diff --git a/.agents/notes/implemented/architecture/2026-06-11-dev-invariants-over-deep-readonly.md b/.agents/notes/implemented/architecture/2026-06-11-dev-invariants-over-deep-readonly.md index 90e215019b..fb1dc93159 100644 --- a/.agents/notes/implemented/architecture/2026-06-11-dev-invariants-over-deep-readonly.md +++ b/.agents/notes/implemented/architecture/2026-06-11-dev-invariants-over-deep-readonly.md @@ -28,11 +28,11 @@ This guarantee belongs in `Session`, not in an optional listener, because every `deriveMessages()` projects logged surface events into detached, deep-frozen `Message` objects and returns a fresh array snapshot. Request assembly can therefore combine derived history with other inputs without exposing a path back into the log. The cache reuses safe immutable projections rather than recloning the complete history for each model call. -### The invariants plugin checks relationships +### Package-owned invariant companions check relationships -`dsh-invariants` is a pure-listener development plugin. It does not freeze records and has no configuration; disposal removes only its assertions. It checks rules that require trace state or observation of another seam, including monotonic sequence numbers, turn and step nesting, tool-call/result pairing, legal agent-status transitions, subject-correct scoped dispatch, and equality between a loop-built request and the request reconstructed from its session-log prefix. +`dsh-invariants` registers the configurable `ctx.invariants` service and contains no product checks. Every package publishes a `./invariant` ownership companion; `dsh-session`, `dsh-agent`, `dsh-scope`, and `dsh-agent-loop` currently add the rules that require trace state or observation of another seam: monotonic sequence numbers, turn and step nesting, tool-call/result pairing, legal agent-status transitions, subject-correct scoped dispatch, and equality between a loop-built request and the request reconstructed from its session-log prefix. Global enablement and package-name regex filters belong to the service ([package-owned invariant service](2026-07-19-package-owned-invariant-service.md)). -When the plugin attaches to an existing or seeded session, it replays the immutable log to rebuild trace state. This makes hot reload safe in the middle of a turn without giving the plugin ownership of session storage. +When the session companion attaches to an existing or seeded session, it replays the immutable log to rebuild trace state. The service gives each contribution a disposable child fiber, so hot reload is safe in the middle of a turn without giving diagnostics ownership of session storage. ## Alternatives considered @@ -53,6 +53,6 @@ Detaching `deriveMessages()` would protect the most common request path but leav - Every accepted live or seeded session event is detached from caller-owned inputs and deeply immutable before any observer can receive it. - `session.events` exposes stable immutable snapshots instead of the private growing array. - Request-side mutation cannot reach stored history through derived messages. -- Development builds can enable relational assertions without changing storage behavior, and disposing or omitting the plugin does not weaken log immutability. -- `dsh-invariants` has no `Config` surface because it has no behavior to tune. +- Development builds can enable relational assertions without changing storage behavior, and disposing or filtering a companion does not weaken log immutability. +- `dsh-invariants` configures global enablement plus package allow/block regex lists; each check remains owned and tested by its product package. - The runtime boundary carries a recursive snapshot-and-freeze cost once per accepted event; later readers and cached projections reuse the owned immutable records. diff --git a/.agents/notes/implemented/architecture/2026-06-11-structured-error-taxonomy.md b/.agents/notes/implemented/architecture/2026-06-11-structured-error-taxonomy.md index eacd409a57..b2f4ea66b4 100644 --- a/.agents/notes/implemented/architecture/2026-06-11-structured-error-taxonomy.md +++ b/.agents/notes/implemented/architecture/2026-06-11-structured-error-taxonomy.md @@ -10,7 +10,7 @@ Failures crossed seams as bare strings. A tool error flattened to a text block A single `HarnessError extends Error` base in `dsh-llm` (the leaf package every other imports — no new dependency edge): a stable `code` distinct from `message`, `cause` chaining via `ErrorOptions`, and `name` defaulting to the subclass. `isHarnessError` narrows at seams. -- `LlmError`, `ToolArgsError` (dsh-tools), and `InvariantError` (dsh-invariants) now extend it, keeping their existing codes. +- `LlmError` and `ToolArgsError` (dsh-tools) extend it, keeping their existing codes. - `ToolExecutionResult` gains optional `error: { name, code }`, populated in the registry's catch when the thrown value is a `HarnessError`. The agent loop forwards it onto the `tool/result` session event (which gained the same optional field), so the structured failure survives into the log for retry/sandbox plugins and replay. The model-facing text block is unchanged. - The loop's `toError` wraps a non-Error throw in a `HarnessError` (`code: 'UNKNOWN'`, original chained as `cause`) instead of a bare `Error`, so even a bad throw carries a routable code into the session `error` event (which already surfaced `code`). @@ -19,6 +19,6 @@ A single `HarnessError extends Error` base in `dsh-llm` (the leaf package every - Errors are machine-routable end-to-end: a plugin can branch on `error.code` rather than substring-matching a message. - One base class is imported widely, but it lives in the package everyone already depends on, so the cost is a single import, not a new edge. - `deriveMessages` does not surface `error` into model history — the model still sees the text block; the structured field is for code and replay. -- Argument validation and dev invariants retain their existing codes and behavior; the shared base adds cross-seam routing metadata without changing model-facing text. +- Argument validation retains its existing code and behavior; package-owned diagnostic invariants carry their stable code independently so the invariant registry does not import a product package. The shared base adds cross-seam routing metadata without changing model-facing text. diff --git a/.agents/notes/implemented/architecture/2026-06-14-session-persistence.md b/.agents/notes/implemented/architecture/2026-06-14-session-persistence.md index e84d7fefd3..ffb3afa51a 100644 --- a/.agents/notes/implemented/architecture/2026-06-14-session-persistence.md +++ b/.agents/notes/implemented/architecture/2026-06-14-session-persistence.md @@ -13,7 +13,7 @@ The [event-sourced model](2026-06-11-event-sourced-sessions.md) makes the append Persistence is an abstract **capability seam** ([capability seams](2026-06-13-capability-seams.md), the `dsh-bash` template), not loop or core logic: 1. **Interface** (`dsh-session-persistence`, `ctx.sessionPersistence`) — an abstract `SessionPersistence` service: `create`/`append`/`load`/`list`. Its persisted unit IS the existing `SessionEvent` (`{ type, seq, time, data }`), reused verbatim — no conversion type. -2. **Implementation** (`dsh-session-persistence-jsonl`) — an append-only JSONL log per session (a `SessionHeader` line then one `SessionEvent` per line, verbatim **including `assistant/chunk`**). +2. **Implementation** (`dsh-session-persistence-jsonl`) — an append-only logical JSONL log per session (a `SessionHeader` line then one `SessionEvent` per line, verbatim **including `assistant/chunk`**), encoded as [checksummed Zstandard frames by default](2026-07-19-zstandard-jsonl-session-logs.md) or raw lines by configuration. Key choices recorded here because they are durable, contested, and surprising: diff --git a/.agents/notes/implemented/architecture/2026-06-15-turn-enclosure-invariant.md b/.agents/notes/implemented/architecture/2026-06-15-turn-enclosure-invariant.md index 74ab8b6bf3..00352b01c4 100644 --- a/.agents/notes/implemented/architecture/2026-06-15-turn-enclosure-invariant.md +++ b/.agents/notes/implemented/architecture/2026-06-15-turn-enclosure-invariant.md @@ -21,7 +21,7 @@ In case 2, if the injected `context/message` is the last event before a flush/di - An `agent.inject()` made while the agent is **running** joins the already-open turn. While the current step executes assistant tool calls, accepted context waits in arrival order until that batch settles, then appends after every recorded result and before the turn closes even when execution is interrupted. - An `agent.inject()` made while **idle** wraps its `context/message` in a one-shot turn: `turn/start{trigger:{kind:'injection'}}` → `context/message` → `turn/end{completed}`. A new `injection` variant joins the merge-extensible `TurnTriggerMap`. - The loop derives the next turn number from the log each iteration (`lastTurnNumber(session) + 1`) instead of keeping a private counter, so an idle injection's one-shot turn cannot collide with the next real turn's number. -- The `dsh-invariants` plugin **enforces** the invariant in dev: a `user/message` / `context/message` / `steering/message` appended while no turn is open throws an `InvariantError`. +- The `dsh-session/invariant` companion registers the check with `ctx.invariants`: when selected, a `user/message` / `context/message` / `steering/message` appended while no turn is open throws an `InvariantError` attributed to `@deepseek-ai/dsh-session`. The serializability invariant is enforced at the same source boundary (`Session.append` throws on non-JSON-serializable data), so "what may enter the log" is now governed in one place rather than discovered downstream by whichever backend happens to be watching. diff --git a/.agents/notes/implemented/architecture/2026-06-18-agent-lifecycle-and-ownership-seams.md b/.agents/notes/implemented/architecture/2026-06-18-agent-lifecycle-and-ownership-seams.md index 125ef0946c..3d84c8074b 100644 --- a/.agents/notes/implemented/architecture/2026-06-18-agent-lifecycle-and-ownership-seams.md +++ b/.agents/notes/implemented/architecture/2026-06-18-agent-lifecycle-and-ownership-seams.md @@ -10,9 +10,9 @@ Several ACP and tool-bash limitations were symptoms of the same missing seam: pl Three seams: the queue-aware cancel, the `AgentHandle` disposer, and the bash owner token. -### 1. Queue-aware `Agent.cancel(reason?)` +### 1. Queue-aware `Agent.cancel(cause?)` -A new `cancel()` verb on the `Agent` interface — the single public stop primitive. (It originally shipped alongside a narrower step-only `abort()`; that verb was later removed as unused, leaving `cancel()` the only public way to stop work.) It clears the inbox's queued + steering FIFOs, aborts the in-flight step if any, and drives a **turn-scoped cancellation marker** the driver loop checks at every turn-decision point — so a prompt that is queued-but-not-yet-started never runs, a cancel landing in the pre-step / continuation window drops the about-to-run turn (ending it `aborted`), and a later prompt cannot be batched into the cancelled turn. `whenIdle()` reaches post-cancel quiescence. ACP `session/cancel` maps to `cancel()`. The marker is armed ONLY when there is something to cancel, so an idle no-op cancel cannot strand the next prompt. +A new `cancel()` verb on the `Agent` interface — the single public stop primitive. (It originally shipped alongside a narrower step-only `abort()`; that verb was later removed as unused, leaving `cancel()` the only public way to stop work.) It clears the inbox's queued + steering FIFOs, aborts the active turn if any, and keeps a cause-less pre-run marker so a prompt cancelled before claim never runs while a later prompt remains independent. An effective call emits `agent/cancel-requested` with the typed `user | parent` cause before clearing or aborting; idle cancellation emits nothing and cannot strand the next prompt. `whenIdle()` reaches post-cancel quiescence, and ACP `session/cancel` maps to `user`. The [explicit turn-cancellation decision](2026-07-16-explicit-turn-cancellation.md) owns the current cause, signal-lifetime, and cooperative-settlement contract. ### 2. `AgentHandle` async disposer @@ -29,7 +29,7 @@ Background-task ownership moved from a `tool-bash` plugin-local `Map/`, so a package's location said nothing about whether it was core product API, a swappable capability seam, a provider adapter, a product integration, or example/test support. The package README carried a `FIXME(package-hierarchy)` and `scripts/publint-all.ts` a `TODO(package-inventory)` flagging exactly this. Core packages, provider integrations, capability seams, example UI support, and snapshot-only replay support all looked equally foundational. diff --git a/.agents/notes/implemented/architecture/2026-06-21-bounded-llm-request-recovery.md b/.agents/notes/implemented/architecture/2026-06-21-bounded-llm-request-recovery.md new file mode 100644 index 0000000000..28de5eb97c --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-06-21-bounded-llm-request-recovery.md @@ -0,0 +1,146 @@ +# Agent Note: Bounded recovery for transient LLM request failures + +Status: implemented + +## Problem + +`dsh-llm` can report provider failures either by throwing during adapter dispatch or iteration or by ending with `finish { kind: 'error' | 'aborted' }`. The final adapter boundary tags thrown failures so `dsh-agent-loop` can distinguish them from middleware and result-processing defects, and the loop normalizes both delivery forms into `agent/request-error` after closing the failed step. The default decision is `fail`; `dsh-compact-basic` is the only shipped recovery listener, and it retries a canonical context-window overflow only after compaction proves that the durable surface shrank. + +That boundary is already safe for another request attempt. Raw `assistant/chunk` events carry the failed `turn` and `step`, message derivation ignores them unless a successful `assistant/message` cites them, tool calls are dispatched only after a successful terminal finish and assembly, and a retry opens a new numbered step from the durable log. The harness therefore does not need a second response lifecycle or tentative-output protocol to keep two attempts separate. + +The prior boundary left three narrower gaps. + +- Provider failures retain only a message and usually a code. HTTP status, retry delay, and provider request id are discarded or recoverable only through provider-specific error objects, so generic recovery cannot make or explain a decision without parsing text. +- Retry ownership differs by adapter. The hand-written DeepSeek adapter makes one attempt, while pi-ai profiles can enable opaque library retries. Combining hidden transport retries with an `agent/request-error` listener would multiply attempts and omit intermediate failures from the session log. +- A recovered failure has no durable status fact. The failed step and chunks remain reconstructable, but an observer cannot tell whether the agent is deliberately backing off, for how long, or why. A long silent wait looks like a stalled loop. + +The goal is bounded recovery from transient failures of the same explicit provider/model request. Provider or model failover, response splicing, and semantic output repair are different problems and have no current consumer. + +## Decision + +### Preserve failure facts without embedding policy + +`@deepseek-ai/dsh-llm` exports one JSON-serializable `LlmFailure` payload: + +```ts ignore-check +type ProviderRequestId = Branded<'ProviderRequestId'> + +interface LlmFailure { + message: string + code: string + status?: number + providerRetryAfterMs?: number + requestId?: ProviderRequestId +} +``` + +`code` remains the provider-neutral machine-routing taxonomy established by `HarnessError`; the new fields are observations from the provider boundary. `ProviderRequestId` is owned and constructed by `dsh-llm`, then serializes as its provider-issued string. The payload deliberately has no `retryable`, `failover`, `partialOutput`, provider, model, phase, or route id fields. Retryability belongs to policy, provider/model are already in the durable request header, and partial output is derived from the failed step's `assistant/chunk` events. + +`LlmError` carries `failure: LlmFailure` and preserves `failure.code === error.code`. `FinishReasonMap.error` and `FinishReasonMap.aborted` carry the same payload instead of parallel failure shapes. An adapter-thrown `Error` keeps its exact object identity: the final-adapter scope associates the normalized facts with that object in call-local sidecar state and rethrows it unchanged; a non-`Error` throw is wrapped as today. `llmFailureOf(stream, error)` retrieves those facts alongside the existing provenance check, while an in-band finish without an error object becomes a new `LlmError`. This preserves listeners that key on error type or identity while giving all final-adapter failures, including unknown SDK exceptions, an `UNKNOWN` terminal payload. + +The agent loop keeps `RequestError` as that exact error object and passes `LlmFailure` as a separate argument to `agent/request-error`; it does not mutate possibly frozen third-party errors. It also uses the payload when converting an in-band finish and when recording an unrecovered `turn/end.reason`. + +Adapters extract structured facts before falling back to message inspection. They validate HTTP status, parse `Retry-After` seconds or dates into a positive finite millisecond delay, brand the provider request id when exposed, and distinguish their own timeout from the caller's abort. Provider-specific codes and messages may refine a mapping, but no recovery listener parses them. + +The initial shared transient-code set is intentionally small: the adapters' existing `RATE_LIMIT` and `SERVER` mappings plus explicit `TIMEOUT` and `TRANSPORT` codes for the two missing remote-failure families. Authentication, quota, invalid request, context overflow, protocol, abort, and unknown failures keep distinct stable codes and are not transient by default. Adding a code requires adapter fixtures and a documented policy decision; it does not require expanding a second failure-class enum. + +### Put retry policy on the existing failed-step seam + +`@deepseek-ai/dsh-llm-retry` is a function plugin that listens to `agent/request-error`. It introduces no service or new loop branch; the agent-loop package changes only the data carried through its existing failed-step recovery control flow. + +The `agent/request-error` seam carries the current `LlmFailure` and an immutable list of prior failures that led to another request attempt in this consecutive recovery sequence. `dsh-llm-retry` counts only prior failures whose codes are in its configured transient set, while `dsh-compact-basic` counts only prior context-overflow failures. A successful model request clears the history. Alternating transient and context-overflow failures therefore consume their owning policy budgets independently; the maximum request count is one plus the sum of the finite budgets of the loaded recovery policies. + +The plugin resolves and validates this deployment configuration at load: + +```ts ignore-check +interface Config { + maxTransientRetries?: number + initialDelayMs?: number + maxDelayMs?: number + jitterRatio?: number + retryableCodes?: string[] +} +``` + +The defaults are two transient retries, a 500 millisecond initial delay, a 10 second delay cap, 10 percent jitter, and the four transient codes above. The count and delay bounds match the conservative edge of the inspected implementations: [OpenCode uses two request retries with 500 ms/10 s bounds](https://github.com/anomalyco/opencode/blob/9976269ab1accfc9f9dc98a4a688c516934de422/%70ackages/llm/src/route/executor.ts#L36-L39), [Pi separates three agent-level retries from provider retries and defaults provider retries to zero](https://github.com/earendil-works/pi/blob/3da591ab74ab9ab407e72ed882600b2c851fae21/%70ackages/coding-agent/docs/settings.md#L139-L147), and [Codex uses finite request/stream budgets plus a five-minute idle timeout](https://github.com/openai/codex/blob/0fb559f0f6e231a88ac02ea002d3ecd248e2b515/codex-rs/model-provider-info/src/lib.rs#L25-L33). Ten percent follows [Codex's bounded jitter](https://github.com/openai/codex/blob/0fb559f0f6e231a88ac02ea002d3ecd248e2b515/codex-rs/codex-client/src/retry.rs#L40-L47). Two retries mean at most three provider requests when no other recovery policy applies. `maxTransientRetries` is a non-negative integer, delays are positive finite numbers with `initialDelayMs <= maxDelayMs`, `jitterRatio` is in `[0, 1]`, and codes are non-empty and unique. These are Cordis config fields rather than hidden constants so deployments can choose different cost and latency budgets. + +For an eligible failure with budget remaining, the one-based transient retry count uses bounded exponential backoff. A valid `providerRetryAfterMs` replaces exponential backoff only when it does not exceed `maxDelayMs`; a longer provider delay causes delegation instead of an earlier retry that violates the provider instruction. Local backoff multiplies by an injected random factor in `[1 - jitterRatio, 1 + jitterRatio]` and clamps the final value to `maxDelayMs`; provider delay is not jittered. + +The plugin owns a lifetime `AbortController` and tracks every active backoff callback. Each wait fuses the waterfall's turn signal with that lifetime signal. Effect cleanup first unregisters the listener, then aborts and awaits the active callbacks; a captured callback whose lifetime signal aborts returns `fail` and can neither retry nor enter the rest of its captured waterfall after disposal. This makes HMR disposal quiescent even though Cordis has already captured the listener. + +Before sleeping, `dsh-llm-retry` appends one non-surface `llm/retry` session event containing the turn, failed step, one-based transient retry