbox: true })
releaseChild.resolve(undefined)
releaseGrandchild.resolve(undefined)
await waitNoActivation(ctx, grandchild.childId)
await waitNoActivation(ctx, started.childId)
})
it('rejects self, sibling, stale, and unrelated ancestor callers without touching the target', async () => {
const releaseA = Promise.withResolvers<undefined>()
const releaseB = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('a'), gate: releaseA.promise },
{ chunks: textResponse('b'), gate: releaseB.promise },
])
const { ctx, parent } = await setupWith(adapter)
const targetStart = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const target = ctx.agents.get(targetStart.childId)!
const siblingStart = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
const sibling = ctx.agents.get(siblingStart.childId)!
const stranger = ctx.agentLoop.create(SessionId('stranger'), { provider: 'mock', model: 'mock' })
const stale = { ...parent, id: parent.id } as unknown as Agent
const cancelSpy = vi.spyOn(target, 'cancel')
expect(() => { ctx.subagents.interrupt(targetStart.childId, { kind: 'ancestor', agent: target }) })
.toThrow(/cannot interrupt itself/)
expect(() => { ctx.subagents.interrupt(targetStart.childId, { kind: 'ancestor', agent: sibling }) })
.toThrow(/not a live descendant/)
expect(() => { ctx.subagents.interrupt(targetStart.childId, { kind: 'ancestor', agent: stranger }) })
.toThrow(/not a live descendant/)
expect(() => { ctx.subagents.interrupt(targetStart.childId, { kind: 'ancestor', agent: stale }) })
.toThrow(/exact live ancestor/)
// A stale caller is rejected before target lookup, even for an absent id.
expect(() => { ctx.subagents.interrupt(SessionId('missing'), { kind: 'ancestor', agent: stale }) })
.toThrow(/exact live ancestor/)
expect(cancelSpy).not.toHaveBeenCalled()
releaseA.resolve(undefined)
releaseB.resolve(undefined)
await waitNoActivation(ctx, targetStart.childId)
await waitNoActivation(ctx, siblingStart.childId)
})
it('accepts absent and one-shot ids as no-ops without touching the one-shot Agent', async () => {
const { ctx, parent } = await setup([textResponse('one shot')])
ctx.subagents.interrupt(SessionId('missing'), { kind: 'user', parentSessionId: parent.id })
ctx.subagents.interrupt(SessionId('missing'), { kind: 'ancestor', agent: parent })
const run = await ctx.subagents.start('spawn', {
label: 'one-shot work',
prompt: message('one-shot work'),
parent,
signal: testSignal,
})
const oneShot = run.localAgent!
const cancelSpy = vi.spyOn(oneShot, 'cancel')
ctx.subagents.interrupt(run.id, { kind: 'user', parentSessionId: parent.id })
ctx.subagents.interrupt(run.id, { kind: 'ancestor', agent: parent })
expect(cancelSpy).not.toHaveBeenCalled()
await run.result
await run.dispose()
})
it('accepts an interrupt after natural completion', async () => {
const { ctx, parent } = await setup([textResponse('done')])
const started = await ctx.subagents.startContinuable(startSpec(parent))
await waitNoActivation(ctx, started.childId)
ctx.subagents.interrupt(started.childId, { kind: 'user', parentSessionId: parent.id })
ctx.subagents.interrupt(started.childId, { kind: 'ancestor', agent: parent })
})
it('accepts an interrupt that lost the race with disposal without signalling twice', async () => {
const hold = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([{ chunks: textResponse('working'), gate: hold.promise }])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable(startSpec(parent))
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const cancelSpy = vi.spyOn(child, 'cancel')
// Scoped teardown opens the disposal transaction synchronously and issues
// its own whole-Activation cancel before this call returns.
const drained = ctx.subagents.drainContinuableDescendants([parent])
expect(cancelSpy).toHaveBeenCalledTimes(1)
// Interrupt after the cutoff: accepted no-op, no second signal, no waiting.
ctx.subagents.interrupt(started.childId, { kind: 'user', parentSessionId: parent.id })
expect(cancelSpy).toHaveBeenCalledTimes(1)
hold.resolve(undefined)
await drained
})
})
packages/subagent/subagent/tests/list-children.spec.ts
+240
View File
@@ -972,3 +972,243 @@ describe('SubagentService.listChildren', () => {
expect((caught as SubagentError).code).toBe('SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE')
})
})
describe('SubagentService.listDescendants', () => {
it('flattens the complete tree in stable pre-order with verified parent and depth', async () => {
const { ctx, parent } = await setup([])
const childA = await authorChild(ctx, '00000000-0000-4000-8000-00000000aaa1', {
parentSession: parent.id,
createdAt: 1,
origin: 'subagent',
}, childEvents(descriptorPayload('branch a')))
const grandchild = await authorChild(ctx, '00000000-0000-4000-8000-00000000aaa2', {
parentSession: childA,
createdAt: 2,
origin: 'subagent',
}, childEvents(descriptorPayload('under a')))
const childB = await authorChild(ctx, '00000000-0000-4000-8000-00000000aaa3', {
parentSession: parent.id,
createdAt: 3,
origin: 'subagent',
}, childEvents(descriptorPayload('branch b')))
const entries = await ctx.subagents.listDescendants(parent.id)
expect(entries).toEqual([
{
kind: 'child', id: childA, label: 'branch a', mode: 'continuable',
activity: 'inactive', hasChildren: true, parentId: parent.id, depth: 1,
},
{
kind: 'child', id: grandchild, label: 'under a', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: childA, depth: 2,
},
{
kind: 'child', id: childB, label: 'branch b', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: parent.id, depth: 1,
},
])
})
it('returns an empty result when the root has no descendants', async () => {
const { ctx, parent } = await setup([])
await ctx.sessions.flush(parent.session)
await expect(ctx.subagents.listDescendants(parent.id)).resolves.toEqual([])
})
it('omits a live creation-window candidate while continuing through its subtree', async () => {
const { ctx, parent } = await setup([])
const bareId = SessionId('live-creation-window')
const bare = ctx.sessions.create(bareId, {
meta: { createdAt: 1, parentSession: parent.id, origin: 'subagent' },
})
bare.append('turn/start', { turn: 1 })
const below = await authorChild(ctx, '00000000-0000-4000-8000-00000000aaaf', {
parentSession: bareId,
createdAt: 2,
origin: 'subagent',
}, childEvents(descriptorPayload('below the creation window')))
await expect(ctx.subagents.listDescendants(parent.id)).resolves.toEqual([{
kind: 'child', id: below, label: 'below the creation window', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: bareId, depth: 2,
}])
})
it('contains a corrupt parent cycle without revisiting the requested root', async () => {
const { ctx } = await setup([])
const rootId = SessionId('cycle-root')
const nodeId = SessionId('cycle-node')
await authorChild(ctx, rootId, {
parentSession: nodeId,
createdAt: 2,
}, childEvents(descriptorPayload('ordinary cycle root')))
await authorChild(ctx, nodeId, {
parentSession: rootId,
createdAt: 1,
origin: 'subagent',
}, childEvents(descriptorPayload('cycle child')))
await expect(ctx.subagents.listDescendants(rootId)).resolves.toEqual([{
kind: 'child', id: nodeId, label: 'cycle child', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: rootId, depth: 1,
}])
})
it('walks a deeply nested ordinary-session chain without consuming the call stack', async () => {
const { ctx, parent } = await setup([])
const depth = 10_000
let parentId = parent.id
for (let level = 1; level < depth; level += 1) {
const session = ctx.sessions.create(SessionId(`deep-ordinary-${level}`), {
meta: { createdAt: level, parentSession: parentId },
})
parentId = session.id
}
const leafId = SessionId('deep-subagent-leaf')
const leaf = ctx.sessions.create(leafId, {
meta: { createdAt: depth, parentSession: parentId, origin: 'subagent' },
})
leaf.append('turn/start', { turn: 1 })
leaf.append('subagent/descriptor', descriptorPayload('deep leaf'))
await expect(ctx.subagents.listDescendants(parent.id)).resolves.toEqual([{
kind: 'child', id: leafId, label: 'deep leaf', mode: 'continuable',
activity: 'running', hasChildren: false, parentId, depth,
}])
})
it('discovers continuable descendants below ordinary and one-shot intermediates', async () => {
const { ctx, parent } = await setup([textResponse('one shot')])
// An ordinary fork has no descriptor: omitted itself, subtree still walked.