number, configured maximum, scheduled delay, and `LlmFailure`. The plugin owns the `SessionEventMap` augmentation; `dsh-session` remains generic persistence and does not absorb the optional policy's vocabulary. The event says what was scheduled, not that the next request completed; cancellation during the delay is subsequently visible on `turn/end`. The event ships only with a production renderer and replay/snapshot coverage, because its purpose is operational state rather than trace collection. + +The listener calls `next()` for a non-transient code, an exhausted policy budget, or an over-cap provider delay. This preserves composition with context-overflow recovery and later policy plugins. It returns `{ action: 'retry' }` only after the delay completes under both signals; turn cancellation and plugin disposal return `fail`, after which the loop's cancellation/disposal checks remain authoritative. + +The agent-spine demo bundle loads the plugin so the shared stdio/TUI, one-shot CLI, and ACP example compositions use the same bounded policy. Library consumers retain explicit plugin composition: omitting the plugin leaves `agent/request-error` at its current fail default. + +### Make one layer own visible attempts + +Adapters perform one provider request per `stream()` call. The pi-ai adapter removes public `maxRetries` and `maxRetryDelayMs` profile fields and disables library retries; the hand-written adapter keeps its current single-attempt behavior. This prevents an SDK budget from multiplying the agent budget and ensures every transient retry is represented by a closed failed step plus `llm/retry`. + +`ctx.llm.stream()` remains the raw one-attempt waterfall. Direct callers such as compaction summarization receive the structured failure but do not gain automatic retry, because they have no agent step boundary or general durable place to separate attempts. A future direct-call consumer may justify a buffering helper that retries only before emitting a chunk; this decision adds no such helper. + +### Bound stalled streams where they can be stopped + +Each adapter exposes a validated `streamIdleTimeoutMs` configuration field with the five-minute prior-art default cited above. The interval is capped at Node's maximum timer delay so it cannot be clamped to one millisecond. It covers each outstanding iterator `next()` from demand to the next valid `StreamChunk`; time a consumer spends between `next()` calls is not provider idle time. + +`@deepseek-ai/dsh-timeout` exposes a rearmable idle-watchdog primitive. One stable local `AbortController` is fused with the caller signal and passed to the transport for the whole adapter call; each outstanding `next()` arms the watchdog, resolution disarms it, and the next demand rearms it. Timeout aborts that stable controller with a capability-owned `TimeoutReason`, and `finally` clears the timer. The adapter classifies its watchdog as `TIMEOUT` and an earlier upstream abort as `ABORTED`. The existing one-shot `deadline()` is not presented as a sliding timer. + +Boundary tests prove termination at both actual transports. The hand-written adapter aborts its fetch/reader, and the pi-ai adapter maps the stable signal through the SDK and proves the SDK closes the response. A timer that merely rejects a consumer promise while leaving the request running does not satisfy the contract. + +### Keep attempts separate in the existing log + +A failed attempt may leave `assistant/chunk` events in its closed step, but it never appends `assistant/message` and never dispatches a tool. A retry opens the next numbered step, reconstructs the request from the durable surface, and produces its own chunks. UIs may render live chunks while a step is open, then mark or clear that transient view when `llm/retry` identifies the failed step or `turn/end` records terminal failure; message derivation continues to ignore the failed chunks. + +If recovery is exhausted, the final failure is stored once on `turn/end.reason` with the structured facts. If transient recovery continues, `llm/retry` is the durable home for that attempt's failure and delay. No standalone final-error event or response-id vocabulary is added. + +## Out of scope + +- Automatic provider or model failover. Requests already select one explicit provider and model, and the provider registry deliberately has one adapter owner per provider. +- Retrying or continuing after a successful terminal finish, or splicing chunks from two attempts into one assistant message. +- Repairing malformed tool arguments, refusals, content filters, or other semantic model output. +- Unbounded retries, unattended retry-until-cancelled behavior, circuit breakers, shared provider health, or cross-agent retry budgets. +- Changing `llm/stream` into a response lifecycle or adding convenience generation APIs without a production consumer. + +## Alternatives considered + +- **Retry inside `llm/stream` or the provider SDK** — rejected because a raw stream has no durable attempt boundary after emitting chunks, hidden SDK retries multiply budgets, and neither path can record each failed attempt consistently. +- **Add response start, interrupted, discarded, failed, and committed events to `dsh-llm`** — rejected because the agent log already separates raw chunks, successful messages, and numbered attempts. A second state machine would duplicate ownership without enabling the bounded same-route retry. +- **Add logical routes, capability matrices, and failover selection** — rejected because current requests already name provider and model explicitly, one adapter owns each provider, and no current consumer requires automatic fallback or can prove semantic compatibility. +- **Put `retryable` or `failover` on `LlmFailure`** — rejected because adapters report facts while deployment policy decides action. The same 429 may be retried in an interactive bundle and rejected in a cost-capped batch. +- **Retry forever while the caller remains active** — rejected because it gives one request unbounded cost and latency. Visible status makes bounded waiting understandable; it does not make an unlimited budget safe. +- **Log retry status only through the process logger** — rejected because process logs do not reconstruct session behavior and cannot drive replayed UI state. +- **Keep only flat codes** — rejected because retry delay and provider request id are structured provider facts, and HTTP status is necessary for diagnosis when different wire failures share one stable code. + +## Verification + +- `LlmFailure` is the single serializable payload for thrown, error-finish, and aborted-finish final-adapter failures; normalization preserves stable code, status, retry delay, branded provider request id, error cause, and caller-abort versus adapter-timeout classification where available. +- An adapter-thrown `Error` reaches `agent/request-error` as the exact same object while its sidecar `LlmFailure` reaches the adjacent argument; tests retain the existing identity assertion for extensible and frozen third-party errors. +- DeepSeek and pi-ai adapter tests cover representative 400, 401/403, 429, 5xx, connection, malformed/truncated stream, timeout, abort, retry-after seconds/date, request-id, and unknown-SDK-error paths without recovery policy parsing message text. +- Pi-ai pins the SDK option to zero retries and performs one observed wire attempt for a retryable provider response; separate tests make removing either boundary fail. +- `agent/request-error` carries current failure facts plus immutable prior-retried failure facts; a success clears that history, and alternating transient/context-overflow integration tests prove the two policies consume only their own finite budgets. +- `dsh-llm-retry` validates every config field at Loader startup, delegates all ineligible paths with `next()`, and makes at most `maxTransientRetries + 1` provider requests when no other policy applies. +- HMR-during-backoff tests prove disposal unregisters the listener, aborts and awaits its captured callbacks, emits no retry decision after disposal, and leaves no timer or promise alive. +- Pure unit tests cover transient-code selection, exponential backoff and jitter bounds, valid and over-cap `Retry-After`, exhausted budgets, deterministic timer/random seams, and abort during backoff. +- Real agent-loop tests cover failure before chunks, partial chunks then failure, thrown and in-band failures, retry to success in a new step, exhaustion to structured `turn/end.reason`, and composition with `dsh-compact-basic` context-overflow recovery. +- The partial-chunk integration test proves failed chunks remain attributed to the failed step, no assistant message or tool side effect is committed for that step, and the successful retry has distinct provenance. +- The plugin-owned `llm/retry` event is non-surface, survives JSONL and SQLite round trips, is ignored by message derivation, and drives TUI retraction plus durable discarded-attempt markers in append-only ACP and stdio streams. Keyless snapshots cover scheduling, cancellation, success, and exhaustion. +- Idle-watchdog tests prove the stable signal is rearmed only while `next()` is outstanding, disarmed during consumer think time and in `finally`, and classified separately from a total-call deadline and an earlier caller abort; adapter tests prove the signal stops the underlying request rather than merely detaching it. +- Direct `ctx.llm.stream()` callers remain single-attempt and receive the same structured failure facts. + +## Consequences + +- Every transient recovery attempt is visible as a closed step plus `llm/retry`, and the bounded policy prevents hidden SDK retries from multiplying cost. A retry can still duplicate provider billing even when no chunk arrived; the finite attempt budget limits but cannot remove that risk. +- Provider SDKs may hide status or retry headers. Those adapters retain the stable facts they expose and otherwise use a coarse code rather than letting recovery policy parse fragile text. +- Durable retry events expand the session protocol and UI state machine. Shipping the event and its consumer together prevents an unused telemetry vocabulary, but later schema changes still require persistence and replay work. +- Clearing a failed step's live chunks can visibly retract output. That is preferable to presenting discarded text or partial tool JSON as committed history, and snapshots pin the transition. +- Adapter-local idle enforcement stops stalled transports without counting consumer think time. Contract tests at each transport boundary guard against SDK drift. +- Multiple recovery plugins add their finite budgets. Their classifiers remain disjoint here; an overlapping classifier would be registration-order policy and must be documented and tested by the plugins that introduce it. + +## Related + +- [Structured error taxonomy](../../implemented/architecture/2026-06-11-structured-error-taxonomy.md) owns stable machine-routable codes and cause chaining. +- [Reconstructable requests](../../implemented/architecture/2026-07-05-reconstructable-requests.md) makes provider/model and complete request inputs durable before dispatch. +- [Timeout deadline library](../../implemented/architecture/2026-07-06-timeout-deadline-library.md) separates shared deadline classification from capability-owned termination. +- [After-call compaction pressure and context-overflow recovery](../../implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md) owns the current closed-step request-recovery seam and bounded overflow retry. +- [Provider-routed LLM adapters](../../implemented/architecture/2026-07-14-provider-routed-llm-adapters.md) owns explicit provider/model routing and the one-adapter-per-provider invariant. diff --git a/.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md b/.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md index 631e5b570c..cdd37091a2 100644 --- a/.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md +++ b/.agents/notes/implemented/architecture/2026-07-05-prompt-variables-and-tool-guidance-ownership.md @@ -8,7 +8,7 @@ The assembled system prompt had four defects, all of one family: facts the harne **The model could not know its own name.** `AgentOptions.model` drives every request, but no prompt text carried it — and nothing COULD carry it: sections in `dsh-system-prompt` were context-global while the model name is per-agent, and `assemble()` took no per-agent input at all. -**Tool guidance was hand-written prose in leaf YAML.** The bash/subagent/todo_write usage guidance lived in the `systemPrompt` strings of `examples/repl-agent/cordis.yml` and `examples/acp-agent/cordis.yml` — two drifting copies (the ACP one was already abridged) — while `dsh-tool-fs` and `dsh-tool-web` owned their guidance as `ctx.systemPrompt.section()` contributions. Loading or dropping a tool plugin meant editing every deployment's persona by hand; both YAMLs carried a `FIXME(config-comments)` apologizing for a symptom of the split, and the stdio welcome banner hand-enumerated the tool set too. +**Tool guidance was hand-written prose in leaf YAML.** The bash/subagent/todo_write usage guidance lived in the coding-agent and ACP persona strings — two drifting copies (the ACP one was already abridged) — while `dsh-tool-fs` and `dsh-tool-web` owned their guidance as `ctx.systemPrompt.section()` contributions. Loading or dropping a tool plugin meant editing every deployment's persona by hand; both YAMLs carried a `FIXME(config-comments)` apologizing for a symptom of the split, and the old terminal welcome banner hand-enumerated the tool set too. **The persona rendered after tool guidance.** The loop string-joined `agent.options.systemPrompt` AFTER the assembled sections, so the model read "Use the read tool…" before "You are a coding agent" — backwards relative to the identity-first convention (Claude Code, Codex) and a second composition path besides the section pipeline. @@ -56,7 +56,7 @@ Per-tool semantics and selection guidance live in tool descriptions. Prompt sect ## Shipped invariants -- The repl-agent prompt renders identity, persona with the interpolated model, then fs/bash/web guidance through one assembly path. +- The tui-agent prompt renders identity, persona with the interpolated model, then fs/bash/web guidance through one assembly path. - Fork and fresh subagent descriptions reflect whether the provider inherits completed conversation turns; the tool appears, disappears, and is reworded with provider lifecycle changes. - Unknown, valueless, malformed, or unbalanced variable references name the section and throw; duplicate section, variable, and tool registrations also throw. - Snapshot replay is prompt-independent: it keys recorded chunk streams by turn and step without comparing the outgoing request. diff --git a/.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md b/.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md index 162a83ad0c..99a7633c5b 100644 --- a/.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md +++ b/.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md @@ -12,7 +12,7 @@ The reference shape for the happy path is MiniCode's `LLMClient`: a stateful con ### The principle -**Model-visible ⟺ logged.** Anything that reaches a model request must be recorded in the session log. The checkable consequence: **every conversation request the loop sends is a pure function of the session log** — anyone holding the log reconstructs it byte-for-byte. Scope, stated precisely: the guarantee covers the loop-built `GenerateOptions`; provider wire bytes follow from it because both adapters' serialization is a pure per-message function at a pinned code version; direct one-shots (compaction's summarize call) log their envelope scalars (`compact/summary.{provider, model, maxTokens}`) and their input is deterministic code over the logged region — reconstructable from log + code, outside the invariant by the unfrozen-request marker. +**Model-visible ⟺ logged.** Anything that reaches a model request must be recorded in the session log. The checkable consequence: **every conversation request the loop sends is a pure function of the session log** — anyone holding the log reconstructs it byte-for-byte. Scope, stated precisely: the guarantee covers the loop-built `GenerateOptions`; provider wire bytes follow from it because both adapters' serialization is a pure per-message function at a pinned code version; direct one-shots (compaction's summarize call) log their envelope scalars (`compact/summary.{provider, model, maxTokens}`) and their input is deterministic code over the logged region — reconstructable from log + code, outside the invariant because only the loop marks request ownership. Prefix-cache stability is corollary #1, not the headline: an append-only log projected by a per-node pure function yields requests that are append-extensions of their predecessors whenever the header is unchanged — stability is emergent, not managed. Byte-exact audit/replay is corollary #2; resume and fork with *attributable* drift is corollary #3. @@ -26,7 +26,7 @@ Each step rebuilds prompt assembly. On the instance's first step, `agent/session **`step/start` is the reconstruction boundary.** A step derives messages from events before that sequence. Injection after the snapshot joins the next request, and reentrant appends are rejected during event publication. `agent/pre-step(agent, turn, step, signal)` remains the generic seam for content needed by the current request. Header reconstruction selects the step's `request/header`, or carries the prior snapshot when no new header is written. -**Enforcement.** In development, `dsh-invariants` independently rebuilds each loop request through a fresh `Session`, so the live cache cannot vouch for itself, then compares messages and folded header fields at `llm/stream`. Loop requests are identified by their frozen shape and session id; direct one-shots are excluded. Correctness depends on sequence-bounded reconstruction rather than listener order. A with-key e2e requires positive cache-read tokens after the first request; per-step usage is the production signal, and a header change or compaction appears as a cache-read drop on the next step. +**Enforcement.** The `dsh-agent-loop/invariant` companion registers with `ctx.invariants` and, when selected, independently rebuilds each loop request through a fresh `Session`, so the live cache cannot vouch for itself, then compares messages and folded header fields at `llm/stream`. The loop records the exact frozen request through `markAgentLoopRequest()` in `dsh-llm`; the process-local identity lets the companion and other request observers recognize conversation work, while direct one-shots remain excluded regardless of their frozen shape or session id. Correctness depends on sequence-bounded reconstruction rather than listener order. A with-key e2e requires positive cache-read tokens after the first request; per-step usage is the production signal, and a header change or compaction appears as a cache-read drop on the next step. ### The MiniCode shape: adopted, with the provenance arrow inverted diff --git a/.agents/notes/implemented/architecture/2026-07-05-windows-fs-permissions.md b/.agents/notes/implemented/architecture/2026-07-05-windows-fs-permissions.md new file mode 100644 index 0000000000..932b6ddbf4 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-05-windows-fs-permissions.md @@ -0,0 +1,31 @@ +# Agent Note: Windows write-permission semantics — inherited DACLs, not mode bits + +Status: implemented + +The replacement-file decision in this record is superseded by [Windows DACL preservation](../bug-fix/2026-07-19-windows-atomic-write-dacl-preservation.md). + +## Problem + +`writeFileAtomic` in `@deepseek-ai/dsh-fs-local` protects write-in-progress content with POSIX mode bits: the staging directory is created `0o700`, the temp file is opened `0o600`, and new files default to `0o600`. On POSIX this keeps temporary content owner-only regardless of the parent directory's permissions. + +Windows has no working equivalent behind the same API. Node's `chmod` there drives only the read-only attribute (every mode this package passes carries owner-write, so the calls are benign no-ops), and `stat().mode` reports synthetic `0o666`/`0o444` bits. The real security state is the file's DACL: a newly created file or directory inherits from its parent, while replacement needs the explicit handling owned by the superseding Agent Note. + +## Decision + +New Windows files use directory inheritance rather than synthetic mode bits: the staging directory is created inside the target's parent directory (`dirname(absolutePath)`), so it and the temp file inherit the destination directory's DACL. Replacement files follow the stricter [DACL preservation contract](../bug-fix/2026-07-19-windows-atomic-write-dacl-preservation.md). + +Tests assert mode bits on POSIX only. Native Windows coverage pins the package-owned replacement behavior; new-file inheritance remains an operating-system contract rather than a machine-specific ACL allowlist. + +## Alternatives considered + +**Explicit owner-only DACLs for new files.** Rejected because they would break inheritance and surprise users whose project directories are deliberately shared. Replacement writes copy the target's existing DACL rather than inventing an owner-only policy. + +**Test-side ACL verification.** A `Get-Acl` SID allowlist or `icacls` would verify Windows inheritance and the machine's `%TEMP%` ACL rather than package behavior; `icacls` also localizes well-known account names, making parsing locale-fragile. + +**Skip `chmod` on Windows.