const fork = ctx.sessions.fork(parent.session, undefined, SessionId('plain-fork'))
await ctx.sessions.flush(fork)
const underFork = await authorChild(ctx, '00000000-0000-4000-8000-00000000bbb1', {
parentSession: fork.header.id,
createdAt: 2,
origin: 'subagent',
}, childEvents(descriptorPayload('under the fork')))
// A real one-shot child, then a continuable authored below it.
const oneShot = await ctx.subagents.start('spawn', {
label: 'one-shot intermediate',
prompt: [{ type: 'text', text: 'one-shot task' }],
parent,
signal: testSignal,
})
await oneShot.result
await ctx.sessions.flush(oneShot.localAgent!.session)
const oneShotId = oneShot.id
await oneShot.dispose()
const underOneShot = await authorChild(ctx, '00000000-0000-4000-8000-00000000bbb2', {
parentSession: oneShotId,
createdAt: 9_999_999_999_999,
origin: 'subagent',
}, childEvents(descriptorPayload('under the one-shot')))
const entries = await ctx.subagents.listDescendants(parent.id)
// The fork is absent (descriptor-less); the one-shot is present with its
// mode so a caller can see the lineage it walked through.
expect(entries.map(entry => entry.id)).not.toContain(fork.header.id)
expect(entries).toContainEqual({
kind: 'child', id: underFork, label: 'under the fork', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: fork.header.id, depth: 2,
})
expect(entries).toContainEqual(expect.objectContaining({
kind: 'child', id: oneShotId, mode: 'one-shot', parentId: parent.id, depth: 1,
}))
expect(entries).toContainEqual({
kind: 'child', id: underOneShot, label: 'under the one-shot', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: oneShotId, depth: 2,
})
// Pre-order: every child appears after its own parent entry.
const position = new Map(entries.map((entry, index) => [entry.id, index]))
expect(position.get(underOneShot)!).toBeGreaterThan(position.get(oneShotId)!)
})
it('diagnoses a settled descriptor-less node while walking its subtree', async () => {
const { ctx, parent } = await setup([])
// A settled origin-marked candidate without an identity is corrupt under
// the projection contract, but its subtree remains independently visible.
const bare = await authorChild(ctx, '00000000-0000-4000-8000-00000000eee1', {
parentSession: parent.id,
createdAt: 1,
origin: 'subagent',
}, [
{ type: 'turn/start', seq: 0, time: 1, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
{ type: 'turn/end', seq: 1, time: 2, data: { turn: 1, reason: { kind: 'completed' } } },
] as SessionEvent[])
const below = await authorChild(ctx, '00000000-0000-4000-8000-00000000eee2', {
parentSession: bare,
createdAt: 2,
origin: 'subagent',
}, childEvents(descriptorPayload('below the bare node')))
await expect(ctx.subagents.listDescendants(parent.id)).resolves.toEqual([
{ kind: 'diagnostic', id: bare, reason: 'corrupt', parentId: parent.id, depth: 1 },
{
kind: 'child', id: below, label: 'below the bare node', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: bare, depth: 2,
},
])
})
it('keeps traversing below a corrupt intermediate and positions its diagnostic', async () => {
const { ctx, parent } = await setup([])
const corrupt = await authorChild(ctx, '00000000-0000-4000-8000-00000000ccc1', {
parentSession: parent.id,
createdAt: 1,
origin: 'subagent',
}, childEvents(descriptorPayload('unsupported descriptor', 999)))
const below = await authorChild(ctx, '00000000-0000-4000-8000-00000000ccc2', {
parentSession: corrupt,
createdAt: 2,
origin: 'subagent',
}, childEvents(descriptorPayload('below the corrupt node')))
const entries = await ctx.subagents.listDescendants(parent.id)
expect(entries).toEqual([
{ kind: 'diagnostic', id: corrupt, reason: 'corrupt', parentId: parent.id, depth: 1 },
{
kind: 'child', id: below, label: 'below the corrupt node', mode: 'continuable',
activity: 'inactive', hasChildren: false, parentId: corrupt, depth: 2,
},
])
})
it('verifies a cold candidate still belongs to its enumerated lifecycle', async () => {
const { ctx, parent } = await setup([])
const childId = await authorChild(ctx, '00000000-0000-4000-8000-00000000ddd1', {
parentSession: parent.id,
createdAt: 1,
origin: 'subagent',
}, childEvents(descriptorPayload('lineage checked')))
const realInspect = ctx.sessionPersistence.inspect.bind(ctx.sessionPersistence)
ctx.sessionPersistence.inspect = async (sessionId, signal) => {
const inspected = await realInspect(sessionId, signal)
// The exact read reports a different durable parent than enumeration did.
return { ...inspected, meta: { ...inspected.meta, parentSession: SessionId('someone-else') } }
}
await expect(ctx.subagents.listDescendants(parent.id)).resolves.toEqual([
{ kind: 'diagnostic', id: childId, reason: 'corrupt', parentId: parent.id, depth: 1 },
])
})
it('a pre-aborted signal stops the descendant scan before persistence reads', async () => {
const { ctx, parent } = await setup([textResponse('done')])
await startChild(ctx, parent, 'never read')
const list = vi.spyOn(ctx.sessionPersistence, 'list')
const controller = new AbortController()
controller.abort()
await expect(ctx.subagents.listDescendants(parent.id, controller.signal)).rejects.toThrow(
expect.objectContaining({ code: 'CANCELLED' }) as Error,
)
expect(list).not.toHaveBeenCalled()
})
it('fails loud when the projection registry is not mounted', async () => {
const { ctx, parent } = await setup([], { sessionProjections: false })
await expect(ctx.subagents.listDescendants(parent.id)).rejects.toThrow(
expect.objectContaining({ code: 'SUBAGENT_CONTROL_PROJECTIONS_UNAVAILABLE' }) as Error,
)
})
})
packages/subagent/subagent/tests/service.spec.ts
+10
View File
@@ -133,6 +133,16 @@ describe('SubagentService', () => {
await expect(subagents.drainContinuableDescendants([])).resolves.toBeUndefined()
})
it('treats interrupt as an accepted no-op when no manager was bound', async () => {
const { subagents } = await service()
// Without a continuation manager no live Activation can exist, so there is
// nothing to stop and nothing to authorize against.
expect(() => { subagents.interrupt(SessionId('child'), {
kind: 'user',
parentSessionId: SessionId('parent-1'),
}) }).not.toThrow()
})
it('rejects continuable operations when their runtime services are absent', async () => {
const { subagents } = await service()
await expect(subagents.startContinuable({
packages/subagent/tool-subagent-control/README.i18n.yaml
+2 -2
View File
@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/subagent/tool-subagent-control/README.md
README.md: ea95a45b85e01d1f5f1c478a35c80c65151724ac
README.zh.md: 2cc876c8b39caa19fdf30eae7c8def0ba81fe7b1
README.md: 4d9991b720bbbff8862e693b6aaed332f414af0e
README.zh.md: dc1c81fc3313a8dfa277f33ae745702ed5b9d34b
packages/subagent/tool-subagent-control/README.md
+22 -6
View File
@@ -2,11 +2,13 @@
English | [中文](README.zh.md)
The optional, globally named `send_message` and `list_agents` tools are thin adapters over `ctx.subagents`. Provider-bound `@deepseek-ai/dsh-tool-subagent` instances register distinct delegation tools per transport; this separately loaded package registers shared control tools once, so multiple delegation tools never register duplicate global controls. The root plugin registers `send_message` and the separately loadable `./list-agents` plugin registers `list_agents`; both require only `subagents`, so a deployment can keep `send_message` while omitting the list tool. Neither tool's presence determines whether a delegation tool starts continuable work. These tools own only the parent-to-child direction; the independently installed [`@deepseek-ai/dsh-tool-subagent-report`](../tool-subagent-report/README.md) owns the child-to-parent direction.
The optional, globally named `send_message`, `interrupt_agent`, and `list_agents` tools are thin adapters over `ctx.subagents`. Provider-bound `@deepseek-ai/dsh-tool-subagent` instances register distinct delegation tools per transport; this separately loaded package registers shared control tools once, so multiple delegation tools never register duplicate global controls. The root plugin registers `send_message` and `interrupt_agent` and requires only `subagents`; the separately loadable `./list-agents` plugin registers `list_agents` and declares `subagents` plus `agents` as load-time dependencies. Its catalog reads additionally require the session store and projection registry at call time, but no query service. A deployment can keep the root tools while omitting the list tool. No tool's presence determines whether a delegation tool starts continuable work. These tools own only the parent-to-child direction; the independently installed [`@deepseek-ai/dsh-tool-subagent-report`](../tool-subagent-report/README.md) owns the child-to-parent direction.
The tool performs no lifecycle routing — residency and cold resume belong to the subagent service. It passes `exec.agent` as the exact live parent that authorizes delivery and attributes every message as durable provenance `{ kind: 'coordinator', senderSessionId: parent.id }`, which the service retains but never treats as authority. Every message becomes the subagent's next FIFO turn through `Agent.followup()`: if the child is still working, the message waits until its current turn finishes, so it cannot redirect work already underway. The tool forwards its execution signal, which owns admission only until inbox acceptance; once the child accepts the message the accepted turn cannot be cancelled through this tool. This call returns no child reply — its transcript by that id is the source of what it did — and a child with `report` sends content on its own initiative as a separate parent message. A delivery failure becomes an errored tool result stating the message was not delivered.
`list_agents` takes no arguments, derives the parent id from the calling agent, and projects `ctx.subagents.listChildren()` to continuable children without a cursor. The service result also contains one-shot session-backed subagents for consumers such as a UI, but those entries are omitted from this model tool because they cannot accept `send_message`. Diagnostics remain visible. Durable identity and mode come from each child's descriptor, while delivery-time authority and Activation ownership checks remain `send_message`'s.
`interrupt_agent(agent_id)` passes `exec.agent` as the exact live ancestor authority for `ctx.subagents.interrupt()`: the target may be a direct child or a deeper descendant, and the service — never this tool — verifies the caller against the target Activation's recorded lineage. Only the target's current turn stops (`keepInbox`): queued messages stay parked until a later `send_message`, published descendants keep running, and the child stays available for follow-ups. The call returns as soon as the stop request is accepted, without waiting for target quiescence; an absent or already-settled target is an accepted no-op, while self, sibling, stale, and non-ancestor callers become errored results.
`list_agents` takes one optional `scope` argument, derives the root id from the calling agent, and projects the service catalog to continuable children without a cursor. The default `children` scope reads `ctx.subagents.listChildren()`; `descendants` reads `ctx.subagents.listDescendants()`, whose one-corpus walk crosses ordinary sessions and one-shot children and renders surviving rows in stable pre-order with `parent=<id> depth=<n>`. The `parent` annotation is the durable direct-parent session id and may name an ordinary session omitted from the output. For the calling agent, only depth-1 child entries are `send_message` candidates; deeper child entries are `interrupt_agent` candidates only. Status comes from the live Agent registry: `running` (active driver), `idle` (resident between turns, possibly waiting on agents it started), `complete` (storage only). The service result also contains one-shot session-backed subagents for consumers such as a UI, but those entries are omitted from this model tool because they cannot accept `send_message`. Diagnostics remain visible, with positions in the descendants scope. Durable identity and mode come from each child's descriptor, while delivery-time authority and Activation ownership checks remain the service's.
## Model Experience
@@ -14,7 +16,7 @@ The tool performs no lifecycle routing — residency and cold resume belong to t
#### What the model sees
The generated [`send_message` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent-control): `subagent_id` and `message`, describing that the message becomes the subagent's next turn, that this call returns no answer from the subagent, and that a failure means the message was not delivered.
The generated [schemas](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent-control): `send_message` takes `subagent_id` and `message`, describing that the message becomes the subagent's next turn, that this call returns no answer from the subagent, and that a failure means the message was not delivered; `interrupt_agent` takes `agent_id`, describing that only the current turn stops, queued messages park, descendants keep running, and acceptance precedes the actual stop; `list_agents` takes the optional `scope` enum.
#### Token effect
@@ -24,6 +26,20 @@ Fixed schema cost per parent request.
Prefix-stable; the schema does not change at runtime.
### Interrupt result
#### What the model sees
`interrupt requested for agent <agent_id>` on acceptance. An unauthorized caller — self, sibling, stale, or non-ancestor — is an errored result naming the rejection; an absent or settled target still renders the acceptance line.
#### Token effect
One short acknowledgement per call; the interrupted turn's abort is visible only in the child's own transcript.
#### KV Cache effect
Append-only; each result follows the reusable request prefix.
### Delivery result
#### What the model sees
@@ -42,11 +58,11 @@ Append-only; newly visible content follows the reusable request prefix and does
#### What the model sees
One line per continuable child in the trace's stable order: `<id> [<status>] — <label>` (`running` = the logical session is live, `complete` = persisted only and resumable by `send_message`), plus `<id> [diagnostic: <reason>]` for a candidate that could not be read (`corrupt`, `unsupported`, or `unavailable`). One-shot children are intentionally absent; `(no subagents)` means no continuable child or diagnostic survived the projection. Diagnostics never expose descriptor contents.
One line per continuable child in stable catalog order: `<id> [<status>] — <label>` (`running` = active driver, `idle` = resident between turns, `complete` = storage only; a direct child in that state can be resumed by `send_message`), plus `<id> [diagnostic: <reason>]` for a candidate that could not be read (`corrupt`, `unsupported`, or `unavailable`). The `descendants` scope inserts ` parent=<id> depth=<n>` before the label dash on every line, in pre-order. One-shot children are intentionally absent; `(no subagents)` means no continuable child or diagnostic survived the projection. Diagnostics never expose descriptor contents.
#### Token effect
Grows linearly with the parent's direct continuable children; there is no cursor or cap, so long-lived parents with many persisted children pay the full list each call.
Grows linearly with the listed continuable children — the whole tree under the `descendants` scope; there is no cursor or cap, so long-lived parents with many persisted children pay the full list each call.
#### KV Cache effect
@@ -56,5 +72,5 @@ Append-only; each result follows the reusable request prefix.
- **A queued message has no independent result** — acceptance returns only its inbox `messageId`; the child's work lands in the durable child Session and is never collected through this tool. A child granted `report` may send selected content back separately, but that message is not this call's result.
- **No steering of the current turn** — every message opens a later FIFO turn, so a message sent while the child is working runs only after its current turn finishes and cannot redirect it.
- **Listing is a snapshot, not a delivery promise** — it may race publication, disposal, or a later message, and another process may activate a child this process reports as `complete`; cross-process accuracy requires a shared lease.
- **Listing is a snapshot, not a delivery promise** — it may race publication, disposal, or a later message, and another process may activate a child this process reports as `complete`; cross-process accuracy requires a shared lease. `interrupt_agent` performs the authoritative live-lineage check itself, so discovery staleness cannot grant authority.
- **No pagination or deletion** — the complete stably ordered set is returned, and persisted children remain listed for as long as their sessions remain in persistence; a service-level bound or delete operation is a later product decision.