** Platform-guarding benign no-op calls adds branches without changing behavior. + +## Consequences + +POSIX keeps owner-only temp content regardless of the parent directory. A new Windows target inside a broadly accessible directory inherits that accessibility by design; a replacement retains the target's narrower DACL when one exists. + +Mode preservation across a replace degenerates to a no-op on Windows: a writable file probes as `0o666`, and replaying that through `chmod` leaves the read-only attribute clear. A read-only target cannot be replaced there because publication fails before the synthetic mode would matter. diff --git a/.agents/notes/implemented/architecture/2026-07-05-windows-jsonl-durable-publish.md b/.agents/notes/implemented/architecture/2026-07-05-windows-jsonl-durable-publish.md new file mode 100644 index 0000000000..60c3b9b627 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-05-windows-jsonl-durable-publish.md @@ -0,0 +1,33 @@ +# Agent Note: Windows-native durable JSONL publication + +Status: implemented + +## Problem + +`dsh-session-persistence-jsonl` publishes a session log lazily on the first append. The POSIX protocol writes a temp file, fsyncs it, links it to the final name, fsyncs the parent directory, and then removes the temp link. The parent-directory fsync is part of the durability contract: a crash after the namespace change must not lose the committed final name while leaving callers believing the session log materialized. + +Windows has atomic namespace operations, but Node does not expose a POSIX-equivalent parent-directory fsync contract there. Treating Windows directory sync failures as success would silently weaken a durable backend. The Windows path therefore needs a different publication primitive rather than a conditional inside the POSIX `syncDir` helper. + +## Decision + +The JSONL backend forks inside `materialize()` before any namespace mutation. Shared code computes the session directory, final log path, and encoded header plus initial event batch; POSIX and Windows then run separate publication protocols. + +POSIX keeps the existing protocol: create the root and cwd bucket with parent directory fsyncs, write and fsync a temp file, publish with `link()` so an existing final log is never overwritten, fsync the bucket directory, then remove the redundant temp hard link. + +Windows creates missing directories through a durable staging publish: create a random sibling directory, then publish it to the final directory name with `MoveFileExW(..., MOVEFILE_WRITE_THROUGH)` without `MOVEFILE_REPLACE_EXISTING` or `MOVEFILE_COPY_ALLOWED`. File materialization writes and fsyncs the temp log, then publishes that temp file to the final path with the same write-through `MoveFileExW` call and no replacement. `koffi` is the minimal Win32 bridge for this API surface; its install script is allowed in `pnpm-workspace.yaml` because the package ships the native loader and prebuilt platform modules. + +## Alternatives considered + +**Ignore Windows directory-sync failures.** Rejected because it reports a first append as durable without forcing the published namespace entry to stable storage. + +**Use `CreateHardLinkW`.** Rejected because hard links are filesystem-dependent, do not publish directories, and expose no write-through option. + +**Use replacement or transactional APIs.** `ReplaceFileW` has replacement semantics that conflict with same-id collision rejection, and Transactional NTFS is not recommended for new application designs. + +## Consequences + +The backend keeps one external contract across platforms: first append either publishes a complete log at the final name or fails without overwriting an existing log. The platform split is an implementation detail; `SessionPersistence` APIs and the logical JSONL record format do not change. The later [Zstandard encoding decision](2026-07-19-zstandard-jsonl-session-logs.md) applies before either platform publishes the opaque bytes. + +Windows tests exercise the real Win32 publish path on native Windows. Power-loss behavior remains an API-contract property rather than something unit tests can prove; the testable invariants are that directory fsync is not called on Windows materialization, final-path collisions fail, temp logs are fsync'd before publication, and the resulting log loads normally. + +Append and repair still use ordinary file-handle fsyncs on both platforms. A failed append closes its append-only handle, reopens the log read/write, truncates it to the pre-append size, and fsyncs the rollback because Windows rejects `ftruncate` on append-only handles. diff --git a/.agents/notes/implemented/architecture/2026-07-06-timeout-deadline-library.md b/.agents/notes/implemented/architecture/2026-07-06-timeout-deadline-library.md index 335581ecff..1c407d777d 100644 --- a/.agents/notes/implemented/architecture/2026-07-06-timeout-deadline-library.md +++ b/.agents/notes/implemented/architecture/2026-07-06-timeout-deadline-library.md @@ -18,7 +18,7 @@ Each new external-process or network tool re-derived the same four things — cl ### The library surface -Three functions plus one reason type: +Four functions, one watchdog interface, and one reason type: ```ts ignore-check /** The internal reason attached to a timeout abort, so consumers can classify it after the fact. */ @@ -51,19 +51,34 @@ export function deadline( code: string, ): { signal: AbortSignal; [Symbol.dispose](): void } +/** A stable signal plus one-at-a-time, timer-guarded async-iterator demand. */ +export interface IdleWatchdog { + readonly signal: AbortSignal + next(iterator: AsyncIterator): Promise> + [Symbol.dispose](): void +} + +/** Arm only while one iterator `next()` is outstanding, then rearm on later demand. */ +export function idleWatchdog( + upstream: AbortSignal | undefined, + timeoutMs: number, + code: string, +): IdleWatchdog + /** Recover the TimeoutReason from an aborted signal (or error); `code` scopes the match to this deadline's timer. */ export function timeoutOf(x: AbortSignal | { reason?: unknown }, code?: string): TimeoutReason | undefined ``` -`deadline` fuses an upstream signal with a timer through `AbortSignal.any`, adds a typed `TimeoutReason`, and exposes disposable timer cleanup. Non-positive timeouts are an internal no-timeout sentinel for backend-owned background work; external hints pass through `clampTimeout` and must be positive and finite. Without a timer or upstream signal, the function returns a never-aborting signal with the same disposal shape. Providers translate timeout reasons into seam-specific results. `timeoutOf(signal, code)` scopes classification so an outer nested deadline is treated as upstream cancellation rather than the inner capability's timeout. +`deadline` fuses an upstream signal with a one-shot timer through `AbortSignal.any`, adds a typed `TimeoutReason`, and exposes disposable timer cleanup. Non-positive timeouts are an internal no-timeout sentinel for backend-owned background work; external hints pass through `clampTimeout` and must be positive and finite. Without a timer or upstream signal, the function returns a never-aborting signal with the same disposal shape. `idleWatchdog` instead requires a positive finite interval, keeps one stable fused signal for the entire stream, and arms its timer only while one iterator `next()` is outstanding; resolution disarms it, later demand rearms it, concurrent demand fails, and disposal clears the active arm. Providers translate timeout reasons into seam-specific results. `timeoutOf(signal, code)` scopes classification so an outer nested deadline is treated as upstream cancellation rather than the inner capability's timeout. ### The division of labor | Concern | Owner | |---|---| | Validate request hint and clamp default/max | `dsh-timeout` (`clampTimeout`) — pure arithmetic plus the shared positive-finite request contract | -| Arm timer, abort on deadline, carry reason, fuse with upstream cancel | `dsh-timeout` (`deadline`) | -| Clear the timer | `dsh-timeout` (`[Symbol.dispose]`) | +| Arm one-shot timer, abort on deadline, carry reason, fuse with upstream cancel | `dsh-timeout` (`deadline`) | +| Arm and rearm only around outstanding iterator demand | `dsh-timeout` (`idleWatchdog`) | +| Clear the timer | `dsh-timeout` (`[Symbol.dispose]` on either primitive) | | Classify the first abort reason after abort | `dsh-timeout` (`timeoutOf`) | | **Actually terminate the work** | the capability's implementation | | The default/max *values* | the capability's config | @@ -75,6 +90,7 @@ The signal only *notifies*; termination is always the listener's job, and the li - **web_fetch** — the tool stays validate-and-forward; the provider's hand-rolled controller + `setTimeout` + manual listener + `finally` + `signal.reason` recovery is replaced by provider-owned `deadline`/`timeoutOf`. A pre-aborted upstream signal still throws `WEB_ABORTED` up front; otherwise `fetch` runs against the fused `d.signal`, and `translateAbortOrNetwork` classifies a thrown error by the signal (`timeoutOf` → `WEB_FETCH_TIMEOUT`, else aborted → `WEB_ABORTED`, else network → `WEB_PROVIDER_ERROR`). The public error-code contract is unchanged, and `TimeoutReason` never crosses the web seam as the public error. - **bash** — `resolve()` clamps the request into an explicit spec. Foreground `run()` creates the deadline and passes its signal to process execution, whose existing abort listener performs the process-group kill. The executor classifies the first abort as timeout or cancellation. Background starts remain timeout-free and forward only upstream cancellation. +- **LLM adapters** — `dsh-llm-deepseek` and `dsh-llm-pi-ai` wrap actual transport iteration with `idleWatchdog`. The five-minute configured interval covers only outstanding provider demand, not time the downstream consumer spends between chunks. The stable signal reaches `fetch` or the SDK for the whole call, so timeout closes the underlying request and maps to `TIMEOUT`, while an earlier caller abort maps to `ABORTED`. ## Consequences @@ -82,6 +98,7 @@ The signal only *notifies*; termination is always the listener's job, and the li - `SpawnSpec.timeoutMs` and `SpawnOutcome.timedOut`/`aborted` were removed rather than kept as always-zero/always-false vestiges: with `runBash` owning no timer and the executor owning classification, they were read nowhere. This is the one deviation from the literal proposal shape (which passed `timeoutMs: 0` into `runBash`); an always-0 field read by nothing is dead weight under the per-file coverage gate. - web_fetch shed its bespoke controller/timer/listener/reason-recovery; the classifier now keys off the deadline signal (`timeoutOf` + `aborted`) rather than the thrown error's shape, which is robust across both the request-phase reject-with-reason and the read-phase bare-`AbortError`. - `AbortSignal.any` and `using`/`Symbol.dispose` enter the repo for the first time here (Node ≥ 24 baseline, already met). +- Model streams now share one rearmable timer contract without turning a sliding idle interval into a total-call deadline or charging consumer think time. The primitive still only notifies; adapter tests prove their transports observe its stable signal and terminate. Out of scope, named to mark the boundary: `web_search` can gain an optional model-facing `timeout_ms` once its tool-schema/snapshot coverage is planned; future ripgrep-backed fs discovery tools can consume the same provider-owned deadline shape once they exist; a `tools/execute` waterfall middleware could arm a default deadline for every tool call by driving `exec.signal` — that would be a plugin that *consumes* this library and still only notifies, the hard kill remaining each capability's job. diff --git a/.agents/notes/implemented/architecture/2026-07-07-tool-call-timeout-policy.md b/.agents/notes/implemented/architecture/2026-07-07-tool-call-timeout-policy.md index 040a6c2fdc..82da8347a6 100644 --- a/.agents/notes/implemented/architecture/2026-07-07-tool-call-timeout-policy.md +++ b/.agents/notes/implemented/architecture/2026-07-07-tool-call-timeout-policy.md @@ -50,9 +50,9 @@ The plugin is `@deepseek-ai/dsh-timeout-policy`, a zero-config function/namespac searchTimeoutMs: 30000 ``` -Timeouts live on tool definitions rather than a free-text name map, eliminating misspelled unused policy. `defineTool` validates a positive finite budget. During dispatch the enforcer derives a deadline signal, restores the caller signal afterward, and converts its own expiry into `TOOL_TIMEOUT`; tools without a budget pass through unchanged. +Timeouts live on tool definitions rather than a free-text name map, eliminating misspelled unused policy. `defineTool` validates a positive finite budget. During dispatch the enforcer derives a deadline signal and assigns it to `exec.signal`; the registry fuses that deadline with the original caller signal before the body under the [tool-cancellation contract](2026-07-19-cooperative-tool-cancellation.md). The enforcer restores the caller signal afterward and converts its own expiry into `TOOL_TIMEOUT`; tools without a budget pass through unchanged. -Signal replacement is by **in-place mutation of `exec.signal`**, not by passing a new object to `next()`. Cordis's waterfall `next()` ignores any arguments handed to it and re-invokes downstream listeners with the shared payload array (`vendor/cordis/src/events.ts`), so the documented cordis idiom — mutate the shared object, then delegate — is the only mechanism that reaches dispatch. The plugin restores `exec.signal` to the caller's original in a `finally` so `tools/post-execute` never sees this plugin's (possibly already-aborted) deadline signal. +Signal replacement is by **in-place mutation of `exec.signal`**, not by passing a new object to `next()`. Cordis's waterfall `next()` ignores any arguments handed to it and re-invokes downstream listeners with the shared payload array (`vendor/cordis/src/events.ts`), so mutation is how the wrapper supplies its deadline to the registry. The registry re-fuses the captured caller signal immediately before the body, and the plugin restores `exec.signal` to the caller's original in a `finally` so `tools/post-execute` never sees the plugin's deadline signal. `timeout-policy` owns both uses of the `TOOL_TIMEOUT` code: the internal deadline code passed to `deadline()`/`timeoutOf()` (scoped so a nested outer deadline reads as an ordinary cancel) and the structured tool-result error code. Its replacement result is: @@ -104,6 +104,6 @@ A future model-facing grep/glob tool can be implemented on top of `ctx.bash` wit - `@deepseek-ai/dsh-tools` gains an around-dispatch surface after the interception seams deliberately split pre/post tool hooks. Its contract is narrow — wrap registry dispatch, not replace the pre-gate or post-result policy — and the base `next()` is dispatch-with-normalization so a wrapper never sees a raw tool throw. - Multiple `tools/execute` listeners compose by ordinary Cordis waterfall order: a listener that calls `next()` wraps downstream listeners plus dispatch; one that returns without `next()` short-circuits them. A deployment combining timeout with a future retry/sandbox/metrics wrapper chooses semantics by registration order ("timeout covers the whole retry" vs "timeout covers each attempt"). -- Opt-in by declaration is a deliberate misconfiguration risk: a tool can declare a `timeoutMs` without honoring `exec.signal`, and that tool will not stop on timeout. The plugin contract states that declaring a budget means cooperative; the web tools prove the pattern on tools that already forward the signal. +- Opt-in by declaration is a deliberate misconfiguration risk: a tool can declare a `timeoutMs` without honoring `exec.signal`, and that tool will not stop on timeout. The registry awaits that non-quiescent body rather than racing it, while the plugin contract states that declaring a budget means cooperative; the web tools prove the pattern on tools that already forward the signal. - During the transition `bash` and the migrated web tools use different timeout paths on purpose: `TOOL_TIMEOUT` is the model-facing tool-call budget, while `BASH_TIMEOUT` remains the bash backend timeout used by bash and hooks. - Deviation from the literal proposal, recorded per the implemented-Agent Note rule: the plugin package is `@deepseek-ai/dsh-timeout-policy` (not `tool-timeout`), signal replacement is in-place `exec.signal` mutation before `next()` (not `next({ ...exec, signal })`, which cordis ignores), and the per-tool budget is declared on the `ToolDefinition` (`timeoutMs`, set by the owning tool plugin from its config) rather than mapped by tool name in this plugin's config — so the enforcer is zero-config and a mistyped tool name is impossible. All three are described in `## Decision` above. diff --git a/.agents/notes/implemented/architecture/2026-07-08-tool-output-spill-files.md b/.agents/notes/implemented/architecture/2026-07-08-tool-output-spill-files.md index 1255dc7328..a9197de179 100644 --- a/.agents/notes/implemented/architecture/2026-07-08-tool-output-spill-files.md +++ b/.agents/notes/implemented/architecture/2026-07-08-tool-output-spill-files.md @@ -162,7 +162,7 @@ Those cases can consume `ctx.spillStore` directly in later work. They are not pa - `dsh-spill-local` unit tests cover `saveText`, `encodeSegment` sanitization (separators/tilde/whole-segment dots/empty), the session-hash directory, owner-only permissions, distinct paths per save, the configured/private root, and a storage-failure rejection. - `dsh-spill-policy` unit tests drive real tools through `ctx.tools.execute`: disabled-mode no-op, oversized-text replacement, small/non-text passthrough, `read` skip, best-effort fallback (save failure / no backend / no owner), and downstream-composition (bounding a replaced result, preserving `additionalContexts`). - `dsh-tool-web` integration drives `web_fetch` through `ctx.tools.execute` with the real `spill-local` backend + policy, proving the model-facing text changes only by the deliberate spill notice while the spill file holds the full formatted result. -- The `repl-agent` example loads `spill-local` + `spill-policy`, so its keyless Loader smoke exercises the real load path (the namespace-plugin export shape + `inject`). +- The `tui-agent` example loads `spill-local` + `spill-policy`, so its keyless Loader/PTY smoke exercises the real load path (the namespace-plugin export shape + `inject`). ## Consequences diff --git a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.i18n.yaml index a6344d276f..b25f335819 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.i18n.yaml +++ b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.i18n.yaml @@ -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 -2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md: f1a1868cd00007fb24efb21779dcc94c098b54e2 -2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md: a4993de2830301610bb2a9b0d28e8bbdf0ed9c46 +2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md: b934f7fd7087006be4f7eb3659e44e78b8ede367 +2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md: 3b5b60a95bef0695a446cdd3d45d299550f449f6 diff --git a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md index f1a1868cd0..b934f7fd70 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md +++ b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.md @@ -16,9 +16,9 @@ Successful calls are not the only pressure signal. A provider can reject a reque `agent/pre-step` is narrowed to `(agent, turn, step, signal)`. It remains a generic serial checkpoint before `step/start`, but it carries no compaction-only prompt or prefix fields. -The loop fires awaited serial `agent/post-step(agent, turn, step, signal)` after assistant output, every dispatched or synthetic tool result, post-tool context, and steering are durable, but before `step/end`. This placement gives pressure policy the complete successful-call state without splitting an assistant tool call from its result. A listener failure is an ordinary turn failure; it never enters model-request recovery. +The loop fires awaited serial `agent/post-step(agent, turn, step, signal)` after assistant output, every dispatched or synthetic tool result, post-tool context, and steering are durable, but before `step/end`. This placement gives pressure policy the complete successful-call state without splitting an assistant tool call from its result. A propagated listener failure is an ordinary turn failure; it never enters model-request recovery. Compact-basic contains its expected operational failures as described below. -`dsh-compact-basic` reads the exact latest routed model from the durable request header only to establish that a completed route exists, then asks the singleton `ctx.tokenMeter` to measure the canonical logged envelope and current surface. It does not fall back to `AgentOptions.model` for automatic pressure. A headerless session has no completed routed request to assess and produces no work; any durable non-empty model name uses the same estimator. Operational measurement or summarization failures warn and continue with full history. +`dsh-compact-basic` reads the exact latest routed model from the durable request header only to establish that a completed route exists, then asks the singleton `ctx.tokenMeter` to measure the canonical logged envelope and current surface. It does not fall back to `AgentOptions.model` for automatic pressure. A headerless session has no completed routed request to assess and produces no work; any durable non-empty model name uses the same estimator. Operational measurement or summarization failures warn and continue from the latest durable surface: full history before any replacement, or the pruned surface if pruning already landed. ### Request recovery is limited to the final model boundary @@ -32,17 +32,17 @@ If cancellation lands after assistant tool calls are durable but before all call `CompactService.compactIfNeeded(agent, trigger, signal)` accepts `trigger: 'pressure' | 'context-overflow'`. The interface gains no estimation methods or token types; `ctx.tokenMeter` remains the reusable accounting owner. -For `pressure`, compact-basic applies the service-wide threshold and retained-tail policy to one unified `ctx.tokenMeter.measure()` result. The same singleton meter owns range pricing, provenance, shadowed token counts, and non-shrinking-summary rejection. The common defaults remain threshold ratio `0.8`, retained history `floor(contextWindow × 0.16)`, summarization provider/model `''`, `maxTokens: 8192`, `compactionRetries: 1`, and `auto: true`. +For `pressure`, compact-basic resolves the durable provider/model target's adapter-owned capacity and exact-target policy, then applies the resulting threshold and retained-tail budgets to one unified `ctx.tokenMeter.measure()` result. Below pressure it returns without pruning. Once pressure qualifies, optional `ctx.toolResultPrune` rewrites oversized current results and compact-basic remeasures through the same meter; safe pressure skips the model call, while remaining pressure selects and summarizes from the pruned surface. The same singleton meter owns range pricing, provenance, shadowed token counts, and non-shrinking-summary rejection. Common defaults remain threshold ratio `0.8`, retained-history ratio `0.16`, summarization provider/model `''`, `maxTokens: 8192`, `compactionRetries: 1`, and `auto: true`; optional `modelPolicies` entries override them for an exact provider/model pair. -For canonical overflow, compact-basic bypasses scalar pressure and the normal retained-token budget. It chooses the maximal tool-balanced head range while leaving the newest indivisible unit, then attempts exactly one shrinking compaction under the same signal. The automatic listener snapshots `session.surface.replaceGeneration` and returns `{ action: 'retry' }` only when compaction succeeds and the generation increases. A backend returning a result without replacement cannot authorize retry. +For canonical overflow, compact-basic requires no capacity metadata and bypasses scalar pressure and the normal retained-token budget. It prunes first, then chooses the maximal tool-balanced head range while leaving the newest indivisible unit and attempts one shrinking summary compaction under the same signal when a range exists. The automatic listener snapshots `session.surface.replaceGeneration` and returns `{ action: 'retry' }` whenever pruning or summarization increases it. This remains true when pruning lands before later summary work throws; cancellation still wins. A backend returning a result without replacement cannot authorize retry, while pruning-only progress can authorize a retry without a `CompactionResult`. -`maxOverflowRetries` is optional and defaults to `1`; `0` disables overflow recovery without disabling pressure. `auto: false` registers neither automatic listener. Noncanonical errors, exhausted attempts, an already-aborted signal, a missing routed model, no safe range, no generation change, and recovery throws all delegate to the next listener. With no later recovery, the loop reports the original provider error object and code. Cancellation or disposal remains authoritative even if recovery work completes concurrently. +`maxOverflowRetries` is optional and defaults to `1`; `0` disables overflow recovery without disabling pressure. `auto: false` registers neither automatic listener. Noncanonical errors, exhausted attempts, an already-aborted signal, a missing routed model, no safe range, no generation change, and recovery throws before any replacement all delegate to the next listener. With no later recovery, the loop reports the original provider error object and code. A recovery throw after generation advances authorizes retry from durable progress; cancellation or disposal remains authoritative even if recovery work completes concurrently. The default summarizer resolves explicit configuration, then the latest logged route, then agent options. Because direct `llm/stream` middleware may reroute that auxiliary call, `compact/summary.{provider, model}` records the final mutable `GenerateOptions` target observed after dispatch rather than the pre-waterfall candidate. ## Testing -Unit tests cover final-adapter failure provenance and identity, closed-step retry numbering and reset, cancellation and disposal, post-step ordering, routed-envelope pressure, balanced overflow reduction, generation proof, caps, delegation, and auxiliary-call routing. Real-loop tests cover thrown and in-band overflow through compaction to a reconstructed retry request. +Unit tests cover final-adapter failure provenance and identity, closed-step retry numbering and reset, cancellation and disposal, post-step ordering, routed-envelope pressure, pressure-gated pruning, pruning-only relief, pruned-input summarization, balanced overflow reduction, durable prune progress before later failure, generation proof, caps, delegation, and auxiliary-call routing. Real-loop tests cover thrown and in-band overflow through pruning or summary compaction to a reconstructed retry request. ## Alternatives considered @@ -54,8 +54,8 @@ Unit tests cover final-adapter failure provenance and identity, closed-step retr ## Consequences -Post-step pressure describes the completed routed request, including durable tool results and request-only prefix fields. Canonical overflow supplies the backstop when no successful usage anchor exists. Recovery is bounded, cancellation-owned, and monotonic: it retries only after a visible surface generation change. +Post-step pressure describes the completed routed request, including durable tool results and request-only prefix fields. Optional model-free pruning removes predictable tool-output bulk before summary selection and can independently create retry-worthy progress. Canonical overflow supplies the backstop when no successful usage anchor exists. Recovery is bounded, cancellation-owned, and monotonic: it retries only after a visible surface generation change. -The cost is one additional serial checkpoint on successful steps and adapter-maintained overflow classification. Provider wording and heuristic character density remain maintenance risks. Surface compaction still cannot repair an envelope that alone exceeds the window or split one indivisible oversized message/tool unit. +The cost is one additional serial checkpoint on successful steps and adapter-maintained overflow classification. Provider wording and heuristic character density remain maintenance risks. Surface compaction still cannot repair an envelope that alone exceeds the window, split an indivisible non-tool node, or repair a tool unit whose non-prunable remainder remains oversized. The optional pruner can repair an otherwise indivisible tool pair when removable text-bearing tool-result content is the bulk. This Agent Note supersedes only the pre-step automatic-trigger portion of the [compaction capability-seam Agent Note](../feature/2026-06-18-compaction-capability-seam.md). The service split, standalone token meter, balanced range contract, log-recorded lock, summary replacement, and sole `summarize()` subclass hook remain unchanged. diff --git a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md index a4993de283..3b5b60a95b 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md +++ b/.agents/notes/implemented/architecture/2026-07-10-after-call-compaction-pressure-and-overflow-recovery.zh.md @@ -16,9 +16,9 @@ Status: implemented `agent/pre-step` 收窄为 `(agent, turn, step, signal)`。它仍是 `step/start` 之前的通用串行检查点，但不再携带压缩专用的提示词或前缀字段。 -循环在 assistant 输出、所有已分发或合成的工具结果、工具后上下文与 steering 都持久化之后、`step/end` 之前，触发等待式串行 `agent/post-step(agent, turn, step, signal)`。该位置让压力策略看到完整的成功调用状态，同时不会拆开 assistant 工具调用与其结果。监听器失败属于普通 turn 失败，绝不会进入模型请求恢复。 +循环在 assistant 输出、所有已分发或合成的工具结果、工具后上下文与 steering 都持久化之后、`step/end` 之前，触发等待式串行 `agent/post-step(agent, turn, step, signal)`。该位置让压力策略看到完整的成功调用状态，同时不会拆开 assistant 工具调用与其结果。向外传播的监听器失败属于普通 turn 失败，绝不会进入模型请求恢复；compact-basic 会按下文所述在内部处理其预期的操作性失败。 -`dsh-compact-basic` 从持久请求头读取精确的最新实际路由模型，只用它确认已经存在完整路由，随后让单例 `ctx.tokenMeter` 计量规范日志信封与当前表层。自动压力不会回退到 `AgentOptions.model`。没有请求头的会话尚无已完成路由请求可供判断，因此不执行工作；任意持久记录的非空模型名都使用同一个估算器。操作性的计量或摘要失败会发出警告，并继续使用完整历史。 +`dsh-compact-basic` 从持久请求头读取精确的最新实际路由模型，只用它确认已经存在完整路由，随后让单例 `ctx.tokenMeter` 计量规范日志信封与当前表层。自动压力不会回退到 `AgentOptions.model`。没有请求头的会话尚无已完成路由请求可供判断，因此不执行工作；任意持久记录的非空模型名都使用同一个估算器。操作性的计量或摘要失败会发出警告，并从最新持久表层继续：任何替换发生前使用完整历史；若剪枝已经落盘，则使用已剪枝表层。 ### 请求恢复只覆盖最终模型边界 @@ -32,17 +32,17 @@ Status: implemented `CompactService.compactIfNeeded(agent, trigger, signal)` 接收 `trigger: 'pressure' | 'context-overflow'`。接口不增加估算方法或 token 类型；`ctx.tokenMeter` 继续作为可复用的核算所有者。 -对于 `pressure`，compact-basic 把服务级阈值与保留尾部策略应用到一次统一的 `ctx.tokenMeter.measure()` 结果。范围定价、来源、被遮蔽 token 数与非缩小摘要拒绝也由同一个单例 meter 完成。通用默认值保持为阈值比例 `0.8`、保留历史 `floor(contextWindow × 0.16)`、摘要提供方/模型 `''`、`maxTokens: 8192`、`compactionRetries: 1` 与 `auto: true`。 +对于 `pressure`，compact-basic 先解析持久提供方/模型目标的适配器所属容量与精确目标策略，再把得到的阈值与保留尾部预算应用到一次统一的 `ctx.tokenMeter.measure()` 结果。低于压力时直接返回，不执行剪枝。压力达到条件后，可选的 `ctx.toolResultPrune` 会改写当前表层中过大的工具结果，compact-basic 再通过同一个 meter 重新计量；若压力恢复安全则跳过模型调用，否则从已剪枝表层选择范围并生成摘要。范围定价、来源、被遮蔽 token 数与非缩小摘要拒绝也由同一个单例 meter 完成。通用默认值保持为阈值比例 `0.8`、保留历史比例 `0.16`、摘要提供方/模型 `''`、`maxTokens: 8192`、`compactionRetries: 1` 与 `auto: true`；可选 `modelPolicies` 项可以按精确提供方/模型组合覆盖这些值。 -对于规范化溢出，compact-basic 绕过标量压力与普通保留 token 预算。它在保留最新不可分割单元的同时，选择最大的工具配对平衡头部范围，并在同一 signal 下只尝试一次缩小压缩。自动监听器先记录 `session.surface.replaceGeneration`，只有压缩成功且 generation 增加时才返回 `{ action: 'retry' }`。后端若只返回结果但没有替换表层，不能授权重试。 +对于规范化溢出，compact-basic 不要求容量元数据，并绕过标量压力与普通保留 token 预算。它先执行剪枝，再在保留最新不可分割单元的同时选择最大的工具配对平衡头部范围；存在范围时，才在同一 signal 下尝试一次缩小摘要压缩。自动监听器先记录 `session.surface.replaceGeneration`，剪枝或摘要让 generation 增加时就返回 `{ action: 'retry' }`。即使剪枝先落盘而后续摘要工作抛错，这条规则仍然成立；取消依然优先。后端若只返回结果但没有替换表层，不能授权重试；只有剪枝取得进展时，即使没有 `CompactionResult` 也可以授权重试。 -`maxOverflowRetries` 可选且默认为 `1`；`0` 只禁用溢出恢复，不会禁用压力检查。`auto: false` 不注册任何自动监听器。非规范化错误、尝试耗尽、已经中止的 signal、缺失路由模型、没有安全范围、generation 未变化，以及恢复抛错都会委托给下一个监听器。若没有后续恢复，循环报告原始提供方错误对象与代码。即使恢复工作并发完成，取消或销毁仍具有最终优先级。 +`maxOverflowRetries` 可选且默认为 `1`；`0` 只禁用溢出恢复，不会禁用压力检查。`auto: false` 不注册任何自动监听器。非规范化错误、尝试耗尽、已经中止的 signal、缺失路由模型、没有安全范围、generation 未变化，以及在任何替换之前恢复抛错，都会委托给下一个监听器。若没有后续恢复，循环报告原始提供方错误对象与代码。generation 增加后的恢复抛错会基于持久进展授权重试；即使恢复工作并发完成，取消或销毁仍具有最终优先级。 默认摘要器依次解析显式配置、最近记录的路由与 agent options。因为直接 `llm/stream` 中间件可以重新路由该辅助调用，`compact/summary.{provider, model}` 记录分发后最终可变的 `GenerateOptions` 目标，而不是 waterfall 之前的候选值。 ## 测试 -单元测试覆盖最终适配器失败的来源与身份、已关闭 step 的重试编号与重置、取消与销毁、post-step 顺序、已路由信封压力、平衡溢出缩减、generation 证明、上限、委托与辅助调用路由。真实循环测试覆盖抛出式和带内溢出，并验证压缩后的重试请求从替换表层重建。 +单元测试覆盖最终适配器失败的来源与身份、已关闭 step 的重试编号与重置、取消与销毁、post-step 顺序、已路由信封压力、压力门控剪枝、剪枝独立解除压力、从已剪枝输入生成摘要、平衡溢出缩减、后续失败前已落盘的剪枝进展、generation 证明、上限、委托与辅助调用路由。真实循环测试覆盖抛出式和带内溢出，并验证剪枝或摘要压缩后的重试请求从替换表层重建。 ## 考虑过的替代方案 @@ -54,8 +54,8 @@ Status: implemented ## 后果 -Post-step 压力描述已完成的路由请求，包括持久工具结果与仅请求前缀字段。当成功 usage 锚点不存在时，规范化溢出提供兜底路径。恢复有明确上限、以取消为准，并保持单调：只有模型可见的表层 generation 变化后才重试。 +Post-step 压力描述已完成的路由请求，包括持久工具结果与仅请求前缀字段。可选的无模型剪枝会在选择摘要前移除可预测的工具输出体积，也能独立产生足以重试的进展。当成功 usage 锚点不存在时，规范化溢出提供兜底路径。恢复有明确上限、以取消为准，并保持单调：只有模型可见的表层 generation 变化后才重试。 -代价是成功 step 增加一个串行检查点，并需要适配器持续维护溢出分类。提供方措辞与启发式字符密度仍是维护风险。表层压缩依然无法修复仅信封本身就超出窗口的情况，也不能拆分单个不可分割的超大消息或工具单元。 +代价是成功 step 增加一个串行检查点，并需要适配器持续维护溢出分类。提供方措辞与启发式字符密度仍是维护风险。表层压缩依然无法修复仅信封本身就超出窗口的情况，也不能拆分不可分割的非工具节点，或修复非可剪枝剩余部分仍然过大的工具单元。若可移除的文本工具结果是主要体积，可选剪枝器仍可修复原本不可分割的工具配对。 本 Agent Note 只取代[压缩能力接缝 Agent Note](../feature/2026-06-18-compaction-capability-seam.md) 中的 pre-step 自动触发部分。服务拆分、独立 token meter、平衡范围契约、日志记录锁、摘要替换与唯一 `summarize()` 子类 hook 均保持不变。 diff --git a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.i18n.yaml index 6db1ef4ece..b76ed1a9ae 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.i18n.yaml +++ b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.i18n.yaml @@ -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 -2026-07-10-single-file-executable-sdk-runtime-distribution.md: 0d4686a5a233785ca4832ef068a118b484a872fe -2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md: dcc9213c6b3a088b8b8bce2a442c5232ed5b7d0b +2026-07-10-single-file-executable-sdk-runtime-distribution.md: 43ba5708d1216c37a7ad7e2904df7d2a6baf016d +2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md: 3b33ff870d745584d2988bb6a7eb1a31e56ec3da diff --git a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md index 0d4686a5a2..43ba5708d1 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md +++ b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md @@ -36,7 +36,7 @@ Config discovery has two channels and fails loudly when both are missing: the `D Inside the exe's VFS sits a **real package tree in build-artifact form** (each package's `lib/` plus a real `node_modules`); the Loader resolves plugin names through standard dynamic `import()`: bare specifiers resolve upward along `node_modules` from the Loader's position inside the VFS, and land inside the VFS naturally. The closed set needs no allowlist code — the set is whatever the VFS has installed, and importing a name outside the set fails. -The deploy root is [`python/sdk-runtime/package.json`](../../../../python/sdk-runtime/package.json) (`dsh-jsonrpc-agent-pkg`, a pnpm workspace member and a zero-code pure dependency manifest) — the unified source of truth for "which plugins the exe ships" and "what the Python runtime distributes". Adding a plugin to the exe = adding one dependency line to the manifest and repackaging. [`scripts/verify-runtime-closure.ts`](../../../../scripts/verify-runtime-closure.ts) traverses every workspace package covered by that manifest and requires every non-optional workspace peer at the runtime root, reporting the complete referencing-package → missing-peer chain; CI static, pre-push, and the single-exe build run it before packaging. Deploy also packs by each package's `files`, so the shared chunks tsdown splits out must be covered by `files`. +The deploy root is [`python/sdk-runtime/package.json`](../../../../python/sdk-runtime/package.json) (`dsh-jsonrpc-agent-pkg`, a pnpm workspace member and a zero-code pure dependency manifest) — the unified source of truth for "which plugins the exe ships" and "what the Python runtime distributes". Adding a plugin to the exe = adding one dependency line to the manifest and repackaging. [`scripts/verify-runtime-closure.ts`](../../../../scripts/verify-runtime-closure.ts) traverses every workspace package covered by that manifest and requires every non-optional workspace peer at the runtime root, reporting the complete referencing-package → missing-peer chain; `pnpm run hygiene`, CI static, and the single-exe build run it before packaging. Deploy also packs by each package's `files`, so the shared chunks tsdown splits out must be covered by `files`. ### Build pipeline and artifacts diff --git a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md index dcc9213c6b..3b33ff870d 100644 --- a/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md +++ b/.agents/notes/implemented/architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.zh.md @@ -36,7 +36,7 @@ exe 使用 [@yao-pkg/pkg](https://github.com/yao-pkg/pkg)（vercel/pkg 归档后 exe 的 VFS 内是**构建产物形态的真实包树**（各包的 `lib/` + 真实 `node_modules`）。loader 通过标准动态 `import()` 解析插件名：裸包名从 VFS 内 loader 所在位置沿 `node_modules` 向上解析，自然落在 VFS 内。封闭集不需要白名单代码——VFS 中安装了什么，集合中就有什么；`import()` 集合外的名称会失败。 -部署根目录是 [`python/sdk-runtime/package.json`](../../../../python/sdk-runtime/package.json)（`dsh-jsonrpc-agent-pkg`，pnpm 工作区成员、零代码纯依赖清单），也是“exe 安装哪些插件”与“Python 运行时分发什么”的统一事实源。向 exe 添加插件，就是在清单中增加一行依赖后重新打包。[`scripts/verify-runtime-closure.ts`](../../../../scripts/verify-runtime-closure.ts) 遍历该清单覆盖的全部工作区包，要求每个非可选的工作区对等依赖（peer dependency）都显式列在运行时根目录，并报告“引用包 → 缺失对等依赖”的完整链路；CI 静态检查、pre-push 与 single-exe 构建都会在打包前运行该门禁。部署还会依据各包的 `files` 字段打包，因此 tsdown 拆出的共享分片必须被 `files` 覆盖。 +部署根目录是 [`python/sdk-runtime/package.json`](../../../../python/sdk-runtime/package.json)（`dsh-jsonrpc-agent-pkg`，pnpm 工作区成员、零代码纯依赖清单），也是“exe 安装哪些插件”与“Python 运行时分发什么”的统一事实源。向 exe 添加插件，就是在清单中增加一行依赖后重新打包。[`scripts/verify-runtime-closure.ts`](../../../../scripts/verify-runtime-closure.ts) 遍历该清单覆盖的全部工作区包，要求每个非可选的工作区对等依赖（peer dependency）都显式列在运行时根目录，并报告“引用包 → 缺失对等依赖”的完整链路；`pnpm run hygiene`、CI 静态检查与 single-exe 构建都会在打包前运行该门禁。部署还会依据各包的 `files` 字段打包，因此 tsdown 拆出的共享分片必须被 `files` 覆盖。 ### 构建管线与产物 diff --git a/.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md b/.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md index 2ddbe0a351..e0de18e90a 100644 --- a/.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md +++ b/.agents/notes/implemented/architecture/2026-07-12-agent-scope-runtime-design.md @@ -12,13 +12,13 @@ The implementation needs enough state to preserve real ownership and settlement ## Decision -The runtime uses one mechanism per independent fact. Scope routing has an opaque carrier; each live registry object has one entry record; each create or resume operation has one transaction; typed same-process calls borrow readonly values; real data boundaries materialize once; the cooperative prompt-assembly result is authoritative; and worker/process code retains separate terminal and quiescence state only where different owners can genuinely race. +The runtime uses one mechanism per independent fact. Scope routing has an opaque carrier and shared layer store; each live registry object has one entry record; each create or resume operation has one transaction; typed same-process calls borrow readonly values; real data boundaries materialize once; the cooperative prompt-assembly result is authoritative; and worker/process code retains separate terminal and quiescence state only where different owners can genuinely race. The design can be skimmed as seven choices: | Problem | Authoritative mechanism | |---|---| -| Select global plus one agent's registrations | Opaque scope key and routing carrier | +| Select global plus one agent's registrations | Opaque scope key, routing carrier, and shared layer store | | Own one live agent or session | One registry entry captured by its disposer | | Coordinate create/resume | One `AgentCreationTransaction` | | Protect durable, queued, model, or wire data | Materialize once at that boundary | @@ -68,11 +68,11 @@ A `ScopeKey` is an opaque object compared by identity. The harness uses the live The receiver is a small carrier rather than a transparent proxy for the domain object. Code that needs the agent receives the explicit event argument; code that needs registration ownership receives `agent.ctx`. -### Registry reads overlay one exact map +### Registry reads overlay one exact layer -Scope-aware registries store global contributions separately from identity-keyed local contributions. A read resolves the global layer and at most one local layer; it never traverses parentage. +Scope-aware registries use `ScopedLayers` to own one eager global aggregate and lazily created identity-keyed aggregates. A read resolves the global layer and at most one exact local layer; it never creates state or traverses parentage. Registration visibility and Cordis effect ownership derive from the same context, and reclamation waits until the concrete layer's complete aggregate is empty ([decision](2026-07-12-scoped-layers-store.md)). -Each service retains its domain rule. Named prompt values and tools use local shadowing, tool restrictions filter globals before local tools are added, and events select listener audiences rather than registered data. Scope supplies identity and ownership, not a universal merge algorithm. +Each service retains its domain rule. Named command and prompt views use the shared insertion-ordered shadow merge; tools keep a richer resolver because restrictions filter globals before local tools are added and the reserved Code Mode transport is inserted separately. Prompt variables and tool guards retain live iteration, while tool-provider membership is materialized per assembly. Scope supplies storage lifecycle and named shadowing, not a universal registry view. ### Fused dispatch helpers prevent subject drift @@ -322,7 +322,7 @@ TypeScript cannot govern JavaScript casts, direct Cordis dispatch, process messa ### Runtime invariants cover cross-service facts -The invariants plugin verifies that every declared scoped event uses a marked carrier and that event families exposing a subject use the matching key. Session trace validation stages before append commit and advances after the same event commits. +The `dsh-scope/invariant` companion verifies, when selected, that every declared scoped event uses a marked carrier and that event families exposing a subject use the matching key. The separate `dsh-session/invariant` contribution stages trace validation before append commit and advances after the same event commits; both register through `ctx.invariants`. The plugin does not police trusted setup by scanning registries or reject prompt assembly objects fabricated through casts. Those checks would turn composition contracts into speculative runtime machinery without protecting a real external boundary. diff --git a/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.i18n.yaml new file mode 100644 index 0000000000..b49506364b --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.i18n.yaml @@ -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 +2026-07-12-scoped-layers-store.md: b850b6bcbb22401b386b4458b6d5c65a160c85cd +2026-07-12-scoped-layers-store.zh.md: 8bfc0a0e8ec1e3de624ff8d9e48b7517833fc025 diff --git a/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.md b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.md new file mode 100644 index 0000000000..b850b6bcbb --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.md @@ -0,0 +1,126 @@ +# Agent Note: Shared scoped-layer storage + +Status: implemented + +English | [中文](2026-07-12-scoped-layers-store.zh.md) + +## Problem + +Agent scoping ([decision](2026-07-08-agent-scope-contexts.md), [runtime design](2026-07-12-agent-scope-runtime-design.md)) gives scope-aware registries the same recurring shape: one global registration layer plus one exact agent layer. Seven registration facades use that shape: `tools.register`, `tools.restrict`, and `tools.guard` in `dsh-tools`; `SystemPrompt.section`, `SystemPrompt.tools`, and `SystemPrompt.variable` in `dsh-system-prompt`; and `CommandService.register` in `dsh-commands`. + +Without a shared primitive, each facade repeats the lifecycle choreography around its domain state: derive visibility from the calling context, create a scoped container on demand, attach ownership to the same Cordis fiber, install undo before notifying observers, return Cordis's exact disposer, and reclaim empty scoped state. Separate maps and collection types also leave a service without one object representing a scope's complete contribution. + +The duplicated code carries three non-obvious requirements: + +- Visibility and ownership must come from the same context; accepting them separately permits a registration visible in one scope but disposed with another. +- Undo must be collected before a change callback runs, so a throwing callback rolls the mutation back. +- The public disposer must be the exact function returned by `ctx.effect()`; wrapping it breaks Cordis's identity-based ordered teardown. + +The shared part is lifecycle and insertion-ordered storage, not registry policy. Tool restrictions, reserved transport handling, prompt evaluation timing, command normalization, exact diagnostics, and callback containment remain different domain contracts. + +## Decision + +`@deepseek-ai/dsh-scope` provides a key-agnostic `store.ts` implementation module. The package continues to peer on Cordis and `@deepseek-ai/dsh-invariants`, and its invariant companion remains unchanged. The package root exports four storage symbols: `ScopeLayer`, `ScopedLayers`, `NamedEntries`, and `AnonymousEntries`. `EntryValues` remains internal, and `store.ts` is not a package subpath. + +`ScopeLayer` keeps the aggregate concept explicit while requiring only whole-layer emptiness. A service defines one concrete layer whose tables and domain helpers fit that service; `ScopedLayers` owns construction, selection, lifecycle attachment, notification, and aggregate reclamation. + +## Public interface + +```ts ignore-check +export interface ScopeLayer { + isEmpty(): boolean +} + +export class ScopedLayers { + constructor( + createLayer: (scope: ScopeKey | undefined) => L, + onChange: () => void, + ) + + readonly global: L + peek(scope: ScopeKey | undefined): L | undefined + + merge( + scope: ScopeKey | undefined, + pick: (layer: L) => NamedEntries, + ): Map + + effect( + ctx: Context, + action: (layer: L) => () => void, + options: { label: string; notify?: boolean }, + ): () => void +} + +export class NamedEntries { + constructor(duplicateError: (name: string) => Error) + insert(name: string, value: V): () => void + get(name: string): V | undefined + has(name: string): boolean + keys(): IterableIterator + entries(): IterableIterator<[string, V]> + values(): IterableIterator + isEmpty(): boolean +} + +export class AnonymousEntries { + append(value: V): () => void + values(): IterableIterator + isEmpty(): boolean +} +``` + +## Storage contract + +- The constructor creates `global` once with `createLayer(undefined)`. A scoped layer is created only by `effect()`; `peek()` and `merge()` never create one, and `peek(undefined)` returns `undefined` because the global layer is already explicit. +- `merge()` is the only materialized generic read. It copies named global entries in insertion order, then applies matching scoped entries in their insertion order so same-name entries shadow without moving unrelated names. +- `NamedEntries.insert()` checks and inserts atomically, returns an idempotent exact-entry undo, and obtains the registry's exact duplicate diagnostic from the caller-supplied factory. Lookup and iterators retain native `Map` order and stay live within one nonempty table generation; draining the table starts a new generation so an in-flight iterator cannot observe a self-replacement. +- `AnonymousEntries.append()` assigns a unique internal key per registration, so equal callbacks or values remain independent. Its iterator is insertion-ordered and uses the same live-generation boundary. +- `effect()` derives the key with `scopeOf(ctx)` and attaches the action to that same `ctx.effect()`. It accepts one synchronous action returning one synchronous undo; actions must either return their undo or throw before retaining a contribution. The helper does not normalize the wider Cordis `Effect` union. +- `effect()` collects the action's undo before calling `onChange` and returns the exact `ctx.effect()` disposer. Disposal runs the action undo before notification, is idempotent through Cordis, and removes a scoped layer only after its complete `ScopeLayer.isEmpty()` becomes true. +- `options.notify` defaults to `true`. The callback's own policy stays authoritative: tool and prompt change callbacks may throw and trigger registration rollback; `CommandService.notifyChange()` contains observer failures; tool guards pass `notify: false`. + +## Registry migrations + +`dsh-tools` defines one `ToolLayer` containing named tools plus anonymous compiled restrictions and guard registrations. `ToolRegistry` retains its private domain resolver for visible definitions, pre-restriction known names, restrictable global names, scoped shadowing, restrictions, and reserved `run_code` insertion. Guard evaluation live-iterates global then scoped registrations: additions to a nonempty generation can run in the current dispatch, while a self-replacement after draining the guard table begins with the next dispatch. + +`dsh-system-prompt` defines one `PromptLayer` containing named sections and variables plus anonymous tool providers. Assembly merges sections before evaluating them, so a shadowed provider is never called. Tool-provider membership is materialized once per assembly. Variable providers live-iterate global then scoped tables: additions to a nonempty generation can run in the current assembly, while a self-replacement after draining the variable table begins with the next assembly. + +`dsh-commands` defines a one-table layer containing `NamedEntries`. Effective views use `merge()`, while `CommandService` retains definition normalization and freezing, exact duplicate diagnostics, sorted immutable descriptors, direct execution, HMR cleanup, and independently contained `commands/change` observers. + +All seven facades keep validation and diagnostics in their owning registry and continue to return the exact Cordis disposer. The migration changes neither public registry behavior nor model-, human-, wire-, persistence-, or configuration-visible output. + +## Alternatives considered + +**Keep the independent implementations.** This avoids a new library interface but leaves lifecycle ordering, disposer identity, and scope reclamation duplicated across seven facades. + +**One helper per table.** This removes some local code but preserves multiple per-scope maps and cannot reclaim one scope's aggregate contribution correctly. + +**Per-scope registry instances.** Child registries would need delegation for global-plus-scoped views, special subtraction for restrictions, and observer discovery across instances. They would move complexity rather than remove it. + +**Explicit scope parameters on registration methods.** Separate visibility and ownership inputs make mismatched lifetimes representable, while an omitted scope silently becomes global. + +**Accept the complete Cordis `Effect` union.** None of the seven registrations has asynchronous setup, multiple undos, or an independent settlement boundary. General normalization would duplicate Cordis lifecycle machinery without a current consumer. + +**Expose `ScopedLayers.values()`, `ScopedLayers.keys()`, or a global-admission predicate.** Those operations encode consumer-specific live/materialized and filtering policies. Direct table iteration preserves explicit live semantics, `merge()` covers the shared named shadowing operation, and `ToolRegistry` keeps its richer private resolver. + +**Put `values()` on `ScopeLayer` or export `EntryValues`.** A layer aggregates heterogeneous tables and has no coherent value type or iteration policy. `EntryValues` is useful only to share implementation details between the two table classes; making it public would enlarge the interface without giving callers a meaningful layer-wide read. + +**Generate layers from a mapped-type table description.** Three-table and one-table concrete layers are short, inspectable, and free to hold domain helpers. A class generator would add a second construction model and generated runtime shape for little leverage. + +## Consequences + +- Scope-aware registries express one aggregate layer and reuse the same construction, ownership, rollback, notification, and reclamation choreography. Domain-specific validation, diagnostics, filtering, evaluation, and observer policy remain in each registry. +- The public read surface stays narrow: direct table iteration preserves explicitly live behavior, while `merge()` is the one shared materialized shadowing operation. A heterogeneous `ScopeLayer` has no layer-wide `values()` contract. +- The helper is deliberately synchronous. A future registration that needs asynchronous setup or several independently owned undos must identify its ownership and settlement boundaries before widening this contract. +- An action must throw before retaining a contribution or return an undo for everything it retained; the helper cannot repair mutation outside that contract. The provided entry operations are atomic, and migrated registries perform fallible validation before insertion. +- A scoped layer remains allocated until every table in its aggregate is empty. Disposing one facade therefore cannot discard sibling contributions owned by the same scope. +- The four public symbols become a reusable package contract. Keeping `EntryValues` internal and consumer policy outside the helper limits the compatibility surface. +- The migration changes no public registry behavior and no model-, human-, wire-, persistence-, configuration-, or dependency-graph output. + +## Verification + +- `dsh-scope` unit tests cover global construction, lazy scoped construction, non-creating reads, named merge order and shadowing, aggregate reclamation, factory and action failure cleanup, notification ordering and rollback, `notify: false`, effect labels, exact disposer identity, idempotent teardown, caller-owned duplicate errors, independent anonymous duplicates, live iterators, and drained-generation detachment. +- Focused tool, system-prompt, and command suites cover restrictions, reserved transport handling, known/restrictable-name agreement, guard re-entrancy and self-replacement, validation order, exact diagnostics, section shadow-before-evaluate, provider snapshot membership, variable re-entrancy and self-replacement, contained command observers, frozen and sorted views, direct execution, and lifecycle disposal. +- The scoped core-data type-equivalence check ties `ScopeLayer` documentation to its source declaration. Repository documentation, module-graph, build, hygiene, coverage, and built-artifact gates exercise the root export and package boundary. +- Existing ACP, headless, and TUI keyless snapshots remain the regression boundary for tool schemas, prompt assembly, and human commands. The implementation does not update any expected transcript. diff --git a/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.zh.md b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.zh.md new file mode 100644 index 0000000000..8bfc0a0e8e --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-12-scoped-layers-store.zh.md @@ -0,0 +1,126 @@ +# Agent Note: 共享作用域分层存储 + +Status: implemented + +[English](2026-07-12-scoped-layers-store.md) | 中文 + +## 问题 + +agent（智能体）作用域机制（[决策](2026-07-08-agent-scope-contexts.md)、[运行时设计](2026-07-12-agent-scope-runtime-design.md)）让支持作用域的注册表反复呈现同一种形态：一个全局注册层，加上一个与具体 agent 精确对应的层。七个注册门面都采用这一形态：`tools.register`、`tools.restrict` 和 `tools.guard`（位于 `dsh-tools`）；`SystemPrompt.section`、`SystemPrompt.tools` 和 `SystemPrompt.variable`（位于 `dsh-system-prompt`）；以及 `CommandService.register`（位于 `dsh-commands`）。 + +如果没有共享原语，每个门面都要围绕自己的领域状态重复相同的生命周期编排：从调用方上下文导出可见性，按需创建专属容器，把属主绑定到同一个 Cordis fiber，先装入 undo 再通知观察者，原样返回 Cordis 的 disposer，并回收空的专属状态。各自分离的映射与集合类型也会让服务缺少一个表示某个 scope 完整贡献的对象。 + +重复代码承载着三项不明显的要求： + +- 可见性与属主必须来自同一个上下文；若分开接受二者，就能登记出对一个 scope 可见、却随另一个 scope 销毁的贡献。 +- change 回调运行前必须收集 undo，抛错的回调才能回滚变更。 +- 公开 disposer 必须就是 `ctx.effect()` 返回的那个函数；包装它会破坏 Cordis 基于身份的有序拆除。 + +共享的是生命周期与保持插入顺序的存储，而不是注册表策略。工具限制、保留传输处理、提示词求值时机、命令规范化、精确诊断和回调异常隔离，仍分别属于不同的领域契约。 + +## 决策 + +`@deepseek-ai/dsh-scope` 提供与键类型无关的 `store.ts` 实现模块。该包（package）继续将 Cordis 和 `@deepseek-ai/dsh-invariants` 列为对等依赖（peer dependency），其不变量配套模块保持不变。包根导出四个存储符号：`ScopeLayer`、`ScopedLayers`、`NamedEntries` 和 `AnonymousEntries`。`EntryValues` 仍是内部接口，`store.ts` 不是包子路径。 + +`ScopeLayer` 保留显式的聚合概念，同时只要求判断整个层是否为空。服务定义一个具体层，使其表结构与领域 helper 适合该服务；`ScopedLayers` 负责构造、选择、生命周期挂接、通知和聚合回收。 + +## 公开接口 + +```ts ignore-check +export interface ScopeLayer { + isEmpty(): boolean +} + +export class ScopedLayers { + constructor( + createLayer: (scope: ScopeKey | undefined) => L, + onChange: () => void, + ) + + readonly global: L + peek(scope: ScopeKey | undefined): L | undefined + + merge( + scope: ScopeKey | undefined, + pick: (layer: L) => NamedEntries, + ): Map + + effect( + ctx: Context, + action: (layer: L) => () => void, + options: { label: string; notify?: boolean }, + ): () => void +} + +export class NamedEntries { + constructor(duplicateError: (name: string) => Error) + insert(name: string, value: V): () => void + get(name: string): V | undefined + has(name: string): boolean + keys(): IterableIterator + entries(): IterableIterator<[string, V]> + values(): IterableIterator + isEmpty(): boolean +} + +export class AnonymousEntries { + append(value: V): () => void + values(): IterableIterator + isEmpty(): boolean +} +``` + +## 存储契约 + +- 构造器只创建一次 `global`，调用的是 `createLayer(undefined)`。只有 `effect()` 会创建专属层；`peek()` 和 `merge()` 从不创建专属层，而 `peek(undefined)` 返回 `undefined`，因为全局层已经显式存在。 +- `merge()` 是唯一会物化结果的通用读取接口。它按插入顺序复制全局命名条目，再按专属条目的插入顺序应用这些条目；同名条目完成遮蔽，但不会移动无关名称。 +- `NamedEntries.insert()` 以原子方式检查并插入，返回幂等且只撤销该精确条目的 undo，并通过调用方提供的工厂取得所属注册表的精确重名诊断。查询与迭代器保留 `Map` 的原生顺序，并在同一个非空表 generation 内保持活遍历；清空表会开启新的 generation，因此尚未结束的迭代器无法观察到自我替换。 +- `AnonymousEntries.append()` 为每次登记分配唯一内部键，因此值相等的回调或其他值仍彼此独立。其迭代器保留插入顺序，并采用同样的 generation 活遍历边界。 +- `effect()` 通过 `scopeOf(ctx)` 导出键，并把 action 挂到同一个 `ctx.effect()` 上。它只接受一个同步 action，且该 action 只返回一个同步 undo；action 要么返回其 undo，要么必须在保留任何贡献之前抛错。helper 不会规范化更宽泛的 Cordis `Effect` union。 +- `effect()` 在调用 `onChange` 前收集 action 的 undo，并原样返回 `ctx.effect()` 的 disposer。销毁时先运行 action undo 再通知；Cordis 保证其幂等性；只有整个层的 `ScopeLayer.isEmpty()` 变为 true 后，helper 才删除专属层。 +- `options.notify` 默认为 `true`。回调自身的策略仍具最终效力：工具与提示词的 change 回调可以抛错并触发登记回滚；`CommandService.notifyChange()` 会隔离观察者失败；工具 guard 传入 `notify: false`。 + +## 注册表迁移 + +`dsh-tools` 定义一个 `ToolLayer`，其中包含命名工具以及匿名的已编译 restriction 和 guard 登记。`ToolRegistry` 保留其私有领域解析器，由它处理可见定义、限制前的已知名称、可限制的全局名称、专属遮蔽、restriction，以及保留的 `run_code` 插入。guard 求值会先活遍历全局登记，再活遍历专属登记：向非空 generation 新增的登记可以在当前分发中运行，而 guard 表清空后的自我替换则从下一次分发开始运行。 + +`dsh-system-prompt` 定义一个 `PromptLayer`，其中包含命名的段落与变量，以及匿名工具提供方。组装流程在求值前合并段落，因此被遮蔽的提供方不会被调用。每次组装只物化一次工具提供方成员集合。变量提供方会先活遍历全局表，再活遍历专属表：向非空 generation 新增的提供方可以在当前组装中运行，而变量表清空后的自我替换则从下一次组装开始运行。 + +`dsh-commands` 定义一个单表层，其中包含 `NamedEntries`。生效视图使用 `merge()`；`CommandService` 则保留对定义的规范化与冻结处理、精确重名诊断、经过排序的不可变描述符、直接执行、HMR（热模块替换）清理，以及对各个 `commands/change` 观察者分别隔离失败的行为。 + +七个门面都把校验与诊断留在所属注册表中，并继续返回 Cordis 的原始 disposer。迁移既不改变公开注册表行为，也不改变模型可见或人类可见的输出，以及协议、持久化或配置层面的可见输出。 + +## 备选方案 + +**保留彼此独立的实现。** 这样不必新增库接口，但七个门面仍会重复生命周期顺序、disposer 身份和 scope 回收。 + +**每张表一个 helper。** 这能减少一部分局部代码，但会保留多张按 scope 划分的映射，而且无法正确回收某个 scope 的聚合贡献。 + +**每 scope 一个注册表实例。** 子注册表需要通过委托获得全局加专属的视图，对 restriction 进行特殊的减法处理，并跨实例发现观察者。这只会转移复杂度，而不会消除复杂度。 + +**注册方法上的显式 scope 参数。** 分开的可见性与属主输入让不匹配的生命周期成为可表达状态，而遗漏 scope 则会静默变成全局登记。 + +**接受完整的 Cordis `Effect` union。** 七个登记口都没有异步 setup、多份 undo 或独立 settlement 边界。通用规范化会在没有现有消费者需要它时重复 Cordis 的生命周期 machinery。 + +**暴露 `ScopedLayers.values()`、`ScopedLayers.keys()` 或全局放行谓词。** 这些操作会编码消费方特有的活遍历或物化策略，以及过滤策略。直接遍历条目表可保留显式的活语义，`merge()` 覆盖共享的命名遮蔽操作，而 `ToolRegistry` 继续保有功能更丰富的私有解析器。 + +**把 `values()` 放在 `ScopeLayer` 上，或导出 `EntryValues`。** 一个层会聚合异构表，因而没有一致的值类型或迭代策略。`EntryValues` 只适合在两个表类之间共享实现细节；将其公开只会扩大接口，却不能为调用方提供有意义的整层读取方式。 + +**通过 mapped-type 表描述生成层。** 三表与单表具体层都很短、易于检查，并可自由持有领域 helper。类生成器会增加第二种构造模型和生成式运行时形状，收益却很小。 + +## 后果 + +- 支持作用域的注册表各自通过一个聚合层表达状态，并复用相同的构造、属主、回滚、通知和回收编排。各注册表仍各自保有领域特有的校验、诊断、过滤、求值和观察者策略。 +- 公开读取接口保持狭窄：直接遍历条目表可保留显式的活语义，`merge()` 是唯一共享的物化遮蔽操作。异构的 `ScopeLayer` 不具备整层 `values()` 契约。 +- helper 刻意保持同步。未来的登记若需要异步 setup 或多份分别拥有属主的 undo，必须先明确属主与 settlement 边界，再拓宽这项契约。 +- action 必须在保留贡献前抛错，或者为自己保留的一切返回 undo；helper 无法修复超出这项契约的变更。提供的条目操作是原子的，迁移后的注册表会在插入前执行可能失败的校验。 +- 专属层会一直保持已分配状态，直到其聚合内的所有表都为空。因此，销毁一个门面不会丢弃同一 scope 拥有的其他贡献。 +- 四个公开符号构成一项可复用的包契约。将 `EntryValues` 保持为内部接口，并把消费方策略留在 helper 之外，可以限制兼容性范围。 +- 迁移不改变任何公开注册表行为，也不改变模型、人类、协议、持久化、配置或依赖图层面的任何输出。 + +## 验证 + +- `dsh-scope` 单元测试覆盖全局构造、专属层延迟构造、非创建式读取、命名合并顺序与遮蔽、聚合回收、工厂与 action 失败清理、通知顺序与回滚、`notify: false`、effect 标签、原始 disposer 身份、幂等拆除、调用方提供的重名错误、相同匿名值的独立登记、活迭代器，以及表清空后的 generation 脱离。 +- 工具、系统提示词和命令专项测试套件覆盖 restriction、保留传输处理、已知名称与可限制名称的一致性、guard 重入与自我替换、校验顺序、精确诊断、section 先遮蔽再求值、提供方快照成员关系、variable 重入与自我替换、隔离失败的命令观察者、冻结且有序的视图、直接执行和生命周期销毁。 +- 作用域核心数据的类型等价性检查将 `ScopeLayer` 文档与其源声明绑定。仓库级的文档、模块图、构建、hygiene、覆盖率与构建产物门禁会覆盖包根导出与包边界。 +- 现有 ACP（Agent Client Protocol）、headless 和 TUI 无密钥快照继续作为工具 schema、提示词组装和人类命令的回归边界。实现不会更新任何预期 transcript（文本记录）。 diff --git a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.i18n.yaml index 3ed38c1282..3a6b18cecf 100644 --- a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.i18n.yaml +++ b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.i18n.yaml @@ -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 -2026-07-14-provider-routed-llm-adapters.md: b7944bd31fdb5f63894e867d7c1224215d694f11 -2026-07-14-provider-routed-llm-adapters.zh.md: 7dcadf2521bab079e328b5f0d0a45185778b3b8d +2026-07-14-provider-routed-llm-adapters.md: 98205d18d07752e0cdba86d7cba80368d45fd816 +2026-07-14-provider-routed-llm-adapters.zh.md: c35225a86baf4c2d09732b5940abbc8046d365fb diff --git a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.md b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.md index b7944bd31f..98205d18d0 100644 --- a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.md +++ b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.md @@ -28,7 +28,7 @@ A provider has exactly one adapter owner in a Cordis context. `dsh-llm-deepseek` ### Explicit pi-ai provider profiles -`dsh-llm-pi-ai` takes one non-empty list of provider profiles. Provider names must be unique within the list and present in pi-ai's `getProviders()` result. Each profile contains the provider name plus optional `apiKey`, `baseURL`, headers, reasoning level and budgets, cache retention, transport, timeouts, and retry settings. Credentials are never global: an explicit key applies only to its profile, while an absent key lets pi-ai resolve its standard environment variable, OAuth token, AWS credential chain, Google ADC, or other provider-native ambient authentication. An explicitly empty key is invalid configuration rather than an environment fallback. +`dsh-llm-pi-ai` takes one non-empty list of provider profiles. Provider names must be unique within the list and present in pi-ai's `getProviders()` result. Each profile contains the provider name plus optional `apiKey`, `baseURL`, headers, reasoning level and budgets, cache retention, transport, SDK timeouts, and a Harness stream-idle timeout. Provider retry fields are deliberately absent: the adapter forces pi-ai's `maxRetries` to zero so one `stream()` call makes one visible provider attempt, while `dsh-llm-retry` owns bounded agent-level recovery. Credentials are never global: an explicit key applies only to its profile, while an absent key lets pi-ai resolve its standard environment variable, OAuth token, AWS credential chain, Google ADC, or other provider-native ambient authentication. An explicitly empty key is invalid configuration rather than an environment fallback. The plugin registers all configured provider names against one `PiAiAdapter` in one all-or-nothing call. A request uses its provider to select the matching profile and finds its model in `getModels(provider)` to obtain the catalog descriptor. An unknown provider fails at plugin load; an unknown model fails before network I/O with `UNKNOWN_MODEL`. The catalog object is never mutated. When a profile supplies `baseURL`, the adapter clones the selected descriptor and overrides only `baseUrl`, so a private endpoint can retain pi-ai's API, capabilities, compatibility flags, context limits, and reasoning map. The private endpoint must implement the selected provider's protocol, and the model id must still exist in the installed pi-ai catalog. @@ -75,14 +75,14 @@ The on-disk session format remains the pre-release pinned version `0`, with no c - Provider names are deployment-wide route ownership keys: two providers may use the same model string, but mounting two adapters for one provider fails at load instead of creating fallback order. - Model selection no longer changes the Cordis plugin graph. Catalog-backed adapters can accept any installed catalog model selected after startup, while the native DeepSeek adapter forwards arbitrary DeepSeek model ids. - A custom `baseURL` preserves the selected catalog model's protocol and capabilities; it does not make catalog-external model ids valid. Private endpoints must implement that catalog entry's protocol. -- pi-ai credentials and transport knobs are scoped per provider profile. An omitted key delegates to pi-ai ambient authentication, while an explicitly empty key is invalid. +- pi-ai credentials, transport knobs, SDK timeouts, and the five-minute-default `streamIdleTimeoutMs` watchdog are scoped per provider profile. Hidden provider retries are disabled; bounded retries belong to the separately composed agent recovery policy. - `dsh-llm-pi-ai` rejects stop sequences because pi-ai's common stream API cannot express them; the native DeepSeek adapter retains its stop support. - Replay state is portable only within the adapter instance that owns both the historical and target providers. Cross-provider and cross-model restoration is an adapter responsibility, and another adapter receives provider-neutral history without the opaque state. - Current pre-release session JSONL requires provider/model request headers and assistant provenance. Older shapes remain version `0` but are rejected rather than migrated. ## Testing -- Unit coverage exercises registry conflicts, request reconstruction, session validation, profile resolution, option forwarding, native API selection including OpenAI Responses, conversion, replay validation, error mapping, cancellation, content rewrites, and same-instance versus different-instance replay dispatch. +- Unit coverage exercises registry conflicts, request reconstruction, session validation, profile resolution, single-attempt option forwarding, native API selection including OpenAI Responses, conversion, replay validation, error mapping, caller cancellation, idle-timeout transport termination, content rewrites, and same-instance versus different-instance replay dispatch. - Keyless loop/session tests and ACP snapshots exercise durable provider/model metadata, resume and fork propagation, workflow/subagent overrides, and unchanged user-visible transcripts; the key-gated DeepSeek e2e retains real provider streaming and tool follow-up coverage. - Public JSDoc, package READMEs, architecture and core-data-structure docs, generated catalogs, examples, session fixtures, and Python SDK pairs use provider/model targets consistently and are checked by the repository documentation and type-equivalence gates. diff --git a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.zh.md b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.zh.md index 7dcadf2521..c35225a86b 100644 --- a/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.zh.md +++ b/.agents/notes/implemented/architecture/2026-07-14-provider-routed-llm-adapters.zh.md @@ -28,7 +28,7 @@ Status: implemented ### 显式 pi-ai 提供方配置 -`dsh-llm-pi-ai` 接受一个非空的提供方配置列表。