packages/subagent/tool-subagent-control/README.zh.md
+22 -6
View File
@@ -2,11 +2,13 @@
[English](README.md) | 中文
可选的全局具名 `send_message` 与 `list_agents` 工具是 `ctx.subagents` 之上的轻量适配器。绑定提供方的 `@deepseek-ai/dsh-tool-subagent` 实例会为每种传输注册不同的委派工具；这个单独加载的包只注册一次共享控制工具，因此多个委派工具绝不会重复注册全局控制工具。根插件注册 `send_message`，可单独加载的 `./list-agents` 插件注册 `list_agents`；两者都只要求 `subagents`，部署可保留 `send_message` 而省略列表工具。是否加载这些工具不会决定委派工具是否启动可继续工作。这些工具只负责父到子的方向；单独安装的 [`@deepseek-ai/dsh-tool-subagent-report`](../tool-subagent-report/README.md) 负责子到父的方向。
可选的全局具名 `send_message`、`interrupt_agent` 与 `list_agents` 工具是 `ctx.subagents` 之上的轻量适配器。绑定提供方的 `@deepseek-ai/dsh-tool-subagent` 实例会为每种传输注册不同的委派工具；这个单独加载的包只注册一次共享控制工具，因此多个委派工具绝不会重复注册全局控制工具。根插件注册 `send_message` 与 `interrupt_agent`，且只要求 `subagents`；可单独加载的 `./list-agents` 插件注册 `list_agents`，并将 `subagents` 与 `agents` 声明为加载时依赖。其目录读取在调用时还要求会话存储与投影注册表，但不要求任何查询服务。部署可保留根插件工具并省略列表工具。是否加载这些工具不会决定委派工具是否启动可继续工作。这些工具只负责父到子的方向；单独安装的 [`@deepseek-ai/dsh-tool-subagent-report`](../tool-subagent-report/README.md) 负责子到父的方向。
本工具不执行生命周期路由：驻留与冷恢复归 subagent 服务所有。它将 `exec.agent` 作为授权投递的确切在线父级传入，并把每条消息的来源标记为持久化来源 `{ kind: 'coordinator', senderSessionId: parent.id }`；服务会保留该来源，但绝不将其视为权限。每条消息都会通过 `Agent.followup()` 成为子 agent（智能体）的下一个 FIFO 轮次：如果子 agent 仍在工作，该消息会等待其当前轮次结束，因此无法重定向已经在进行的工作。本工具会转发其执行信号，该信号只在 inbox 接受之前掌管准入；一旦子 agent 接受消息，已接受的轮次便无法再通过本工具取消。本次调用不会返回子 agent 的回复；通过该 id 查看其 transcript（文本记录），才是了解它完成了哪些工作的真源。拥有 `report` 的子 agent 会自行把内容作为一条单独的父级消息发回。投递失败会变为出错的工具结果，并明确说明消息未送达。
`list_agents` 不接受参数，会从调用它的 agent 推导 parent id，并且不使用 cursor，将 `ctx.subagents.listChildren()` 的结果投影为可继续 child。服务结果还包含由会话支撑的一次性 subagent，以供 UI 等消费方使用；但这些条目无法接受 `send_message`，因此会从这个模型工具中排除。diagnostic 仍然可见。持久化身份和模式来自每个子 agent 的描述符，消息送达时的鉴权和 Activation 所有权检查仍归 `send_message` 负责。
`interrupt_agent(agent_id)` 将 `exec.agent` 作为 `ctx.subagents.interrupt()` 的确切在线 ancestor 授权传入：目标可以是直接 child 或更深的后代，由服务——而不是本工具——依据目标 Activation 记录的 lineage 校验调用方。只有目标的当前轮次会停止（`keepInbox`）：已排队的消息保持暂停直到之后的 `send_message`，已发布的后代继续运行，child 也仍可接受后续消息。调用在停止请求被接受后立即返回，不等待目标静止；目标不存在或已结算是被接受的 no-op，而 self、sibling、过期与非 ancestor 调用方会成为出错结果。
`list_agents` 接受一个可选的 `scope` 参数，会从调用它的 agent 推导根 id，并且不使用 cursor，将服务目录投影为可继续 child。默认的 `children` scope 读取 `ctx.subagents.listChildren()`；`descendants` 读取 `ctx.subagents.listDescendants()`，其单份语料的遍历会穿过普通会话与一次性 child，并按稳定 pre-order 以 `parent=<id> depth=<n>` 渲染保留下来的条目。`parent` 注释是持久化直接 parent 会话 id，可能指向输出中省略的普通会话。对于调用本工具的 agent，只有 depth-1 child 条目可作为 `send_message` 候选；更深的 child 条目只能作为 `interrupt_agent` 候选。状态来自在线 Agent 注册表：`running`（driver 活跃）、`idle`（驻留但处于轮次之间，可能在等待它启动的 agent）、`complete`（仅存于存储）。服务结果还包含由会话支撑的一次性 subagent，以供 UI 等消费方使用；但这些条目无法接受 `send_message`，因此会从这个模型工具中排除。diagnostic 仍然可见，并在 descendants scope 中带有位置。持久化身份和模式来自每个子 agent 的描述符，消息送达时的鉴权和 Activation 所有权检查仍归服务负责。
## 模型体验
@@ -14,7 +16,7 @@
#### 模型看到的内容
已生成的 [`send_message` schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent-control)：包含 `subagent_id` 和 `message`，说明消息会成为子 agent 的下一个轮次、本次调用不会返回子 agent 的回答，以及失败即表示消息未送达。
已生成的 [schema](../../../docs/tool-catalog.md#deepseek-aidsh-tool-subagent-control)：`send_message` 包含 `subagent_id` 和 `message`，说明消息会成为子 agent 的下一个轮次、本次调用不会返回子 agent 的回答，以及失败即表示消息未送达；`interrupt_agent` 包含 `agent_id`，说明只有当前轮次会停止、已排队消息保持暂停、后代继续运行，以及接受先于实际停止；`list_agents` 包含可选的 `scope` 枚举。
#### Token 影响
@@ -24,6 +26,20 @@
前缀保持稳定；schema 不会在运行时改变。
### 中断结果
#### 模型看到的内容
接受时返回 `interrupt requested for agent <agent_id>`。未授权的调用方——self、sibling、过期或非 ancestor——会成为指明拒绝原因的出错结果；目标不存在或已结算仍渲染接受行。
#### Token 影响
每次调用产生一条简短确认消息；被中断轮次的中止只在 child 自己的 transcript 中可见。
#### KV Cache 影响
仅追加；每个结果都位于可复用请求前缀之后。
### 投递结果
#### 模型看到的内容
@@ -42,11 +58,11 @@
#### 模型看到的内容
按追踪结果的稳定顺序，每个可继续 child 占一行：渲染为 `<id> [<status>] — <label>`（`running` 表示逻辑会话存活，`complete` 表示仅存在于持久化存储中，可通过 `send_message` 恢复），另为无法读取的候选项渲染 `<id> [diagnostic: <reason>]`（`corrupt`、`unsupported` 或 `unavailable`）。一次性 child 会被有意排除；`(no subagents)` 表示投影后没有留下可继续 child 或 diagnostic。诊断信息绝不会暴露描述符内容。
按稳定目录顺序，每个可继续 child 占一行：渲染为 `<id> [<status>] — <label>`（`running` 表示 driver 活跃，`idle` 表示驻留但处于轮次之间，`complete` 表示仅存于存储；处于该状态的直接 child 可通过 `send_message` 恢复），另为无法读取的候选项渲染 `<id> [diagnostic: <reason>]`（`corrupt`、`unsupported` 或 `unavailable`）。`descendants` scope 会在每行 label 破折号之前插入 ` parent=<id> depth=<n>`，按 pre-order 排列。一次性 child 会被有意排除；`(no subagents)` 表示投影后没有留下可继续 child 或 diagnostic。诊断信息绝不会暴露描述符内容。
#### Token 影响
随 parent 的直接可继续 child 数量线性增长；没有 cursor 或上限，因此长期存活且有许多持久化 child 的 parent 每次调用都会承担完整列表成本。
随所列可继续 child 数量线性增长——`descendants` scope 下为整棵树；没有 cursor 或上限，因此长期存活且有许多持久化 child 的 parent 每次调用都会承担完整列表成本。
#### KV Cache 影响
@@ -56,5 +72,5 @@
- **已排队的消息没有独立结果**：接受时只返回其 inbox `messageId`；子 agent 的工作会落入持久化子 agent 会话，绝不会通过本工具收集。获得 `report` 的子 agent 可以单独发回选定内容，但该消息不是本次调用的结果。
- **不对当前轮次进行 steering（中途引导）**：每条消息都会开启后续 FIFO 轮次，因此在子 agent 工作时发送的消息只会在其当前轮次结束后运行，无法将其重定向。
- **列表是快照，而非投递承诺**：它可能与发布、dispose（资源释放）或后续消息发生竞态，另一个进程也可能激活当前进程报告为 `complete` 的 child；跨进程准确性需要共享租约。
- **列表是快照，而非投递承诺**：它可能与发布、dispose（资源释放）或后续消息发生竞态，另一个进程也可能激活当前进程报告为 `complete` 的 child；跨进程准确性需要共享租约。`interrupt_agent` 自己执行权威的在线 lineage 检查，因此过期的发现结果不会授予权限。
- **没有分页或删除**：系统返回完整且稳定排序的集合；只要 child 会话仍在持久化存储中，它就会继续出现在列表中，服务级上限或删除操作留待后续产品决策。
packages/subagent/tool-subagent-control/package.json
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@deepseek-ai/dsh-tool-subagent-control",
"description": "Globally named send_message and list_agents tools over ctx.subagents continuations",
"description": "Globally named send_message, interrupt_agent, and list_agents tools over ctx.subagents continuations",
"version": "0.0.1",
"private": true,
"type": "module",
packages/subagent/tool-subagent-control/src/index.ts
+50 -6
View File
@@ -1,9 +1,11 @@
/**
* The globally named `send_message` tool: a thin model-facing adapter over
* `ctx.subagents.followup()`. It performs no lifecycle routing of its own —
* residency and cold resume belong to the subagent service — and it lives apart
* from the provider-bound `@deepseek-ai/dsh-tool-subagent` instances so multiple
* delegation tools share one control tool.