列表内的提供方名称必须唯一，并且存在于 pi-ai 的 `getProviders()` 结果中。每项配置包含提供方名称，以及可选的 `apiKey`、`baseURL`、headers、推理级别和预算、缓存保留设置、传输方式、超时和重试设置。凭据不设全局值：显式密钥仅对所属配置生效；未提供密钥时，pi-ai 使用标准环境变量、OAuth token、AWS 凭据链、Google ADC 或其他提供方原生环境认证。显式空密钥属于无效配置，不会回退到环境认证。 +`dsh-llm-pi-ai` 接受一个非空的提供方配置列表。列表内的提供方名称必须唯一，并且存在于 pi-ai 的 `getProviders()` 结果中。每项配置包含提供方名称，以及可选的 `apiKey`、`baseURL`、headers、推理级别和预算、缓存保留设置、传输方式、SDK 超时和 Harness 流空闲超时。配置中有意不提供重试字段：适配器强制将 pi-ai 的 `maxRetries` 设为零，使一次 `stream()` 调用只发起一次可见的提供方请求；有界的 agent 层恢复由 `dsh-llm-retry` 负责。凭据不设全局值：显式密钥仅对所属配置生效；未提供密钥时，pi-ai 使用标准环境变量、OAuth token、AWS 凭据链、Google ADC 或其他提供方原生环境认证。显式空密钥属于无效配置，不会回退到环境认证。 插件通过一次全有或全无调用，将所有已配置的提供方名称注册到同一个 `PiAiAdapter`。请求按 provider 选择对应配置，并在 `getModels(provider)` 中查找模型以取得目录描述符。未知提供方会在插件加载时失败；未知模型会在网络 I/O 前以 `UNKNOWN_MODEL` 失败。适配器不会修改目录对象。当配置提供 `baseURL` 时，适配器复制选中的描述符，仅覆盖 `baseUrl`，使私有端点保留 pi-ai 的 API、能力、兼容标志、上下文限制与推理映射。私有端点必须实现所选提供方的协议，模型 ID 也仍须存在于已安装的 pi-ai 目录中。 @@ -75,14 +75,14 @@ JSON-RPC 运行时显式接收 provider 与 model。仅当 `deepseek` 提供方 - 提供方名称是部署范围内的路由所有权键：两个提供方可以使用相同的模型字符串，但为同一个提供方挂载两个适配器会在加载时失败，不会形成回退顺序。 - 模型选择不再改变 Cordis 插件图。目录型适配器可以接受启动后选择的任意已安装目录模型，原生 DeepSeek 适配器则会转发任意 DeepSeek 模型 ID。 - 自定义 `baseURL` 会保留所选目录模型的协议与能力，但不会让目录外模型 ID 变为有效。私有端点必须实现该目录项对应的协议。 -- pi-ai 凭据与传输选项按提供方配置隔离。省略密钥时委托 pi-ai 使用环境认证；显式空密钥无效。 +- pi-ai 凭据、传输选项、SDK 超时，以及默认五分钟的 `streamIdleTimeoutMs` 空闲超时机制均按提供方配置隔离。系统禁用隐藏的提供方重试；有界重试由单独组合的 agent 恢复策略负责。 - pi-ai 的通用流 API 无法表达停止序列，因此 `dsh-llm-pi-ai` 会拒绝停止序列；原生 DeepSeek 适配器仍支持停止序列。 - 仅当历史提供方与目标提供方归同一个适配器实例所有时，回放状态才可移植。适配器负责跨提供方和跨模型恢复；其他适配器只接收不含不透明状态的提供方无关历史。 - 当前预发布会话 JSONL 要求请求头包含 provider/model，助手消息包含来源信息。旧格式仍使用版本 `0`，但会被拒绝，不执行迁移。 ## 测试 -- 单元测试覆盖注册表冲突、请求重建、会话验证、配置解析、选项转发、包括 OpenAI Responses 在内的原生 API 选择、转换、回放验证、错误映射、取消、内容重写，以及同一实例与不同实例间的回放分发。 +- 单元测试覆盖注册表冲突、请求重建、会话验证、配置解析、单次请求的选项转发、包括 OpenAI Responses 在内的原生 API 选择、转换、回放验证、错误映射、调用方取消、空闲超时导致的传输终止、内容重写，以及同一实例与不同实例间的回放分发。 - 无密钥的 agent loop/会话测试和 ACP 快照覆盖持久化 provider/model 元数据、恢复与 fork 传播、工作流/subagent 覆盖，以及不变的用户可见 transcript（文本记录）；密钥门控的 DeepSeek e2e 测试保留真实提供方的流式输出与工具后续调用覆盖率。 - 公共 JSDoc、package README、架构与核心数据结构文档、生成目录、示例、会话 fixture（测试前置数据）和 Python SDK 配对文档统一使用 provider/model 目标，并由仓库文档与类型等价门禁校验。 diff --git a/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.i18n.yaml new file mode 100644 index 0000000000..70b98f0cf1 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.i18n.yaml @@ -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 +2026-07-15-lsp-capability-seam.md: 7265b04ac9b2f83764bdd13f07b2d3404c4c1708 +2026-07-15-lsp-capability-seam.zh.md: 10e8956005045d0934dd9dada5718b85a34cda3f diff --git a/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.md b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.md new file mode 100644 index 0000000000..7265b04ac9 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.md @@ -0,0 +1,198 @@ +# Agent Note: LSP capability seam and model-facing query tool + +Status: implemented + +English | [中文](2026-07-15-lsp-capability-seam.zh.md) + +## Problem + +The harness has text search and file reads, but neither identifies a program symbol. A textual match cannot reliably distinguish two same-named functions, follow an import alias, connect an interface to its implementations, or report an inferred type. Before changing code, an agent therefore lacks the semantic navigation that a human gets from an editor's language server. + +LSP support has three owners: the model needs a stable query schema, the harness needs provider selection and normalized results, and the local implementation needs process, JSON-RPC, workspace, synchronization, and filesystem behavior. Combining them would bind the model contract to local subprocesses and obstruct remote or sandbox-native providers. + +Many language servers behave best when the queried document is opened with current text. A compatible agent client must bound that state, define whether its source read is a model observation, and keep the document snapshot in the same filesystem namespace as the server's workspace index. + +## Decision + +Add LSP as a three-package capability seam with one read-only model tool and one generic local provider implementation: + +1. `@deepseek-ai/dsh-lsp` at `packages/lsp/lsp` owns `ctx.lsp`, provider registration and selection, normalized requests/results, execution control, and structured LSP errors. +2. `@deepseek-ai/dsh-lsp-local` at `packages/lsp/lsp-local` adapts configured stdio language servers to the seam. One plugin instance accepts a named server table and registers one isolated provider for each command and extension-to-language-id mapping. +3. `@deepseek-ai/dsh-tool-lsp` at `packages/lsp/tool-lsp` owns the model-facing `lsp` schema, prompt guidance, argument validation, result limits and formatting, and ACP presentation. + +`dsh-lsp-local` is a generic host, not a language-server catalog or installer. Deployments explicitly configure commands and mappings; future presets belong in composition plugins or `cordis.yml` overlays. + +The model and seam expose exactly `goToDefinition`, `findReferences`, `goToImplementation`, and `hover`; no arbitrary JSON-RPC method escapes through `ctx.lsp`. These operation literals match Claude Code's familiar camelCase names while the tool name and `file_path` field remain harness-owned. + +The prompt positions LSP as a precision aid: `Use search/read for ordinary navigation. Use lsp when textual matches are ambiguous or before a change requires precise definitions, implementations, or references.` + +## Package and ownership boundaries + +`dsh-lsp` registers providers by branded id and extension-to-language-id mapping. `registerProvider()` atomically reserves the id and every normalized extension: invalid input or any conflict publishes nothing, and its disposer releases all reservations. Provider plugins register through `ctx.effect()`. Selection is per query and order-independent; no match returns a structured unavailable error. The first version has no glob, language-id, or explicit route selector and no statically declared operation capabilities. + +The seam exposes one `query(request, signal?)` operation because no fields need implementation defaulting: `workspaceRoot` is required, `languageId` comes from the registration, and consumers own timeouts and result limits. `query()` selects and derives without hidden `??` fallbacks, leaving no executable spec to resolve. `dsh-tool-lsp` validates model arguments and passes only `exec.signal` as a bare `AbortSignal`, matching web and keeping `dsh-lsp` independent of `dsh-tools`. Removal before selection fails as unavailable; later disposal follows the selected provider's cancellation lifecycle without rerouting. + +The intended contract shape is: + +```ts +import type { Branded } from '@deepseek-ai/dsh-brand' + +type LspOperation = 'goToDefinition' | 'findReferences' | 'goToImplementation' | 'hover' +type LspProviderId = Branded<'LspProviderId'> + +interface LspPosition { + readonly line: number + readonly character: number +} + +interface LspRange { + readonly start: LspPosition + readonly end: LspPosition +} + +interface LspQueryRequest { + readonly operation: LspOperation + readonly filePath: string + readonly position: LspPosition + readonly workspaceRoot: string +} + +interface LspProviderQuery extends LspQueryRequest { + readonly languageId: string +} + +type LspQueryResult = + | { readonly kind: 'locations'; readonly locations: readonly { readonly uri: string; readonly range: LspRange }[]; readonly resolvedWorkspaceRoot: string } + | { readonly kind: 'hover'; readonly hover: { readonly contents: string; readonly range?: LspRange } | null } + +interface LspProvider { + readonly id: LspProviderId + readonly extensionToLanguage: Readonly> + query(request: LspProviderQuery, signal?: AbortSignal): Promise +} + +interface LspService { + registerProvider(provider: LspProvider): () => void + query(request: LspQueryRequest, signal?: AbortSignal): Promise +} +``` + +Mapping keys normalize to lowercase, leading-dot extensions selected from `filePath`'s final extension; language ids only synchronize documents. Seam positions and ranges are zero-based UTF-16. `findReferences` always includes declarations: providers enforce this internally, the local mapping sets `context.includeDeclaration: true`, and callers get no flag. Closed result unions normalize navigation to locations and hover to content or `null`; navigation results carry the provider's resolved workspace root so consumers relativize file URIs in the same canonical namespace. The seam exposes no protocol types, process or document controls, or generic request escape hatch. + +`dsh-lsp-local` owns host files, server configuration, JSON-RPC, process and transient-document state, and protocol translation; it depends on `dsh-lsp` and Node APIs, not `dsh-fs`. The server-table key is its provider id. The plugin resolves every server-local setting before registration, rolls back earlier registrations if a later mapping is invalid or conflicts, and retains an independent process pool per provider. `dsh-tool-lsp` runtime-injects only `tools`, `lsp`, and `systemPrompt`, obtains the workspace from `exec.agent?.session.header.cwd` through a package-local `sessionCwd(exec)` helper matching the filesystem tools' lookup, and imports no provider. + +## Model-facing contract + +The single `lsp` tool accepts: + +```ts +interface LspToolInput { + readonly operation: 'goToDefinition' | 'findReferences' | 'goToImplementation' | 'hover' + readonly file_path: string + readonly line: number + readonly character: number +} +``` + +`line` and `character` are positive, one-based UTF-16 cursor coordinates; the tool converts them to the seam's zero-based `LspPosition` and converts rendered locations back. `findReferences` includes declarations so impact analysis does not omit the defining site. Provider, language id, workspace root, limits, timeout, initialization, and executable remain outside model input. + +The tool requires `workspaceRoot` from session `header.cwd`, with no fallback; absence fails as `LSP_WORKSPACE_REQUIRED` before querying or startup. The local provider resolves relative paths against that root and accepts absolute paths directly; both forms are canonicalized and rejected before startup when the target is outside the canonical workspace. + +Locations render as stable, file-grouped `path:line:character` entries. A `file:` URI accepted by Node `fileURLToPath()` becomes a relative path inside the workspace or an absolute path outside it; other URIs remain verbatim. `maxLocations` defaults to `100` and reports omitted items; `maxResultChars` defaults to `16_000` and bounds every complete rendered result, including its truncation metadata. Empty locations and `null` hover are successful no-result responses; missing or malformed server payloads fail with structured `LSP_MALFORMED_RESPONSE` errors. + +ACP uses `{ card: 'generic', kind: 'search', title, locations: [{ path: file_path, line }] }` with an args-derived operation/cursor `title`. Because `FileLocation` has no character, follow-along focuses the input line while the title preserves the cursor; presentation remains pure. + +## Timeout ownership + +`dsh-tool-lsp` attaches one configurable `timeoutMs` budget, default `60_000`, to the tool definition. `dsh-timeout-policy` enforces it and supplies `exec.signal`, which reaches `ctx.lsp.query`; the budget covers the complete queued open/query/close lifecycle and is not model-configurable. + +The seam and provider add no startup or request deadline. Non-tool callers therefore receive no hidden timeout and must supply an `AbortSignal`, using `deadline()` when they need a budget. + +Provider disposal occurs outside tool execution, so `dsh-lsp-local` keeps `shutdownTimeoutMs` (default `5_000`) for `shutdown`/`exit` and `killGraceMs` (default `2_000`) for both request-cancel grace and SIGTERM-to-SIGKILL escalation; the same bounds govern failed-instance cleanup. Timer values above Node's `2_147_483_647` ms scheduling range fail at load. The provider uses `deadline()` and `timeoutOf()` but owns request cancellation, process signals, and awaiting close because timeout notification does not terminate work. + +## Workspace, filesystem, and document synchronization + +`dsh-lsp-local` canonicalizes and reads through Node APIs in the subprocess's host namespace. It rejects missing, non-regular, non-UTF-8, oversized, or canonical out-of-workspace sources and keeps one `O_NOFOLLOW | O_NONBLOCK` handle through validation and reading, so a FIFO with no writer cannot block before the regular-file check. It observes caller cancellation around each filesystem operation. It does not consume `ctx.fs` or emit `fs/observed`: only the LSP result is model-visible, so the query does not satisfy read-before-write policy. + +The `read` tool is unsuitable source because its output is windowed, numbered, transcript-visible, and observed. Reading in `tool-lsp` would also assign provider-specific synchronization to the consumer and preclude non-local providers. + +The local provider uses a compatibility-first transient-open sequence for every query. It accepts legacy `textDocumentSync` `Full` or `Incremental`, or options with `openClose: true`; omitted, `None`, or explicitly incompatible synchronization fails as unsupported before `didOpen`. + +1. Canonicalize and validate the host path, then read the current source with Node filesystem APIs. +2. Send `textDocument/didOpen` with version `1`, full text, and the configured language id. Its write remains abortable; failure or cancellation invalidates the instance and awaits bounded process termination before the pool can reuse it. +3. Send the requested `textDocument/definition`, `textDocument/references`, `textDocument/implementation`, or `textDocument/hover` request. +4. If `didOpen` succeeded, attempt `textDocument/didClose` in `finally` after the request settles or aborts. A close-write failure does not replace the settled result or error, but invalidates the instance and awaits bounded process termination. + +Documents close after each call, so the first version needs no `didChange`, `didSave`, content cache, mutation listener, or document LRU. One abortable per-workspace provider queue serializes source-read/open/query/close lifecycles, so a waiting query reads current bytes only when its turn starts; the instance also keeps protocol lifecycles serialized. Distinct workspaces may run in parallel. The server's workspace index remains responsible for closed files reached from the source. + +The canonical workspace `realpath` must be a directory and supplies process cwd, `rootUri`, the sole `workspaceFolders` entry, and pool identity; symlink aliases therefore share an instance. Result locations may be external, but an external path cannot become a query source. Remote, virtual, or independently sandboxed filesystems require another provider. + +## Local server lifecycle and protocol behavior + +`dsh-lsp-local` lazily single-flights one server per `(provider id, canonical workspace realpath)`. At load it resolves the executable after credential scrubbing and environment overrides, failing before registration if unavailable; server process launch stays lazy (first query spawns it) and uses no shell. `maxMessageBytes` defaults to `16_000_000`, `maxStderrBytes` to `1_000_000`, and `maxDocumentBytes` to `4_000_000`. A crash fails the active query without replay; a later query may replace the process. Each query starts at most one process, so the MVP has no cross-request restart counter. + +Initialization advertises `general.positionEncodings: ['utf-16']`, `workspace: { workspaceFolders: true, configuration: true }`, `textDocument.hover.contentFormat: ['markdown', 'plaintext']`, and `linkSupport: true` for definition and implementation, with no dynamic registration. Returned operation and synchronization capabilities are authoritative. An omitted server `positionEncoding` defaults to `utf-16`; any other value is a protocol error. Configuration may supply initialization options and `workspace/configuration` responses, but the client rejects `workspace/applyEdit` and never executes commands or edits. + +Navigation maps `Location` directly and `LocationLink` from `targetUri` plus `targetSelectionRange`. Positions must be nonnegative integers. Hover normalization accepts only valid `MarkupContent` and `MarkedString` shapes, preserves string values, renders language-tagged values as fenced code, and joins arrays with one blank line. The model-facing tool applies `maxResultChars` after rendering. + +Abort reaches every query phase and sends `$/cancelRequest` once an id exists. An unresponsive server is terminated and awaited without collateral active work because the instance is serialized. Disposal rejects and cancels work, attempts graceful shutdown, escalates through bounded termination, and awaits quiescence. + +## Deliberately deferred surface + +Symbols are deferred because they need different schemas and overlap read/search; a future workspace-symbol tool must accept a search query. Call hierarchy is deferred because support is uneven, and `prepareCallHierarchy` remains an internal prerequisite rather than a model operation. + +Diagnostics need separate freshness, accumulation, and transcript rules. Mutations such as rename, code actions, and formatting require separate tools with preview, permission, and write-policy integration. + +The local provider trusts its configured server and claims no sandbox confinement. Supporting untrusted binaries requires a later process/filesystem contract for workspace reads plus private cache and temporary writes; restricted, remote, or virtual workspaces require another provider. + +## Alternatives considered + +**Copy Claude Code's unified schema.** Its cursor operations validate the core use case, but symbols and call hierarchy need different arguments. Copying all nine operations would freeze speculative surface, so the proposal aligns only on the four semantic queries. + +**Let providers register tools.** Loaded servers would then control model schema and prompts, preventing one stable contract across local and remote providers. + +**Expose arbitrary LSP methods.** A JSON-RPC escape hatch would leak protocol payloads and admit unreviewed mutation or command execution; the operation union stays closed. + +**Expose `resolve(request)` / `query(spec)`.** With no defaulted fields, resolution would only expose provider selection, and a public spec could outlive provider disposal or replacement. One operation keeps selection and invocation atomic to the registration lifetime. + +**Wrap the signal in a per-seam execution-context object.** Web passes a bare `AbortSignal`; wrapping this single field would add unexplained asymmetry. `query()` gains a context object only when another field requires it. + +**Read through `ctx.fs` or the `read` tool.** This could mix the document with a server index from another filesystem namespace; tool output is also windowed, numbered, and observed. The host-local provider reads unobserved full text beside its subprocess. + +**Keep documents open.** Mirroring edits requires version ownership, all-path `didChange`, HMR recovery, eviction, and stale-state rules. Transient opens avoid that MVP state machine. + +**Configure phase timeouts.** Nested timers create competing classifications and fresh budgets. One caller-owned deadline covers query work; only out-of-call teardown keeps local bounds. + +**Query without `didOpen`.** Although permitted, support is inconsistent and may use stale server state. Transient open supplies an explicit current snapshot. + +**Add routes or select the first match.