* The globally named `send_message` and `interrupt_agent` tools: thin
* model-facing adapters over `ctx.subagents.followup()` and
* `ctx.subagents.interrupt()`. They perform no lifecycle routing of their own —
* residency, cold resume, and interrupt authorization belong to the subagent
* service — and they live apart from the provider-bound
* `@deepseek-ai/dsh-tool-subagent` instances so multiple delegation tools share
* one control surface.
* @module @deepseek-ai/dsh-tool-subagent-control
*/
@@ -17,7 +19,7 @@ export const name = 'tool-subagent-control'
export const inject = ['tools', 'subagents']
/**
* Register the `send_message` tool.
* Register the `send_message` and `interrupt_agent` tools.
* @param ctx - context carrying the tool registry and subagent service.
*/
export function apply(ctx: Context): void {
@@ -73,4 +75,46 @@ export function apply(ctx: Context): void {
return { messageId }
},
}))
ctx.tools.register(defineTool({
name: 'interrupt_agent',
description:
'Request cancellation of a background agent\'s current turn by its agent id. The target may be your '
+ 'direct child or a deeper agent created under you. Only the current turn stops: messages already '
+ 'queued for the agent stay parked until a later send_message, agents it started keep running, and '
+ 'the agent itself stays available for follow-ups. This call returns as soon as the stop request is '
+ 'accepted, so the target may keep running briefly; interrupting an agent that already finished is '
+ 'an accepted no-op.',
parameters: {
agent_id: {
type: 'string',
required: true,
description: 'The agent id of the running agent to interrupt.',
},
},
output: {
schema: {
type: 'object',
additionalProperties: false,
properties: {
accepted: { type: 'boolean', required: true },
},
},
render: (args, _value) => [{
type: 'text',
text: `interrupt requested for agent ${args.agent_id}`,
}],
},
execute(args, exec) {
const caller = exec.agent
if (!caller) {
// Ancestor authority requires an exact live calling agent.
throw new Error('interrupt_agent requires a calling agent (exec.agent was undefined)')
}
// The service authorizes the exact live caller against the target's
// recorded lineage; the tool adds no authority of its own.
ctx.subagents.interrupt(SessionId(args.agent_id), { kind: 'ancestor', agent: caller })
return Promise.resolve({ accepted: true })
},
}))
}
packages/subagent/tool-subagent-control/src/list-agents.ts
+90 -34
View File
@@ -1,46 +1,95 @@
/**
* The globally named `list_agents` tool: a thin model-facing adapter over
* the continuable projection of `ctx.subagents.listChildren()`. It stays
* separately loadable from the root `send_message` plugin so a deployment
* can register `send_message` without exposing the list tool.
* the continuable projection of `ctx.subagents.listChildren()` and, for the
* `descendants` scope, `ctx.subagents.listDescendants()`. It stays separately
* loadable from the root `send_message` plugin so a deployment can register
* continuation delivery without exposing discovery.
* @module @deepseek-ai/dsh-tool-subagent-control/list-agents
*/
import type { Context } from 'cordis'
import { defineTool } from '@deepseek-ai/dsh-tools'
import type {} from '@deepseek-ai/dsh-subagent'
import type { Agent } from '@deepseek-ai/dsh-agent'
import type { SessionId } from '@deepseek-ai/dsh-session'
import type { SubagentDescendantListEntry, SubagentListEntry } from '@deepseek-ai/dsh-subagent'
export const name = 'tool-subagent-list-agents'
export const inject = ['tools', 'subagents']
export const inject = ['tools', 'subagents', 'agents']
type ListAgentsEntry =
| {
readonly kind: 'child'
readonly id: string
readonly label: string
readonly status: 'running' | 'complete'
readonly status: 'running' | 'idle' | 'complete'
readonly parent?: string
readonly depth?: number
}
| {
readonly kind: 'diagnostic'
readonly id: string
readonly reason: 'corrupt' | 'unsupported' | 'unavailable'
readonly parent?: string
readonly depth?: number
}
/**
* Refine one candidate's status through the live Agent registry: `running`
* for an active driver, `idle` for a resident Agent between turns (possibly
* waiting on agents it started), and `complete` when no live Agent remains.
*/
function statusOf(agents: { get(id: SessionId): Agent | undefined }, id: SessionId): 'running' | 'idle' | 'complete' {
const agent = agents.get(id)
if (agent === undefined) return 'complete'
return agent.status === 'running' ? 'running' : 'idle'
}
/** Project one service row into the model-facing entry, or omit a one-shot child. */
function project(
agents: { get(id: SessionId): Agent | undefined },
entry: SubagentListEntry,
position?: Pick<SubagentDescendantListEntry, 'parentId' | 'depth'>,
): ListAgentsEntry | undefined {
const at = position === undefined ? {} : { parent: position.parentId as string, depth: position.depth }
if (entry.kind === 'diagnostic') {
return { kind: 'diagnostic', id: entry.id, reason: entry.reason, ...at }
}
// One-shot children cannot be continued by send_message, so the model
// never selects them; discovery still traversed them for descendants.
if (entry.mode !== 'continuable') return undefined
return {
kind: 'child',
id: entry.id,
label: entry.label,
status: statusOf(agents, entry.id),
...at,
}
}
/**
* Register the `list_agents` tool.
* @param ctx - context carrying the tool registry and subagent service.
* @param ctx - context carrying the tool registry, subagent service, and live Agent registry.
*/
export function apply(ctx: Context): void {
ctx.tools.register(defineTool({
name: 'list_agents',
description:
'List your continuable background subagents by durable id and label. Status is a snapshot of the stored '
+ 'record: running means the subagent session is currently live in this process, complete means '
+ 'it exists only in storage and a `send_message` starts a new turn on the same conversation. '
+ 'The snapshot is not a delivery promise — `send_message` performs the authoritative check and '
+ 'may still fail. Children that could not be read are reported as diagnostics instead of being '
+ 'silently dropped.',
parameters: {},
'List your continuable background subagents by durable id and label. Status comes from the live '
+ 'registry: running means the agent is working right now, idle means it is loaded but between turns '
+ '(it may be waiting on agents it started), and complete means it exists only in storage — a '
+ 'direct child remains a `send_message` candidate in every status. The snapshot is not a delivery '
+ 'promise — `send_message` performs the authoritative check and may still fail. Children that could '
+ 'not be read are reported as diagnostics instead of being silently dropped. Scope `descendants` '
+ 'walks the whole tree below you in stable pre-order, annotating each entry with its durable direct-parent '
+ 'session id and depth. You may use `send_message` only for depth-1 entries; deeper entries are '
+ 'candidates for `interrupt_agent` only.',
parameters: {
scope: {
type: 'string',
enum: ['children', 'descendants'],
description: 'children (default) lists direct children only; descendants walks the complete tree below you.',
},
},
output: {
schema: {
type: 'array',
@@ -53,7 +102,9 @@ export function apply(ctx: Context): void {
kind: { type: 'string', required: true, enum: ['child'] },
id: { type: 'string', required: true },
label: { type: 'string', required: true },
status: { type: 'string', required: true, enum: ['running', 'complete'] },
status: { type: 'string', required: true, enum: ['running', 'idle', 'complete'] },
parent: { type: 'string' },
depth: { type: 'number' },
},
},
{
@@ -63,21 +114,31 @@ export function apply(ctx: Context): void {
kind: { type: 'string', required: true, enum: ['diagnostic'] },
id: { type: 'string', required: true },
reason: { type: 'string', required: true, enum: ['corrupt', 'unsupported', 'unavailable'] },
parent: { type: 'string' },
depth: { type: 'number' },
},
},
],
},
},
render: (_args, entries) => [{
render: (args, entries) => [{
type: 'text',
text: entries.length === 0
? '(no subagents)'
: entries.map(entry => entry.kind === 'child'
? `${entry.id} [${entry.status}] — ${entry.label}`
: `${entry.id} [diagnostic: ${entry.reason}]`).join('\n'),
: entries.map((entry) => {
// A descendants row always carries its position; children rows
// never render it. String() spans the schema-optional shape
// without a dead fallback branch.
const at = args.scope === 'descendants'
? ` parent=${String(entry.parent)} depth=${String(entry.depth)}`
: ''
return entry.kind === 'child'
? `${entry.id} [${entry.status}]${at} — ${entry.label}`
: `${entry.id} [diagnostic: ${entry.reason}]${at}`
}).join('\n'),
}],
},
async execute(_args, exec) {
async execute(args, exec) {
const parent = exec.agent
if (!parent) {
// Non-agent callers have no session whose children could be listed.