** Registration order and HMR timing are not product semantics, while a route table duplicates unique extension ownership. Overlaps therefore fail registration. + +**Run concurrent queries in one instance.** If cancellation fails, terminating the shared process would kill unrelated work. Per-instance serialization limits that blast radius; instances remain parallel. + +**Ship presets or PATH discovery.** A catalog would make the generic host own language policy, while discovery cannot infer arguments, language ids, or initialization. Deployments configure providers explicitly; composition plugins may package presets. + +## Testing + +- Package tests pin the three-package dependency direction, runtime injections, and `ctx.lsp`-only communication. +- Tool tests pin the four operations, coordinate validation, configured bounds and omission markers, prompt, and ACP presentation. +- Registry tests pin atomic reservation/release, order-independent selection, and structured unavailable, disposed, conflict, and unsupported-operation errors. +- Fake-stdio tests pin exact initialization capabilities, four protocol mappings, `Location`/`LocationLink` and hover normalization, and `findReferences` mapping to `references.includeDeclaration`. +- Synchronization tests pin UTF-16 negotiation and conversion, supported and rejected `textDocumentSync` forms, blocked and failed open writes, balanced transient open/close, close-write failure, and malformed-response rejection. +- Timeout tests pin one `TOOL_TIMEOUT` budget, unclassified upstream cancellation, no hidden seam deadline, and bounded awaited teardown. +- Lifecycle tests pin startup single-flight, complete-lifecycle serialization with fresh queued source reads, cross-workspace parallelism, abortable queues, crash replacement without replay, failed-stdin teardown, and quiescent disposal. +- Host-filesystem tests pin session-cwd requirements, relative and absolute source containment through symlinks, document validation, file/non-file URI rendering, unformatted source, and no `fs/observed` event. +- A keyless pinned TypeScript real-server e2e exercises all four operations; runnable configuration uses the same explicit provider mapping. +- Snapshots cover model-visible schema, prompt, results, omissions, and ACP rendering; a built-artifact smoke test covers framing and cleanup. +- Package and architecture docs cover configuration, security boundaries, and search/read guidance; the new `packages/lsp/` group is added to the AGENTS.md repository-layout block, the packages/README.md group table, and architecture.md in the same change. + +## Consequences + +Language servers vary in method support, capability interpretation, and indexing readiness; LSP has no universal “index complete” signal. Servers without compatible transient-open synchronization are unsupported even if closed-document queries work. Supported servers may still return empty or partial results, so the tool promises no cross-server completeness. The pinned TypeScript e2e establishes one compatibility floor, not a cross-language claim. + +Transient opens repeat parsing and notifications. Per-instance serialization increases latency under parallel agents, and long-lived workspace processes consume memory until disposal. + +Extension ownership is exclusive within one runtime. Two providers cannot both claim `.ts`, even with different language ids; this is a conscious MVP limit. The intended extension is a deployment-configured selector above registrations that can relax exclusive reservations without adding provider choice to model input or changing `LspProvider.query`. + +UTF-16 cursor columns are exact for the protocol but difficult for a model to count around non-BMP characters. Invalid or off-symbol positions may produce empty results, so error text and prompt examples must explain the coordinate convention without encouraging broad LSP use. + +Direct Node access aligns the query snapshot with the server index but bypasses `ctx.fs` and its policy. Canonical containment rejects source files outside the workspace; a trusted server may still read the workspace and use caches. The first implementation therefore requires trusted host-local deployment and provides no sandbox guarantee. diff --git a/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.zh.md b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.zh.md new file mode 100644 index 0000000000..10e8956005 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-15-lsp-capability-seam.zh.md @@ -0,0 +1,198 @@ +# Agent Note: LSP 能力服务边界与面向模型的查询工具 + +Status: implemented + +[English](2026-07-15-lsp-capability-seam.md) | 中文 + +## 问题 + +harness 已具备文本搜索与文件读取能力，但二者都无法识别程序符号。文本匹配无法可靠地区分同名函数、跟踪导入别名、关联接口与具体实现，也无法报告推断类型。因此，agent（智能体）在修改代码前缺少人类通过编辑器语言服务器获得的语义导航能力。 + +语言服务器协议（Language Server Protocol，LSP）支持分属三个职责方：模型需要稳定的查询 schema，harness 需要提供方选择与规范化结果，本地实现则负责进程、JSON-RPC、工作区、同步与文件系统行为。将三者合并会使模型契约绑定本地子进程，并阻碍远程或沙箱原生提供方。 + +许多语言服务器在查询文档已按当前文本打开时表现最佳。兼容的 agent 客户端必须限制这项状态、定义内部读取是否算作模型观察，并确保文档快照与服务器工作区索引位于同一文件系统命名空间。 + +## 决策 + +将 LSP 建成由三个包（package）组成的能力服务边界，其中包含一个只读模型工具和一个通用本地提供方实现： + +1. `packages/lsp/lsp` 下的 `@deepseek-ai/dsh-lsp` 负责 `ctx.lsp`、提供方注册与选择、标准化请求与结果、执行控制，以及结构化 LSP 错误。 +2. `packages/lsp/lsp-local` 下的 `@deepseek-ai/dsh-lsp-local` 将配置的 stdio 语言服务器适配到该服务边界。一个插件实例接收具名服务器表，并为每组命令及扩展名到语言 id 的映射注册一个隔离的提供方。 +3. `packages/lsp/tool-lsp` 下的 `@deepseek-ai/dsh-tool-lsp` 负责面向模型的 `lsp` schema、提示词指导、参数校验、结果限制与格式化，以及 ACP（Agent Client Protocol）展示。 + +`dsh-lsp-local` 是通用 host，不是语言服务器目录或安装器。部署显式配置命令与映射；未来 preset 属于组合插件或 `cordis.yml` overlay。 + +模型与服务边界仅公开 `goToDefinition`、`findReferences`、`goToImplementation` 和 `hover`；`ctx.lsp` 不提供任意 JSON-RPC 方法。这些操作字面量与 Claude Code 熟悉的 camelCase 命名一致，而工具名与 `file_path` 字段仍由 harness 自行定义。 + +提示词将 LSP 定位为精确查询手段：`Use search/read for ordinary navigation. Use lsp when textual matches are ambiguous or before a change requires precise definitions, implementations, or references.` + +## 包与职责边界 + +`dsh-lsp` 按带品牌类型的 id 和扩展名到语言 id 的映射注册提供方。`registerProvider()` 以原子方式占用 id 与所有规范化扩展名：输入无效或存在冲突时不发布任何状态，清理函数释放全部占用。提供方插件通过 `ctx.effect()` 注册。系统按查询且不受顺序影响地选择提供方；没有匹配项时返回结构化不可用错误。第一版不提供 glob、language-id 或显式路由选择器，也不静态声明操作能力。 + +服务边界只公开 `query(request, signal?)`，因为没有字段需要实现层填充默认值：`workspaceRoot` 是必填项，`languageId` 来自注册映射，超时与结果限制由消费方负责。`query()` 执行选择与推导时不使用隐藏的 `??` 后备逻辑，因此没有需要 resolve 的可执行 spec。`dsh-tool-lsp` 校验模型参数，并只把 `exec.signal` 作为裸 `AbortSignal` 传递，与 web 一致，并使 `dsh-lsp` 不依赖 `dsh-tools`。提供方在选择前被移除时按不可用失败；之后的释放遵循已选提供方的取消生命周期，不改路由。 + +预期契约如下： + +```ts +import type { Branded } from '@deepseek-ai/dsh-brand' + +type LspOperation = 'goToDefinition' | 'findReferences' | 'goToImplementation' | 'hover' +type LspProviderId = Branded<'LspProviderId'> + +interface LspPosition { + readonly line: number + readonly character: number +} + +interface LspRange { + readonly start: LspPosition + readonly end: LspPosition +} + +interface LspQueryRequest { + readonly operation: LspOperation + readonly filePath: string + readonly position: LspPosition + readonly workspaceRoot: string +} + +interface LspProviderQuery extends LspQueryRequest { + readonly languageId: string +} + +type LspQueryResult = + | { readonly kind: 'locations'; readonly locations: readonly { readonly uri: string; readonly range: LspRange }[]; readonly resolvedWorkspaceRoot: string } + | { readonly kind: 'hover'; readonly hover: { readonly contents: string; readonly range?: LspRange } | null } + +interface LspProvider { + readonly id: LspProviderId + readonly extensionToLanguage: Readonly> + query(request: LspProviderQuery, signal?: AbortSignal): Promise +} + +interface LspService { + registerProvider(provider: LspProvider): () => void + query(request: LspQueryRequest, signal?: AbortSignal): Promise +} +``` + +映射键规范化为带前导点的小写扩展名，并按 `filePath` 的最后一个扩展名选择；语言 id 仅用于文档同步。服务边界中的位置和范围从零开始按 UTF-16 计数。`findReferences` 始终包含声明：提供方在内部执行该约束，本地映射设置 `context.includeDeclaration: true`，调用方不能配置。封闭结果联合将导航统一为位置，将 `hover` 统一为内容或 `null`；导航结果携带提供方解析后的工作区根目录，使消费方依据同一规范化根目录相对化文件 URI。服务边界不公开协议类型、进程或文档控制，也不提供通用请求逃生口。 + +`dsh-lsp-local` 负责主机文件、服务器配置、JSON-RPC、进程与临时文档状态和协议转换；它依赖 `dsh-lsp` 与 Node API，不依赖 `dsh-fs`。服务器表的键是提供方 id。插件在注册前解析每个服务器的本地设置；如果后续映射无效或发生冲突，插件会撤销此前的注册，并为每个提供方保留独立进程池。`dsh-tool-lsp` 在运行时只注入 `tools`、`lsp` 和 `systemPrompt`，通过包内的 `sessionCwd(exec)` 辅助函数从 `exec.agent?.session.header.cwd` 取得工作区，其取值方式与文件系统工具一致，也不导入提供方。 + +## 面向模型的契约 + +单一 `lsp` 工具接受以下参数： + +```ts +interface LspToolInput { + readonly operation: 'goToDefinition' | 'findReferences' | 'goToImplementation' | 'hover' + readonly file_path: string + readonly line: number + readonly character: number +} +``` + +`line` 和 `character` 是从一开始计数的正数 UTF-16 光标坐标；工具将其转换为服务边界中从零开始的 `LspPosition`，并将渲染位置转回。`findReferences` 包含声明，避免影响分析漏掉定义位置。提供方、语言 id、工作区根目录、限制、超时、初始化和可执行文件均不进入模型输入。 + +工具必须从会话 `header.cwd` 取得 `workspaceRoot`，没有后备值；缺失时在查询或启动前以 `LSP_WORKSPACE_REQUIRED` 失败。本地提供方基于根目录解析相对路径并直接接受绝对路径；两种路径都会进行规范化，如果目标位于规范工作区外，则在启动前拒绝。 + +位置按文件稳定分组并渲染为 `path:line:character`。Node `fileURLToPath()` 可接受的 `file:` URI 在工作区内转换为相对路径，在工作区外转换为绝对路径；其他 URI 保持原样。`maxLocations` 默认值为 `100`，并报告省略的条目；`maxResultChars` 默认值为 `16_000`，并限制每个完整渲染结果，其中包括截断元数据。空位置与 `null` hover 是成功的无结果响应；服务器载荷缺失或格式错误时，以结构化 `LSP_MALFORMED_RESPONSE` 错误失败。 + +ACP 使用 `{ card: 'generic', kind: 'search', title, locations: [{ path: file_path, line }] }`，`title` 由参数推导并标明操作与光标。由于 `FileLocation` 没有 character，跟随位置聚焦输入行，标题保留完整光标；展示保持纯函数。 + +## 超时归属 + +`dsh-tool-lsp` 将一个可配置的 `timeoutMs` 预算附加到工具定义，默认值为 `60_000`。`dsh-timeout-policy` 执行预算并提供传入 `ctx.lsp.query` 的 `exec.signal`；该预算覆盖排队、打开、查询和关闭的完整生命周期，模型不可配置。 + +服务边界和提供方不增加启动或请求截止时间。非工具调用方不会获得隐藏超时，必须自行提供 `AbortSignal`，并在需要预算时使用 `deadline()`。 + +提供方释放发生在工具执行之外，因此 `dsh-lsp-local` 保留 `shutdownTimeoutMs`（默认 `5_000`）限制 `shutdown`/`exit`，以及 `killGraceMs`（默认 `2_000`），同时用于限制请求取消宽限期和从 SIGTERM 升级到 SIGKILL 的宽限期；失败实例的清理也使用相同边界。定时器值超过 Node `2_147_483_647` ms 的调度范围时，插件加载失败。提供方使用 `deadline()` 和 `timeoutOf()`，但仍负责请求取消、进程信号和等待关闭，因为超时通知不会终止工作。 + +## 工作区、文件系统与文档同步 + +`dsh-lsp-local` 通过 Node API 在子进程所在的主机命名空间中规范化并读取文件。它拒绝缺失、非普通、非 UTF-8、超大或规范路径越出工作区的源文件，并在校验与读取期间保持同一个 `O_NOFOLLOW | O_NONBLOCK` 句柄，因此没有写入方的 FIFO 不会在普通文件校验前造成阻塞。它在每项文件系统操作前后检查调用方是否取消。它不使用 `ctx.fs` 或发送 `fs/observed`：只有 LSP 结果对模型可见，因此查询不满足写前读取策略。 + +`read` 工具的输出带窗口与行号，进入 transcript（文本记录）且已被观察，不适合作为源文件。在 `tool-lsp` 内读取还会把提供方专用同步职责交给消费方，并排除非本地提供方。 + +本地提供方对每次查询都采用兼容优先的临时打开流程。它接受旧式 `textDocumentSync` 的 `Full` 或 `Incremental`，也接受设置了 `openClose: true` 的选项；同步能力缺失、为 `None` 或明确不兼容时，在 `didOpen` 前以不支持错误失败。 + +1. 规范化并校验主机路径，再使用 Node 文件系统 API 读取当前源文件。 +2. 发送 `textDocument/didOpen`，其中包含版本 `1`、完整文本和配置的语言 id。该写入仍可取消；写入失败或遭取消会使实例失效，并等待有界进程终止完成，池才能复用它。 +3. 发送所请求的 `textDocument/definition`、`textDocument/references`、`textDocument/implementation` 或 `textDocument/hover` 请求。 +4. 如果 `didOpen` 成功，则在请求完成或取消后于 `finally` 中尝试发送 `textDocument/didClose`。关闭写入失败不会覆盖已经确定的结果或错误，但会使实例失效，并等待有界进程终止完成。 + +每次调用后都关闭文档，因此第一版不需要 `didChange`、`didSave`、内容缓存、变更监听器或文档 LRU。每个工作区的提供方队列可取消，并串行执行源文件读取、打开、查询和关闭的完整生命周期，因此等待中的查询只在轮到它时才读取当前字节；实例也会串行执行协议生命周期。不同工作区可以并行。服务器工作区索引仍负责从源文件跳转到的已关闭文件。 + +规范工作区 `realpath` 必须是目录，并用于进程 cwd、`rootUri`、唯一的 `workspaceFolders` 条目和进程池 identity；符号链接别名因此共享实例。结果位置可以在工作区外，但外部路径不能成为查询源。远程、虚拟或独立沙箱化文件系统需要另一种提供方。 + +## 本地服务器生命周期与协议行为 + +`dsh-lsp-local` 按 `(provider id, canonical workspace realpath)` 懒启动一个服务器，并通过 single-flight 合并启动。插件加载时，它在清除凭据并应用环境变量覆盖后解析可执行文件；命令不可用时在注册前失败。服务器进程的启动保持懒执行（首次查询时才拉起），且不经过 shell。`maxMessageBytes` 默认值为 `16_000_000`，`maxStderrBytes` 默认值为 `1_000_000`，`maxDocumentBytes` 默认值为 `4_000_000`。崩溃使当前查询失败且不重放；后续查询可以替换进程。每次查询最多启动一个进程，因此 MVP 不设置跨请求重启计数器。 + +初始化声明 `general.positionEncodings: ['utf-16']`、`workspace: { workspaceFolders: true, configuration: true }`、`textDocument.hover.contentFormat: ['markdown', 'plaintext']`，以及 definition 与 implementation 的 `linkSupport: true`，但不支持动态注册。服务器返回的操作与同步能力均为真源。服务器省略 `positionEncoding` 时默认为 `utf-16`；其他值均属于协议错误。配置可以提供初始化选项和 `workspace/configuration` 响应，但客户端拒绝 `workspace/applyEdit`，绝不执行命令或编辑。 + +导航结果直接映射 `Location`，并将 `LocationLink` 的 `targetUri` 与 `targetSelectionRange` 映射为统一位置。位置必须是非负整数。`hover` 归一化只接受有效的 `MarkupContent` 和 `MarkedString` 结构，保留字符串值，把带语言标签的值渲染为围栏代码块，并以一个空行连接数组。面向模型的工具在渲染后应用 `maxResultChars`。 + +取消信号传递到查询的所有阶段，请求 id 创建后还会发送 `$/cancelRequest`。无响应的服务器会被终止并等待关闭；实例串行化保证没有其他正在执行的工作被连带中断。资源释放会拒绝并取消工作、尝试优雅关闭、通过有界终止流程升级处理，并等待完全停稳。 + +## 明确延后的接口 + +符号操作因需要不同 schema 且与读取或搜索重叠而延后；未来的工作区符号工具必须接收搜索词。调用层级因支持度不一而延后，`prepareCallHierarchy` 仍是内部准备步骤，不是模型操作。 + +诊断需要独立的新鲜度、累积与 transcript 规则。重命名、代码操作和格式化等变更能力需要单独工具，并集成预览、权限和写入策略。 + +本地提供方信任配置的服务器，不声称具备沙箱隔离。支持不受信任的二进制文件需要后续补充允许读取工作区并写入私有缓存与临时目录的进程/文件系统契约；受限、远程或虚拟工作区需要另一种提供方。 + +## 备选方案 + +**照搬 Claude Code 的统一 schema。** 它的光标操作验证了核心场景，但符号与调用层级需要不同参数。照搬九种操作会固化尚未验证的接口，因此本提案只对齐四种语义查询。 + +**允许提供方注册工具。** 已加载服务器会控制模型 schema 和提示词，无法在本地与远程提供方之间维持统一契约。 + +**公开任意 LSP 方法。** JSON-RPC 逃生口会泄露协议载荷，并允许未经评审的变更或命令执行；操作联合保持封闭。 + +**公开 `resolve(request)` / `query(spec)`。** 没有需要填充默认值的字段时，resolve 只会暴露提供方选择，而公开 spec 可能活过提供方释放或替换。单一操作让选择与调用共用注册生命周期。 + +**将信号包装为每服务边界的执行上下文对象。** Web 传递裸 `AbortSignal`；仅包装这一个字段会造成无谓的不对称。只有另一个字段确有需要时，`query()` 才引入上下文对象。 + +**通过 `ctx.fs` 或 `read` 工具读取。** 这可能把文档与另一文件系统命名空间中的服务器索引混合；工具输出还带窗口、行号且已被观察。host-local 提供方在子进程旁读取未观察的完整文本。 + +**保持文档打开。** 镜像编辑需要版本归属、覆盖所有路径的 `didChange`、HMR 恢复、淘汰和陈旧状态规则。临时打开避免在 MVP 引入这套状态机。 + +**配置分阶段超时。** 嵌套定时器会产生相互竞争的分类与新预算。一个由调用方负责的截止时间覆盖查询；只有调用外清理保留本地限制。 + +**不发送 `didOpen`。** 协议虽允许，但支持不一致且可能使用陈旧服务器状态。临时打开提供明确的当前快照。 + +**增加路由或选择首个匹配项。** 注册顺序与 HMR 时机不是产品语义，路由表又会重复唯一扩展名所有权。因此，扩展名重叠时注册失败。 + +**在一个实例中并发查询。** 取消失败时，终止共享进程会杀死无关工作。实例内串行可限制影响范围；不同实例仍可并行。 + +**内置 preset 或 PATH 发现。** 目录会让通用 host 承担语言策略，而发现机制无法推断参数、语言 id 或初始化配置。部署显式配置提供方，组合插件可以封装 preset。 + +## 测试 + +- 包测试固定三个包的依赖方向、运行时注入和仅通过 `ctx.lsp` 通信的边界。 +- 工具测试固定四种操作、坐标校验、配置限制与省略标记、提示词和 ACP 展示。 +- 注册表测试固定原子占用/释放、不受顺序影响的选择，以及结构化的不可用、已释放、冲突和不支持操作错误。 +- 测试用 stdio server 固定精确的初始化能力、四种协议映射、`Location`/`LocationLink` 与 `hover` 归一化，以及 `findReferences` 到 `references.includeDeclaration` 的映射。 +- 同步测试固定 UTF-16 协商与转换、受支持和被拒绝的 `textDocumentSync` 形式、打开写入阻塞与失败、配对的临时打开/关闭、关闭写入失败和错误响应拒绝。 +- 超时测试固定一个 `TOOL_TIMEOUT` 预算、不对上游取消错误分类、服务边界无隐藏截止时间，以及受限且等待完成的清理。 +- 生命周期测试固定启动 single-flight、完整生命周期串行化及排队查询读取最新源文件、跨工作区并行、可取消队列、崩溃后不重放的替换、stdin 失败后的进程拆除，以及释放后完全停稳。 +- 主机文件系统测试固定 session cwd 要求、符号链接下相对与绝对源路径的规范 containment、文档校验、file/non-file URI 渲染、无格式源文本和不发送 `fs/observed`。 +- 无密钥且固定版本的 TypeScript 真实服务器 e2e 覆盖四种操作；可运行配置使用同一项显式提供方映射。 +- 快照覆盖模型可见 schema、提示词、结果、省略提示和 ACP 渲染；构建产物冒烟测试覆盖分帧与清理。 +- 包与架构文档覆盖配置、安全边界和搜索/读取指导；同一改动中，新的 `packages/lsp/` 包组要加入 AGENTS.md 的仓库布局块、packages/README.md 的分组表和 architecture.md。 + +## 影响 + +各语言服务器对方法支持、能力解释和索引就绪时机的处理不同；LSP 没有统一的“索引完成”信号。无法声明兼容临时打开同步能力的服务器不受支持，即使它能查询已关闭文档。受支持的服务器仍可能返回空结果或不完整结果，因此工具不承诺跨服务器完整性。固定的 TypeScript e2e 只建立一条兼容性基线，不代表跨语言承诺。 + +临时打开会重复解析并产生通知。实例内串行会增加并发 agent 的延迟，长期运行的工作区进程则持续占用内存直到释放。 + +同一运行时内的扩展名所有权互斥。即使 language id 不同，两个提供方也不能同时占用 `.ts`；这是有意接受的 MVP 限制。预期扩展方式是在注册之上增加由部署配置的 selector，允许放宽互斥占用，同时不向模型输入增加提供方选择，也不改变 `LspProvider.query`。 + +UTF-16 光标列与协议完全一致，但模型难以在包含非 BMP 字符的文本中准确计数。无效位置或不在符号上的位置可能返回空结果，因此错误文本和提示词示例必须说明坐标约定，同时避免鼓励模型广泛使用 LSP。 + +直接访问 Node 文件系统会对齐查询快照与服务器索引，但绕过 `ctx.fs` 及其策略。规范路径 containment 会拒绝工作区外的源文件；受信任的服务器仍可读取工作区并使用缓存。因此，第一版要求受信任的 host-local 部署，不提供沙箱保证。 diff --git a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.i18n.yaml index d49aedb545..dd16a5f327 100644 --- a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.i18n.yaml +++ b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.i18n.yaml @@ -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 -2026-07-15-replay-token-meter-service.md: 9bbc177f456e006179c466f8c245e4599db3dd5a -2026-07-15-replay-token-meter-service.zh.md: 4437626c8651a80537d45197a93733271a592173 +2026-07-15-replay-token-meter-service.md: 3496364663c1f73b8161461d1a229b19d9730c6d +2026-07-15-replay-token-meter-service.zh.md: 0bc4d9decac36bd5674cd0fb04f82fdcd277554e diff --git a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.md b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.md index 9bbc177f45..3496364663 100644 --- a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.md +++ b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.md @@ -6,7 +6,7 @@ English | [中文](2026-07-15-replay-token-meter-service.zh.md) ## Problem -Context pressure is useful outside compaction. A compaction backend, an overflow guard, or a future request-policy plugin can all need the same answer: how much of the configured context window does the durable request consume? Keeping that fold inside `dsh-compact-basic` duplicates replay logic, makes measurement unavailable without compaction, and encourages callers to reuse stale accounting. +Context pressure is useful outside compaction. A compaction backend, an overflow guard, or a future request-policy plugin can all need the same answer: how many tokens does the durable request consume? Keeping that fold inside `dsh-compact-basic` duplicates replay logic, makes measurement unavailable without compaction, and encourages callers to reuse stale accounting. Provider usage is not a complete answer. It describes one successful call under one exact request envelope, while the current surface can grow, shrink, or be replaced afterward. Sessions also switch providers and models, old logs can lack chunk provenance, and usage fields separate input, cache-read, cache-write, output, and reasoning counts. A useful service therefore combines the latest exact anchor with conservative heuristic repricing and exposes the log revision consumed by each result. @@ -14,9 +14,9 @@ Provider usage is not a complete answer. It describes one successful call under ### One concrete LLM-family service -`@deepseek-ai/dsh-token-meter` is one concrete package under `packages/llm/` and registers `ctx.tokenMeter`. It is not split into an interface and backend before a second implementation exists. `TokenMeterService` itself exposes `contextWindow`, `measure(session, requestHeader?)`, and `estimateMessage(message)`; consumers call the singleton service directly. +`@deepseek-ai/dsh-token-meter` is one concrete package under `packages/llm/` and registers `ctx.tokenMeter`. It is not split into an interface and backend before a second implementation exists. `TokenMeterService` itself exposes `measure(session, requestHeader?)` and `estimateMessage(message)`; consumers call the singleton service directly. -The service has one `contextWindow`, defaulting to 128,000 tokens and configurable as a positive integer. Estimation uses a fixed four-characters-per-token heuristic plus structural overhead. There are no model profiles, density settings, tokenizer backends, or language-specific strategies. +The service has no configuration. Estimation uses a fixed four-characters-per-token heuristic plus structural overhead. There are no model profiles, capacity settings, density settings, tokenizer backends, or language-specific strategies. Exact provider/model capacity is a separate adapter-owned query, as specified by the [routed model context and compaction policy Agent Note](2026-07-20-routed-model-context-and-compaction-policy.md). ### Per-session replay folds @@ -32,9 +32,9 @@ Usage sums the disjoint input, cache-read, cache-write, and output buckets. Reas `dsh-compact-basic` requires `ctx.tokenMeter`; `CompactService` gains no token methods or types. Configuration, the region transaction, and summarization stay in separate modules; the service registers automatic listeners itself, while `summarize()` remains its sole subclass hook. The singleton meter consistently prices pressure, retention, shadowed content, provenance, and non-shrinking-summary rejection. -Automatic compaction uses one unified measurement for each threshold-and-retention decision. The region transaction measures after appending its durable `compact/start` lock and again after asynchronous summarization; any intervening durable append changes `logRevision` and prevents replacement. +Automatic compaction uses one unified measurement for each threshold-and-retention decision. The region transaction measures after appending its durable `compact/start` lock and again after asynchronous summarization, then compares the detached surface-node vectors. An intervening surface mutation prevents replacement; `logRevision` may advance for unrelated log-only facts without invalidating an unchanged selected span. -Compact policy has service-wide defaults: threshold ratio `0.8`, retained tail `floor(contextWindow × 0.16)`, `summarizationProvider: ''`, `summarizationModel: ''`, `maxTokens: 8192`, `compactionRetries: 1`, `maxOverflowRetries: 1`, and `auto: true`. Top-level `thresholdRatio` and `retainTokens` override the pressure policy; retention must remain below the resulting threshold. The summarization provider and model must both be set or both be empty; an empty pair resolves the latest logged request target, then the `AgentOptions` pair. +Compact policy has service-wide defaults: threshold ratio `0.8`, retained-tail ratio `0.16`, `summarizationProvider: ''`, `summarizationModel: ''`, `maxTokens: 8192`, `compactionRetries: 1`, `maxOverflowRetries: 1`, and `auto: true`. Top-level fields apply to every routed target; exact provider/model entries in `modelPolicies` partially override them. Pressure scales ratios against capacity resolved from the owning adapter, and `retainTokens` may replace `retainRatio`; retention must remain below the resulting threshold. The summarization provider and model must both be set or both be empty; an empty pair resolves the latest logged request target, then the `AgentOptions` pair. Automatic pressure runs at `agent/post-step` and measures the canonical durable envelope produced under the provider/model actually selected by `agent/request`. A headerless session has no completed routed request to assess and produces no work; any routed target can use the singleton estimator. Canonical overflow recovery uses the same measurement for forced range selection and retries only after a proven surface replacement. @@ -46,14 +46,14 @@ Unit tests cover fixed estimation, envelope invalidation and anchor replacement, - **Keep estimation inside `CompactService`** — rejected because measurement has consumers and replay semantics independent of compaction; it would also force every compactor to expose the same unrelated API. - **Split a token-meter interface from a heuristic backend immediately** — rejected because only one implementation exists. One concrete service preserves the future seam without speculative packages or configuration. -- **Keep model-keyed windows and density profiles** — rejected because the deployment currently has one context policy and one estimator. Model registries, unknown-model failures, and configurable density add branches without a second behavior to select. +- **Put model-keyed windows and density profiles in the meter** — rejected because replay estimation does not own model routing or capacity facts. The route-owning adapter exposes capacity, while compact-basic owns the consumer-specific threshold and retention policy. - **Keep separate scalar and surface measurements** — rejected because callers would need two reads and revision matching for one decision. A scalar-only read could avoid cloning nodes below threshold, but the split API introduces a caller-side race window; the unified snapshot accepts O(surface) cloning in exchange for coherence. - **Treat provider usage as portable between envelopes** — rejected because model, tools, prefixes, and call config are request facts. Mismatch reprices the whole current request. ## Consequences - Token pressure has one replay-aware owner that compaction and future plugins can share. -- The default makes the bundled composition usable with two zero-config plugin entries; deployments override one context capacity when needed. +- The default makes the meter a zero-config composition entry; deployments configure capacity on each route-owning adapter and optional policy overrides on compact-basic. - Fixed heuristic pricing remains an estimate of provider behavior and is not an exact tokenizer or request serializer. - Every measurement clones the current positional surface and therefore costs O(surface), including pressure checks that finish below threshold. - Measurements fail loudly on malformed durable boundaries. This turns corrupted replay into a named integration failure instead of silently drifting pressure. diff --git a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.zh.md b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.zh.md index 4437626c86..0bc4d9deca 100644 --- a/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.zh.md +++ b/.agents/notes/implemented/architecture/2026-07-15-replay-token-meter-service.zh.md @@ -6,7 +6,7 @@ Status: implemented ## 问题 -上下文压力并不只对压缩有用。