@@ -85,21 +146,16 @@ export function apply(ctx: Context): void {
}
// The registry drains started tool bodies, so the scan must observe the
// call's signal rather than finish a slow catalog after cancellation.
const entries = await ctx.subagents.listChildren(parent.id, exec.signal)
const visible: ListAgentsEntry[] = []
for (const entry of entries) {
if (entry.kind === 'diagnostic') {
visible.push(entry)
} else if (entry.mode === 'continuable') {
visible.push({
kind: 'child',
id: entry.id,
label: entry.label,
status: entry.activity === 'running' ? 'running' : 'complete',
})
}
if (args.scope === 'descendants') {
const entries = await ctx.subagents.listDescendants(parent.id, exec.signal)
return entries
.map(entry => project(ctx.agents, entry, entry))
.filter(entry => entry !== undefined)
}
return visible
const entries = await ctx.subagents.listChildren(parent.id, exec.signal)
return entries
.map(entry => project(ctx.agents, entry))
.filter(entry => entry !== undefined)
},
}))
}
packages/subagent/tool-subagent-control/tests/list-agents.spec.ts
+156 -9
View File
@@ -12,9 +12,37 @@ import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection'
import SubagentService from '@deepseek-ai/dsh-subagent'
import type { SubagentListEntry } from '@deepseek-ai/dsh-subagent'
import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn'
import type { GenerateOptions, StreamChunk } from '@deepseek-ai/dsh-llm'
import { LlmAdapter } from '@deepseek-ai/dsh-llm'
import { MockAdapter, textResponse } from '../../../core/agent-loop/tests/mock-adapter.ts'
import * as tool from '../src/list-agents.ts'
/** One scripted response that may wait on a caller-released gate before streaming. */
interface GatedEntry {
chunks: StreamChunk[]
gate?: Promise<undefined>
}
/** Adapter whose entries can hold a model call open until the test releases it. */
class GatedAdapter extends LlmAdapter {
readonly requests: GenerateOptions[] = []
constructor(private script: GatedEntry[]) {
super()
}
async * stream(options: GenerateOptions): AsyncIterable<StreamChunk> {
this.requests.push(options)
const entry = this.script.shift()
if (!entry) throw new Error('GatedAdapter: script exhausted')
if (entry.gate) await entry.gate
for (const chunk of entry.chunks) {
if (options.signal?.aborted) throw new Error('aborted')
yield chunk
}
}
}
const testToolSignal = new AbortController().signal
const roots: string[] = []
@@ -22,7 +50,7 @@ afterEach(() => {
for (const root of roots.splice(0)) rmSync(root, { recursive: true, force: true })
})
async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
async function setupWith(adapter: MockAdapter | GatedAdapter) {
const ctx = new Context()
await mountAgentLoopTestDependencies(ctx)
const root = mkdtempSync(join(tmpdir(), 'dsh-tool-list-agents-'))
@@ -33,9 +61,13 @@ async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
await ctx.plugin(SubagentService)
await ctx.plugin(SubagentSpawn, { providerName: 'spawn' })
await ctx.plugin(tool)
ctx.llm.registerAdapter(['mock'], new MockAdapter(script))
ctx.llm.registerAdapter(['mock'], adapter)
const parent = ctx.agentLoop.create(SessionId('parent'), { provider: 'mock', model: 'mock' })
return { ctx, parent }
return { ctx, parent, adapter }
}
async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
return setupWith(new MockAdapter(script))
}
function text(result: { content: { type: string; text?: string }[] }): string {
@@ -67,13 +99,19 @@ async function waitNoActivation(ctx: Context, childId: SessionId): Promise<void>
}
describe('dsh-tool-subagent-control/list-agents', () => {
it('registers list_agents once, globally, with no parameters', async () => {
it('registers list_agents once, globally, with only the optional scope parameter', async () => {
const { ctx } = await setup([])
const schemas = ctx.tools.schemas().filter(schema => schema.name === 'list_agents')
expect(schemas).toHaveLength(1)
const props = (schemas[0]!.parameters as { properties?: Record<string, unknown> }).properties ?? {}
expect(Object.keys(props)).toEqual([])
const parameters = schemas[0]!.parameters as {
properties?: Record<string, { enum?: string[] }>
required?: string[]
}
expect(Object.keys(parameters.properties ?? {})).toEqual(['scope'])
expect(parameters.properties?.scope?.enum).toEqual(['children', 'descendants'])
expect(parameters.required ?? []).toEqual([])
expect(schemas[0]!.description).toContain('send_message')
expect(schemas[0]!.description).toContain('interrupt_agent')
})
it('renders the empty result as (no subagents)', async () => {
@@ -84,7 +122,7 @@ describe('dsh-tool-subagent-control/list-agents', () => {
expect(text(result)).toBe('(no subagents)')
})
it('renders children and diagnostics in array order with the fixed text forms', async () => {
it('renders children and diagnostics in array order with registry-derived statuses', async () => {
const { ctx, parent } = await setup([textResponse('done')])
const started = await ctx.subagents.startContinuable({
provider: 'spawn',
@@ -94,7 +132,8 @@ describe('dsh-tool-subagent-control/list-agents', () => {
})
await waitNoActivation(ctx, started.childId)
// Pin the render deterministically past the service: the tool is a thin
// adapter, so its fixed text forms are what this test pins.
// adapter, so its fixed text forms are what this test pins. Status comes
// from the live Agent registry, stubbed per candidate id.
const entries: SubagentListEntry[] = [
{
kind: 'child',
@@ -120,14 +159,28 @@ describe('dsh-tool-subagent-control/list-agents', () => {
activity: 'running',
hasChildren: true,
},
{
kind: 'child',
id: SessionId('waiting-child'),
label: 'waiting on descendants',
mode: 'continuable',
activity: 'running',
hasChildren: true,
},
{ kind: 'diagnostic', id: SessionId('broken-child'), reason: 'corrupt' },
]
ctx.subagents.listChildren = () => Promise.resolve(entries)
const agents = new Map<string, { status: 'running' | 'idle' }>([
['running-child', { status: 'running' }],
['waiting-child', { status: 'idle' }],
])
vi.spyOn(ctx.agents, 'get').mockImplementation(id => agents.get(id) as never)
const result = await callTool(ctx, 'list_agents', {}, parent)
expect(result.isError).toBe(false)
expect(text(result)).toBe(
`${started.childId} [complete] — real child\n`
+ 'running-child [running] — still working\n'
+ 'waiting-child [idle] — waiting on descendants\n'
+ 'broken-child [diagnostic: corrupt]',
)
})
@@ -186,7 +239,101 @@ describe('dsh-tool-subagent-control/list-agents', () => {
it('has the namespace-plugin export shape', () => {
expect('default' in tool).toBe(false)
expect(tool.name).toBe('tool-subagent-list-agents')
expect(tool.inject).toEqual(['tools', 'subagents'])
expect(tool.inject).toEqual(['tools', 'subagents', 'agents'])
expect(typeof tool.apply).toBe('function')
})
it('walks the complete descendant tree in pre-order with parent and depth annotations', async () => {
const releaseChild = Promise.withResolvers<undefined>()
const releaseGrandchild = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('child'), gate: releaseChild.promise },
{ chunks: textResponse('grandchild'), gate: releaseGrandchild.promise },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'waiting branch',
request: { prompt: [{ type: 'text', text: 'branch work' }], parent },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const grandchild = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'nested leaf',
request: { prompt: [{ type: 'text', text: 'leaf work' }], parent: child },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
// The branch finishes its own turn but stays resident waiting on the
// grandchild it owns: the live-registry `idle` status.
releaseChild.resolve(undefined)
await vi.waitFor(() => {
expect(ctx.agents.get(started.childId)?.status).toBe('idle')
}, { timeout: 5_000 })
const result = await callTool(ctx, 'list_agents', { scope: 'descendants' }, parent)
expect(result.isError).toBe(false)
expect(text(result)).toBe(
`${started.childId} [idle] parent=${parent.id} depth=1 — waiting branch\n`
+ `${grandchild.childId} [running] parent=${started.childId} depth=2 — nested leaf`,
)
releaseGrandchild.resolve(undefined)
await waitNoActivation(ctx, grandchild.childId)
await waitNoActivation(ctx, started.childId)
})
it('omits one-shot intermediates from descendants output while surfacing what they own', async () => {
const { ctx, parent } = await setup([])
// Deterministic service rows: a one-shot intermediate owning a continuable
// leaf, plus a positioned diagnostic. The tool filters only the one-shot.
ctx.subagents.listDescendants = () => Promise.resolve([
{
kind: 'child',
id: SessionId('one-shot-mid'),
label: 'one-shot intermediate',
mode: 'one-shot',
activity: 'inactive',
hasChildren: true,
parentId: parent.id,
depth: 1,
},
{
kind: 'child',
id: SessionId('deep-leaf'),
label: 'deep leaf',
mode: 'continuable',
activity: 'inactive',
hasChildren: false,
parentId: SessionId('one-shot-mid'),
depth: 2,
},
{
kind: 'diagnostic',
id: SessionId('broken-node'),
reason: 'unavailable',
parentId: parent.id,
depth: 1,
},
])
const result = await callTool(ctx, 'list_agents', { scope: 'descendants' }, parent)
expect(result.isError).toBe(false)
expect(text(result)).toBe(
'deep-leaf [complete] parent=one-shot-mid depth=2 — deep leaf\n'
+ `broken-node [diagnostic: unavailable] parent=${parent.id} depth=1`,
)
})
it('forwards the tool cancellation signal to descendant enumeration', async () => {
const { ctx, parent } = await setup([])
const signal = new AbortController().signal
const listDescendants = vi.spyOn(ctx.subagents, 'listDescendants').mockResolvedValue([])
const result = await callTool(ctx, 'list_agents', { scope: 'descendants' }, parent, signal)
expect(result.isError).toBe(false)
expect(listDescendants).toHaveBeenCalledWith(parent.id, signal)
})
})
packages/subagent/tool-subagent-control/tests/tool-subagent-control.spec.ts
+200 -2
View File
@@ -11,9 +11,37 @@ import JsonlSessionPersistence from '@deepseek-ai/dsh-session-persistence-jsonl'
import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection'
import SubagentService from '@deepseek-ai/dsh-subagent'
import * as SubagentSpawn from '@deepseek-ai/dsh-subagent-spawn'
import type { GenerateOptions, StreamChunk } from '@deepseek-ai/dsh-llm'
import { LlmAdapter } from '@deepseek-ai/dsh-llm'
import { MockAdapter, textResponse } from '../../../core/agent-loop/tests/mock-adapter.ts'
import * as tool from '../src/index.ts'
/** One scripted response that may wait on a caller-released gate before streaming. */
interface GatedEntry {
chunks: StreamChunk[]
gate?: Promise<undefined>
}
/** Adapter whose entries can hold a model call open until the test releases it. */
class GatedAdapter extends LlmAdapter {
readonly requests: GenerateOptions[] = []
constructor(private script: GatedEntry[]) {
super()
}
async * stream(options: GenerateOptions): AsyncIterable<StreamChunk> {
this.requests.push(options)
const entry = this.script.shift()
if (!entry) throw new Error('GatedAdapter: script exhausted')
if (entry.gate) await entry.gate
for (const chunk of entry.chunks) {
if (options.signal?.aborted) throw new Error('aborted')
yield chunk
}
}
}
const testToolSignal = new AbortController().signal
const roots: string[] = []
@@ -21,7 +49,7 @@ afterEach(() => {
for (const root of roots.splice(0)) rmSync(root, { recursive: true, force: true })
})
async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
async function setupWith(adapter: MockAdapter | GatedAdapter) {
const ctx = new Context()
await mountAgentLoopTestDependencies(ctx)
const root = mkdtempSync(join(tmpdir(), 'dsh-tool-subagent-control-'))
@@ -32,12 +60,15 @@ async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
await ctx.plugin(SubagentService)
await ctx.plugin(SubagentSpawn, { providerName: 'spawn' })
await ctx.plugin(tool)
const adapter = new MockAdapter(script)
ctx.llm.registerAdapter(['mock'], adapter)
const parent = ctx.agentLoop.create(SessionId('parent'), { provider: 'mock', model: 'mock' })
return { ctx, parent, adapter }
}
async function setup(script: ConstructorParameters<typeof MockAdapter>[0]) {
return setupWith(new MockAdapter(script))
}
function text(result: { content: { type: string; text?: string }[] }): string {
return result.content.filter(block => block.type === 'text').map(block => block.text).join('')
}
@@ -177,8 +208,10 @@ describe('dsh-tool-subagent-control', () => {
await ctx.plugin(SubagentService)
const fiber = await ctx.plugin(tool)
expect(ctx.tools.schemas().some(schema => schema.name === 'send_message')).toBe(true)
expect(ctx.tools.schemas().some(schema => schema.name === 'interrupt_agent')).toBe(true)
await fiber.dispose()
expect(ctx.tools.schemas().some(schema => schema.name === 'send_message')).toBe(false)
expect(ctx.tools.schemas().some(schema => schema.name === 'interrupt_agent')).toBe(false)
})
it('has the namespace-plugin export shape (no stray default)', () => {
@@ -188,3 +221,168 @@ describe('dsh-tool-subagent-control', () => {
expect(typeof tool.apply).toBe('function')
})
})
describe('dsh-tool-subagent-control interrupt_agent', () => {
it('registers interrupt_agent with the single agent_id parameter and current-turn wording', async () => {
const { ctx } = await setup([])
const schemas = ctx.tools.schemas().filter(schema => schema.name === 'interrupt_agent')
expect(schemas).toHaveLength(1)
const props = (schemas[0]!.parameters as { properties?: Record<string, unknown> }).properties ?? {}
expect(Object.keys(props)).toEqual(['agent_id'])
expect(schemas[0]!.description).toContain('current turn')
expect(schemas[0]!.description).toContain('send_message')
})
it('interrupts a running direct child with the parent cause, parking its queue', async () => {
const releaseFirst = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('held'), gate: releaseFirst.promise },
{ chunks: textResponse('parked answer') },
{ chunks: textResponse('waking answer') },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'long work',
request: { prompt: [{ type: 'text', text: 'long work' }], parent },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const queued = await callTool(ctx, 'send_message', {
subagent_id: started.childId,
message: 'parked follow-up',
}, parent)
expect(queued.isError).toBe(false)
const cancelSpy = vi.spyOn(child, 'cancel')
const result = await callTool(ctx, 'interrupt_agent', { agent_id: started.childId }, parent)
expect(result.isError).toBe(false)
expect(text(result)).toBe(`interrupt requested for agent ${started.childId}`)
expect(cancelSpy).toHaveBeenCalledExactlyOnceWith({ kind: 'parent' }, { keepInbox: true })
releaseFirst.resolve(undefined)
await child.whenIdle()
// Parked, not resumed: the queued follow-up waits for a waking send.
expect(adapter.requests).toHaveLength(1)
expect(child.inbox.nextTurn).toHaveLength(1)
const waking = await callTool(ctx, 'send_message', {
subagent_id: started.childId,
message: 'wake up',
}, parent)
expect(waking.isError).toBe(false)
await waitNoActivation(ctx, started.childId)
const loaded = await ctx.sessionPersistence.load(started.childId)
const prompts = loaded.events.flatMap(event => event.type === 'user/message'
? event.data.content.flatMap(block => block.type === 'text' ? [block.text] : [])
: [])
expect(prompts).toEqual(['long work', 'parked follow-up', 'wake up'])
})
it('lets a deep live ancestor interrupt a descendant it did not directly create', async () => {
const releaseChild = Promise.withResolvers<undefined>()
const releaseGrandchild = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('child'), gate: releaseChild.promise },
{ chunks: textResponse('grandchild'), gate: releaseGrandchild.promise },
])
const { ctx, parent } = await setupWith(adapter)
const started = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'child',
request: { prompt: [{ type: 'text', text: 'child work' }], parent },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const child = ctx.agents.get(started.childId)!
const grandchild = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'grandchild',
request: { prompt: [{ type: 'text', text: 'grandchild work' }], parent: child },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
const grandchildAgent = ctx.agents.get(grandchild.childId)!
const cancelSpy = vi.spyOn(grandchildAgent, 'cancel')
const result = await callTool(ctx, 'interrupt_agent', { agent_id: grandchild.childId }, parent)
expect(result.isError).toBe(false)
expect(cancelSpy).toHaveBeenCalledExactlyOnceWith({ kind: 'parent' }, { keepInbox: true })
releaseChild.resolve(undefined)
releaseGrandchild.resolve(undefined)
await waitNoActivation(ctx, grandchild.childId)
await waitNoActivation(ctx, started.childId)
})
it('rejects self, sibling, and unrelated callers without touching the target', async () => {
const releaseA = Promise.withResolvers<undefined>()
const releaseB = Promise.withResolvers<undefined>()
const adapter = new GatedAdapter([
{ chunks: textResponse('a'), gate: releaseA.promise },
{ chunks: textResponse('b'), gate: releaseB.promise },
])
const { ctx, parent } = await setupWith(adapter)
const target = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'target',
request: { prompt: [{ type: 'text', text: 'a' }], parent },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(1) })
const sibling = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'sibling',
request: { prompt: [{ type: 'text', text: 'b' }], parent },
signal: testToolSignal,
})
await vi.waitFor(() => { expect(adapter.requests).toHaveLength(2) })
const targetAgent = ctx.agents.get(target.childId)!
const siblingAgent = ctx.agents.get(sibling.childId)!
const stranger = ctx.agentLoop.create(SessionId('stranger'), { provider: 'mock', model: 'mock' })
const cancelSpy = vi.spyOn(targetAgent, 'cancel')
const self = await callTool(ctx, 'interrupt_agent', { agent_id: target.childId }, targetAgent)
expect(self.isError).toBe(true)
expect(text(self)).toContain('cannot interrupt itself')
const fromSibling = await callTool(ctx, 'interrupt_agent', { agent_id: target.childId }, siblingAgent)
expect(fromSibling.isError).toBe(true)
expect(text(fromSibling)).toContain('not a live descendant')
const fromStranger = await callTool(ctx, 'interrupt_agent', { agent_id: target.childId }, stranger)
expect(fromStranger.isError).toBe(true)
expect(text(fromStranger)).toContain('not a live descendant')
expect(cancelSpy).not.toHaveBeenCalled()
releaseA.resolve(undefined)
releaseB.resolve(undefined)
await waitNoActivation(ctx, target.childId)
await waitNoActivation(ctx, sibling.childId)
})
it('accepts an absent target as a no-op without cold-resuming it', async () => {
const { ctx, parent } = await setup([textResponse('done')])
const started = await ctx.subagents.startContinuable({
provider: 'spawn',
label: 'settled child',
request: { prompt: [{ type: 'text', text: 'child work' }], parent },
signal: testToolSignal,
})
await waitNoActivation(ctx, started.childId)
const settled = await callTool(ctx, 'interrupt_agent', { agent_id: started.childId }, parent)
expect(settled.isError).toBe(false)
expect(text(settled)).toBe(`interrupt requested for agent ${started.childId}`)
const unknown = await callTool(ctx, 'interrupt_agent', { agent_id: 'no-such-agent' }, parent)
expect(unknown.isError).toBe(false)
// No cold resume: the settled target never rematerialized.
expect(ctx.agents.get(started.childId)).toBeUndefined()
})
it('fails loud when invoked without a calling agent', async () => {
const { ctx } = await setup([])
const result = await callTool(ctx, 'interrupt_agent', { agent_id: 'x' })
expect(result.isError).toBe(true)
expect(text(result)).toContain('requires a calling agent')
})
})
scripts/gen-cordis-catalog.ts
+2
View File
@@ -172,7 +172,9 @@ export const LINK_MAP: Readonly<Record<string, string>> = {
ContinuableStart: 'subagent.md',
ContinuableStartSpec: 'subagent.md',
CoordinatorMessageSource: 'subagent.md',
SubagentDescendantListEntry: 'subagent.md',
SubagentFollowupOptions: 'subagent.md',
SubagentInterruptAuthority: 'subagent.md',
SubagentListEntry: 'subagent.md',
SubagentProvider: 'subagent.md',
SubagentReportDelivery: 'subagent.md',
scripts/gen-tool-catalog.ts
+13 -2
View File
@@ -399,10 +399,11 @@ const TOOL_PACKAGES: ToolPackage[] = [
pkg: '@deepseek-ai/dsh-tool-subagent-control',
dir: 'tool-subagent-control',
source: {
interrupt_agent: 'packages/subagent/tool-subagent-control/src/index.ts',
list_agents: 'packages/subagent/tool-subagent-control/src/list-agents.ts',
send_message: 'packages/subagent/tool-subagent-control/src/index.ts',
},
requires: ['ctx.tools', 'ctx.subagents', 'ctx.sessionProjections (list_agents catalog rows)'],
requires: ['ctx.tools', 'ctx.subagents', 'ctx.agents and ctx.sessionProjections (list_agents only)'],
writes: ['tool/call', 'tool/result', 'child session events through ctx.subagents'],
async mount(ctx) {
await ctx.plugin(SubagentService)
@@ -414,7 +415,7 @@ const TOOL_PACKAGES: ToolPackage[] = [
await ctx.plugin(ToolSubagentListAgents)
},
note:
'The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows are served through the sessionProjections registry).',
'The globally named control tools over continuable background subagents: provider-bound `tool-subagent` instances register distinct delegation tools, while this package registers `send_message` and `interrupt_agent` once, plus `list_agents` from its separately loaded `/list-agents` plugin (whose catalog rows use the sessionProjections and live Agent registries).',
},
{
pkg: '@deepseek-ai/dsh-tool-subagent-report',
@@ -654,6 +655,16 @@ async function main(): Promise<void> {
process.exit(0)
}
console.error(`gen-tool-catalog: ${OUT} is stale. Run \`pnpm run gen-tool-catalog\` and commit ${OUT}.`)
const committedLines = committed?.split('\n') ?? []
const generatedLines = content.split('\n')
const lineCount = Math.max(committedLines.length, generatedLines.length)
for (let index = 0; index < lineCount; index += 1) {
if (committedLines[index] === generatedLines[index]) continue
console.error(`gen-tool-catalog: first difference at line ${index + 1}`)
console.error(` committed: ${JSON.stringify(committedLines[index])}`)
console.error(` generated: ${JSON.stringify(generatedLines[index])}`)
break
}
process.exit(1)
}
scripts/type-equiv.manifest.json
+10
View File
@@ -1150,6 +1150,16 @@
"symbol": "SubagentFollowupOptions",
"source": "packages/subagent/subagent/src/continuation.ts"
},
{
"doc": "docs/core-data-structures/subagent.md",
"symbol": "SubagentInterruptAuthority",
"source": "packages/subagent/subagent/src/continuation.ts"
},
{
"doc": "docs/core-data-structures/subagent.md",
"symbol": "SubagentDescendantListEntry",
"source": "packages/subagent/subagent/src/list-children.ts"
},
{
"doc": "docs/core-data-structures/subagent.md",
"symbol": "ContinuableStart",
tsconfig.host.json
+2
View File
@@ -54,6 +54,8 @@
"apps/web/tests/produced-files.e2e.ts",
"apps/web/tests/produced-file-mentions.e2e.ts",
"apps/web/tests/subagent-conversation.e2e.ts",
"apps/web/tests/subagent-interrupt.e2e.ts",
"apps/web/tests/subagent-interrupt-ui.e2e.ts",
"apps/web/tests/sidebar-subagent-activity.e2e.ts",
"apps/web/tests/bash-abort-row.e2e.ts",
"apps/web/tests/skill-tool-row.e2e.ts",