压缩后端、溢出保护或未来的请求策略插件都可能需要回答同一个问题：持久请求占用了已配置上下文窗口的多少容量？如果把该折叠逻辑留在 `dsh-compact-basic` 内部，就会重复实现回放逻辑，使未加载压缩的调用方无法使用计量，并诱使调用方复用陈旧的核算结果。 +上下文压力并不只对压缩有用。压缩后端、溢出保护或未来的请求策略插件都可能需要回答同一个问题：持久请求消耗了多少 token？如果把该折叠逻辑留在 `dsh-compact-basic` 内部，就会重复实现回放逻辑，使未加载压缩的调用方无法使用计量，并诱使调用方复用陈旧的核算结果。 提供方 usage 也不是完整答案。它只描述某个精确请求信封下的一次成功调用，而当前表层之后还可能增长、缩小或被替换。会话也可能切换提供方与模型，旧日志可能缺少分片来源，usage 字段还会分别报告输入、缓存读取、缓存写入、输出与推理计数。因此，可用的服务必须把最新精确锚点与保守的启发式重新定价结合起来，并公开每个结果已经消费的日志修订号。 @@ -14,9 +14,9 @@ Status: implemented ### 一个具体的 LLM 家族服务 -`@deepseek-ai/dsh-token-meter` 是 `packages/llm/` 下的单个具体包，并注册 `ctx.tokenMeter`。在第二种实现出现之前，它不会被拆成接口与后端。`TokenMeterService` 本身公开 `contextWindow`、`measure(session, requestHeader?)` 与 `estimateMessage(message)`；消费方直接调用这个单例服务。 +`@deepseek-ai/dsh-token-meter` 是 `packages/llm/` 下的单个具体包，并注册 `ctx.tokenMeter`。在第二种实现出现之前，它不会被拆成接口与后端。`TokenMeterService` 本身公开 `measure(session, requestHeader?)` 与 `estimateMessage(message)`；消费方直接调用这个单例服务。 -服务只有一个 `contextWindow`，默认值为 128,000 token，并允许配置为正整数。估算采用固定的每 token 四个字符启发式规则，并加上结构开销。服务不提供模型 profile、密度设置、分词器后端或语言专用策略。 +服务没有配置。估算采用固定的每 token 四个字符启发式规则，并加上结构开销。服务不提供模型 profile、容量设置、密度设置、分词器后端或语言专用策略。精确提供方/模型容量由独立的适配器查询拥有，具体见[路由模型上下文与压缩策略 Agent Note](2026-07-20-routed-model-context-and-compaction-policy.md)。 ### 逐会话回放折叠 @@ -32,9 +32,9 @@ Usage 会对互不重叠的输入、缓存读取、缓存写入与输出 bucket `dsh-compact-basic` 要求 `ctx.tokenMeter`；`CompactService` 不增加 token 方法或类型。配置、区域事务与摘要器分别保留在独立模块中，服务自身注册自动监听器，而 `summarize()` 仍是唯一的子类 hook。单例计量器一致用于压力、保留、被遮蔽内容、来源以及非缩小摘要拒绝的定价。 -自动压缩的每次阈值与保留联合决策只使用一次统一计量。区域事务先追加持久 `compact/start` 锁，再执行一次计量，并在异步摘要完成后再次计量；期间任何持久追加都会改变 `logRevision`，从而阻止替换。 +自动压缩的每次阈值与保留联合决策只使用一次统一计量。区域事务会在追加持久 `compact/start` 锁后执行计量，在异步摘要完成后再次计量，随后比较分离的表层节点向量。期间发生的表层变更会阻止替换；`logRevision` 可以因无关的纯日志事实而推进，而不会使未变的选定范围失效。 -压缩策略采用服务级默认值：阈值比例 `0.8`、保留尾部 `floor(contextWindow × 0.16)`、`summarizationProvider: ''`、`summarizationModel: ''`、`maxTokens: 8192`、`compactionRetries: 1`、`maxOverflowRetries: 1` 与 `auto: true`。顶层 `thresholdRatio` 与 `retainTokens` 覆盖压力策略；保留值必须小于最终阈值。摘要提供方与模型必须同时设置或同时为空；空组合先解析最近记录的请求目标，再使用 `AgentOptions` 中的组合。 +压缩策略采用服务级默认值：阈值比例 `0.8`、保留尾部比例 `0.16`、`summarizationProvider: ''`、`summarizationModel: ''`、`maxTokens: 8192`、`compactionRetries: 1`、`maxOverflowRetries: 1` 与 `auto: true`。顶层字段适用于每个路由目标；`modelPolicies` 中的精确提供方/模型项可以部分覆盖这些字段。压力检查根据所属适配器解析的容量缩放比例，`retainTokens` 可以替代 `retainRatio`；保留值必须小于最终阈值。摘要提供方与模型必须同时设置或同时为空；空组合先解析最近记录的请求目标，再使用 `AgentOptions` 中的组合。 自动压力检查运行在 `agent/post-step`，并计量 `agent/request` 实际所选提供方/模型产生的规范持久信封。没有请求头的会话尚无已完成的路由请求可供判断，因此不执行工作；任意路由目标都可使用这个单例估算器。规范化溢出恢复使用同一计量结果强制选择范围，并且只有在表层替换得到证明后才重试。 @@ -46,14 +46,14 @@ Usage 会对互不重叠的输入、缓存读取、缓存写入与输出 bucket - **把估算保留在 `CompactService` 内**——不予采纳，因为计量拥有独立于压缩的消费方与回放语义；它还会强迫每个压缩器暴露同一套无关 API。 - **立即把 token meter 拆成接口与启发式后端**——不予采纳，因为目前只有一种实现。单个具体服务保留未来接缝，同时避免推测性的包与配置。 -- **保留模型键控的窗口与密度 profile**——不予采纳，因为当前部署只有一种上下文策略与一个估算器。模型注册表、未知模型错误和可配置密度只增加分支，却没有第二种行为可供选择。 +- **把模型键控窗口与密度 profile 放进 meter**——不予采纳，因为回放估算不拥有模型路由或容量事实。路由所属适配器公开容量，compact-basic 则拥有消费方专用的阈值与保留策略。 - **保留独立的标量与表层计量**——不予采纳，因为消费方必须为一次决策执行两次读取并匹配修订号。仅读取标量可以避免在低于阈值时复制节点，但拆分 API 会在消费方引入竞态窗口；统一快照接受 O(surface) 复制成本，以换取结果一致性。 - **在不同信封之间移用提供方 usage**——不予采纳，因为模型、工具、前缀与调用配置都是请求事实。不匹配时会重新定价完整当前请求。 ## 后果 - Token 压力拥有一个可供压缩与未来插件共享的回放感知所有者。 -- 默认值让内置组合只需两个零配置插件条目即可使用；部署需要时只覆盖一个上下文容量。 +- 默认值让 meter 成为零配置组合项；部署在各个路由所属适配器上配置容量，并在 compact-basic 上配置可选策略覆盖。 - 固定启发式定价仍然只是提供方行为的估计，并不是精确分词器或请求序列化器。 - 每次计量都会复制当前的位置表层，因此成本为 O(surface)，低于阈值即可结束的压力检查也不例外。 - 遇到畸形持久边界时，计量会明确失败。这会把损坏的回放转化为具名集成错误，而不是让压力静默漂移。 diff --git a/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.i18n.yaml new file mode 100644 index 0000000000..158a78acd8 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.i18n.yaml @@ -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 +2026-07-16-explicit-turn-cancellation.md: 7ac743221084e663294954bfd048ba7ef1114f60 +2026-07-16-explicit-turn-cancellation.zh.md: 3dca6339787ebef749c0d6a15609376ede994a97 diff --git a/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md new file mode 100644 index 0000000000..7ac7432210 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md @@ -0,0 +1,55 @@ +# Agent Note: Explicit turn cancellation capability + +Status: implemented + +English | [中文](2026-07-16-explicit-turn-cancellation.zh.md) + +## Problem + +Cancellation is a control capability with a shorter lifetime than an Agent driver. A free-form string cannot distinguish callers exhaustively, and a step-local controller cannot interrupt prompt submission, prompt assembly, continuation, or terminal turn policy. Storing `Error`, `AbortSignal.reason`, or backend-private objects would also expose unstable runtime details to durable replay. + +The [initiating Agent scope decision](2026-07-15-agent-initiator-scope.md) intentionally carries only the exact Agent through AsyncLocalStorage. Adding turn, step, or signal state to that driver-lifetime boundary would make stale asynchronous descendants appear to retain authority over later turns. Cancellation therefore needs one turn owner and explicit propagation without creating another ambient context or public turn wrapper. + +## Decision + +Agent owns the runtime-only `AgentCancelCause` union `{ kind: 'user' } | { kind: 'parent' }`; `agent.cancel()` defaults to `user`. TypeScript enforces that vocabulary at this typed same-process seam, with no runtime validator, fallback, or special compatibility contract for untyped callers. An active `TurnCancellation` copies the typed discriminant into a fresh frozen signal reason; idle cancellation has no holder to mutate and does not arm later work. + +An interrupted live turn ends with the coarse durable `{ kind: 'aborted' }` outcome. The terminal event records what happened to the turn, while the runtime signal identifies who requested cancellation; it does not duplicate `user` or `parent` into replay. Session seed/load rejects legacy aborted records with a reason or any other extra field, so replay cannot reintroduce caller-owned cancellation detail. The process-local `agent/cancel-requested` notification is not durable; a future audit requirement uses a separate durable control-request event so a request and its eventual outcome remain distinct. Durable events contain no stack, signal, error object, free-form cancellation text, or backend-private detail. + +AgentLoop privately owns one `TurnCancellation` per prospective turn. It installs the holder before notifying `agent/status = running`, retains its single `AbortController` through prompt processing, prompt assembly, every step, model and tool execution, continuation, and `agent/turn-stop`, then clears the exact holder immediately before publishing `turn/end`. Terminal event observers and the following durability flush therefore cannot cancel already-completed turn work even though driver status may remain `running` until the flush settles. Every participating method, event, and request value receives that same explicit signal; the next turn receives a fresh signal. + +The driver keeps only a cause-less pre-run marker for queued work cancelled before a turn is claimed. An effective `cancel()` emits the observe-only `agent/cancel-requested` notification with its resolved typed cause before clearing queued and steering work or aborting the holder; notification failures cannot veto the stop, and an idle call emits nothing. Work synchronously queued by a notification observer is included in that clear, while work queued by a later signal abort observer belongs to the next turn. If a `running` listener synchronously cancels old work and sends a replacement, the driver discards the aborted holder and creates a fresh one for the replacement. Repeated cancellation is first-wins for the active holder, while later calls may still clear newly queued pending work. + +The explicit event signatures keep their positional form and place `signal` immediately before a waterfall's final `next`. Prompt submission, request configuration, step-result processing, continuation, and terminal stop join the pre-existing explicit signal seams for pre-step, session prefix, model generation, tool execution, approval, and subagent or workflow requests. Hook bridges must also supply `RunHookOptions.signal`, so a turn cancellation reaches the bash executor's process-group kill and join boundary. `SystemPrompt.assemble()` carries `signal?: AbortSignal` in `AssembleContext` because that object is an explicit request value that can also represent signal-less assembly outside a turn. Listeners may cooperate with the signal but must not retain it to control another turn. + +`ctx.agents` continues to carry only the initiating Agent. Ambient Agent presence does not imply liveness, a current turn, or cancellation authority, and `agentInterruptReasonOf(signal)` reads only its explicit argument. Concurrent Agents isolate both their initiator identities and their turn signals; a child driver shadows the parent initiator while its parent request signal still travels through the subagent seam. + +Agent disposal requests the runtime-only `{ kind: 'disposed' }` interruption on the active holder. If cancellation already won the controller reason, the reason cannot be rewritten, so terminal classification first checks lifecycle state: disposed wins, then a supported `user` or `parent` cause becomes the coarse aborted outcome, and unrelated exceptions retain the existing error path. ACP cancellation maps to `user`; in-process spawn and fork propagation map to `parent`. Remote ACP subagents retain their existing wire protocol. + +Cancellation remains cooperative. The loop checks interruption before and after awaited boundaries but does not use `Promise.race` to abandon an in-process listener, adapter, or tool Promise. Work that ignores the signal must settle before `whenIdle()`, handle disposal, and scope teardown report quiescence. + +## Verification + +Contract tests verify the typed caller union, frozen detachment, default and first-wins behavior, the coarse Session JSON round trip and legacy-record rejection, ACP `user`, in-process subagent `parent`, and disposal precedence. Loop tests make cooperative listeners wait on the signal at prompt submission, system-prompt assembly, session prefix, pre-step, request, model stream, step result, tool execution, continuation, and terminal stop; they assert one signal within a turn, a fresh signal across turns, and no cancellation authority during terminal publication or a blocked durability flush. A real hook bridge test cancels and reaps a blocked prompt hook before idle. + +Initiator-scope tests assert that every hook still observes the exact Agent and no ambient turn signal, concurrent Agents retain independent identities and signals, and a nested child driver shadows only identity. Race tests cover idle cancellation, pre-run cancellation, replacement submission from a `running` listener, repeated cancellation, and cancel-versus-dispose quiescence. + +## Alternatives considered + +**Store the signal in ALS.** ALS follows asynchronous descendants for the entire driver lifetime, while cancellation authority ends with one turn. A leaked callback could observe a stale signal or require mutable ambient state, so the initiator scope continues to carry only the Agent and control remains explicit. + +**Persist a free-form string reason.** Strings admit spelling drift, prevent exhaustive switching, and encourage consumers to parse presentation text. The runtime uses a closed discriminated union, while the terminal record needs only the stable aborted outcome. + +**Persist the typed caller cause in `turn/end`.** No production replay, UI, ACP, telemetry, or workflow consumer distinguishes `user` from `parent`. Copying the request source into the terminal result would conflate two facts and add Session-specific validation without a consumer; a future audit surface can record a separate cancellation-request event. + +**Define speculative `superseded`, `timeout`, and `shutdown` variants now.** No current Agent cancellation producer implements those semantics. `shutdown` is already lifecycle disposal, and timeout or supersession should enter the union only with an owning policy and unique terminal meaning. + +**Expose public turn or step context wrappers.** Existing positional seams already identify Agent, turn, and step. A wrapper would widen every API, duplicate ownership, and tempt callers to treat a captured object as durable authority. + +**Abandon uncooperative work after a grace period.** Returning idle while same-process work still runs breaks teardown and resource-ownership guarantees. Hard termination requires a worker or process isolation boundary and is outside this control seam. + +## Consequences + +Cancellation has one runtime owner, one signal per live turn, and one typed runtime caller vocabulary. Session retains the coarse `aborted` outcome that its consumers actually use, rejects reason-bearing legacy forms, and stays isolated from runtime objects. Cooperative cancellation reaches every asynchronous turn seam, including work before the first step and after the last one, while terminal publication and persistence remain outside its authority. + +The explicit signal adds parameters to several public events and requires plugins to forward cancellation deliberately. This is intentional: authority is visible at the call boundary, lifetime matches the turn, and stale ambient descendants cannot acquire control. Uncooperative in-process work may delay cancellation, but the reported quiescent state remains truthful. diff --git a/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.zh.md b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.zh.md new file mode 100644 index 0000000000..3dca633978 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.zh.md @@ -0,0 +1,55 @@ +# Agent Note：显式轮次取消能力 + +Status: implemented + +[English](2026-07-16-explicit-turn-cancellation.md) | 中文 + +## 问题 + +取消是一种生命周期短于 Agent（智能体）驱动器的控制能力。自由文本字符串无法穷尽地区分调用方，步骤级控制器也无法中断提示词提交、提示词组装、继续决策或轮次终止策略。持久化 `Error`、`AbortSignal.reason` 或后端私有对象还会向持久化回放暴露不稳定的运行时细节。 + +[发起 Agent 作用域决策](2026-07-15-agent-initiator-scope.md)有意让 AsyncLocalStorage 只携带同一个 Agent。若把轮次、步骤或 signal 状态加入这个与驱动器同生命周期的边界，陈旧的异步后代就会看似仍对后续轮次拥有权限。因此，取消需要一个轮次归属方并显式传播，且不创建另一套环境上下文或公开的轮次包装层。 + +## 决策 + +Agent 拥有仅用于运行时的 `AgentCancelCause` 联合类型 `{ kind: 'user' } | { kind: 'parent' }`；`agent.cancel()` 默认使用 `user`。TypeScript 在这份类型化同进程契约中强制执行该词汇，不提供运行时校验器、后备行为，也不为无类型调用方提供特殊兼容性契约。活跃的 `TurnCancellation` 会把类型化判别字段复制为一个全新且已冻结的 signal 原因；空闲状态下没有可修改的持有者，也不会让后续工作预先进入取消状态。 + +正在运行的轮次被中断后，以粗粒度的持久化结果 `{ kind: 'aborted' }` 结束。终态事件记录轮次发生了什么，运行时 signal 标识谁请求了取消；回放不会重复保存 `user` 或 `parent`。Session seed/load 会拒绝携带取消原因或任何其他额外字段的旧式中止记录，因此回放无法重新引入由调用方持有的取消细节。仅限进程内的 `agent/cancel-requested` 通知不会持久化；未来若有审计需求，应使用独立的持久化控制请求事件，让请求与最终结果保持为两项事实。持久化事件不包含调用栈、signal、错误对象、自由文本取消原因或后端私有细节。 + +AgentLoop 为每个待启动轮次私有地持有一个 `TurnCancellation`。它在通知 `agent/status = running` 前安装该持有者，使其中唯一的 `AbortController` 持续覆盖提示词处理、提示词组装、每个步骤、模型与工具执行、继续决策和 `agent/turn-stop`；随后在发布 `turn/end` 前立即清除所安装的那个持有者。因此，即使驱动器状态可能在持久化刷新结算前保持 `running`，终态事件观察者及其后的持久化刷新也无法取消已完成的轮次工作。所有参与的方法、事件和请求值都会收到同一个显式 signal；下一个轮次会收到全新的 signal。 + +对于轮次被认领前已取消的排队工作，驱动器只保留一个不携带取消原因的运行前标记。实际生效的 `cancel()` 会先发出仅供观察的 `agent/cancel-requested` 通知并携带最终确定的类型化取消原因，然后才清除排队工作和 steering（中途引导）工作或中止持有者；通知失败不能阻止此次停止，空闲状态下调用则不发出任何通知。通知观察者同步加入队列的工作也会被这次清除，而稍后由 signal 中止观察者加入队列的工作属于下一个轮次。若 `running` 监听器同步取消旧工作并发送替代提示词，驱动器会丢弃已中止的持有者，并为替代提示词创建全新的持有者。同一活跃持有者上的重复取消遵循首次请求优先，后续调用仍可清除新入队的待处理工作。 + +显式事件签名保留位置参数形式，并把 `signal` 放在 waterfall（瀑布式事件）的最后一个参数 `next` 之前。提示词提交、请求配置、步骤结果处理、继续决策和终止停止加入已有的步骤前处理、会话前缀、模型生成、工具执行、审批以及 subagent 或工作流请求的显式 signal seam。钩子桥接器也必须提供 `RunHookOptions.signal`，使轮次取消能够到达 Bash 执行器终止进程组并等待其退出的边界。`SystemPrompt.assemble()` 在 `AssembleContext` 中携带 `signal?: AbortSignal`，因为该对象是显式请求值，也可表示轮次之外不携带 signal 的组装。监听器可以配合该 signal 取消，但不得保留它来控制其他轮次。 + +`ctx.agents` 仍只携带发起 Agent。环境中的 Agent 并不代表存活、当前轮次或取消权限，`agentInterruptReasonOf(signal)` 也只读取其显式参数。并发 Agent 会同时隔离各自的发起方身份和轮次 signal；子驱动会遮蔽父发起方，而父请求 signal 仍通过 subagent seam 传递。 + +Agent dispose（资源释放）会在活跃持有者上请求仅用于运行时的 `{ kind: 'disposed' }` 中断。若取消已经先占用控制器的中断原因，该原因便无法改写，因此终态分类会先检查生命周期状态：资源释放结果优先，之后受支持的 `user` 或 `parent` 取消原因形成粗粒度的中止结果，其他异常保留现有错误路径。ACP（Agent Client Protocol）取消映射为 `user`；进程内 spawn 和 fork 的传播映射为 `parent`。远程 ACP subagent 保持现有协议。 + +取消仍然是协作式的。AgentLoop 会在异步等待边界前后检查中断，但不会用 `Promise.race` 放弃进程内监听器、适配器或工具 Promise。忽略 signal 的工作必须真正结算，`whenIdle()`、句柄 dispose 和作用域清理才会报告静止状态。 + +## 验证 + +契约测试验证类型化调用方联合类型、冻结且与调用方分离、默认行为与首次请求优先行为、粗粒度的会话 JSON 往返与旧式记录拒绝、ACP `user`、进程内 subagent `parent` 以及 dispose 优先级。AgentLoop 测试让协作式监听器在提示词提交、系统提示词组装、会话前缀、步骤前处理、请求、模型流、步骤结果、工具执行、继续决策和终止停止处等待 signal；并断言同一轮次使用一个 signal，不同轮次使用全新的 signal，终态发布期间和持久化刷新受阻期间不存在取消权限。真实钩子桥接器测试会在报告空闲状态前取消并回收受阻的提示词钩子。 + +发起方作用域测试断言所有钩子仍观察到同一个 Agent 且没有环境中的轮次 signal，并发 Agent 保持独立的身份与 signal，嵌套子驱动只遮蔽身份。竞态测试覆盖空闲状态取消、运行前取消、从 `running` 监听器提交替代提示词、重复取消以及取消与 dispose 竞争下的静止状态。 + +## 考虑过的替代方案 + +**把 signal 存入 ALS。** ALS 会在整个驱动器生命周期内跟随异步后代，而取消权限在一个轮次结束时就已终止。泄漏的回调可能观察到陈旧 signal，或者迫使实现使用可变的环境状态，因此发起方作用域继续只携带 Agent，控制能力继续显式传递。 + +**持久化自由文本原因。** 字符串允许拼写漂移、阻碍穷尽分支判断，还会鼓励消费方解析展示文本。运行时使用封闭的可辨识联合类型，终态记录只需要稳定的中止结果。 + +**在 `turn/end` 中持久化类型化调用方取消原因。** 当前没有任何生产环境中的回放、UI、ACP、遥测或工作流消费方区分 `user` 与 `parent`。把请求来源复制到终态结果会混淆两项事实，还会在没有消费方的情况下引入会话特有校验；未来的审计接口可以记录独立的取消请求事件。 + +**现在就定义推测性的 `superseded`、`timeout` 和 `shutdown` 变体。** 当前没有 Agent 取消生产方实现这些语义。`shutdown` 已经属于生命周期 dispose；超时或替代只有在拥有明确归属策略和唯一终态含义时才应进入联合类型。 + +**公开轮次或步骤上下文包装类型。** 现有位置参数 seam 已经标识 Agent、轮次和步骤。包装类型会加宽所有 API、重复归属，并诱导调用方把捕获的对象当成持久权限。 + +**在宽限期后放弃不协作的工作。** 同进程工作仍在运行时就报告空闲状态，会破坏资源清理与资源归属保证。硬终止需要 worker 或进程隔离边界，不属于该控制 seam。 + +## 后果 + +取消拥有一个运行时归属方、每个活跃轮次一个 signal，以及一套类型化的运行时调用方词汇。会话保留其消费方实际使用的粗粒度 `aborted` 结果，拒绝携带原因的旧式形式，并与运行时对象保持隔离。协作式取消覆盖每个异步轮次 seam，包括第一个步骤之前和最后一个步骤之后的工作，而终态发布和持久化仍在其权限范围之外。 + +显式 signal 会给多个公开事件增加参数，并要求插件有意识地转发取消。这是有意设计：权限在调用边界可见，生命周期与轮次匹配，陈旧的环境异步后代无法获得控制能力。不协作的进程内工作可能延迟取消，但所报告的静止状态仍然真实。 diff --git a/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.i18n.yaml new file mode 100644 index 0000000000..b27a40cc7f --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.i18n.yaml @@ -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 +2026-07-19-cooperative-tool-cancellation.md: 559012f10d41963698cc932727125de1b9ccfef7 +2026-07-19-cooperative-tool-cancellation.zh.md: 6af8e57349bba026ab22f257014c084c5c3c3f54 diff --git a/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.md b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.md new file mode 100644 index 0000000000..559012f10d --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.md @@ -0,0 +1,73 @@ +# Agent Note: Cooperative tool cancellation at the registry boundary + +Status: implemented + +English | [中文](2026-07-19-cooperative-tool-cancellation.zh.md) + +## Problem + +Every typed tool invocation needs a caller-owned cancellation signal. An optional `ToolExecutionInput.signal` lets direct callers omit ownership, makes `exec.signal` optional in every tool body, and encourages registry fallbacks that cannot represent the caller's actual lifetime. + +The pipeline also has different mutability needs at different stages. Tool implementations, pre-policy, post-policy, and result observers only borrow cancellation state, while an around-dispatch wrapper must temporarily replace the signal to add a deadline or another lexical cancellation scope. One mutable public type either grants mutation too broadly or prevents that composition. + +Cancellation can arrive before policy, during approval, inside an around-dispatch wait, after a tool body starts, or while post-policy waits. One undifferentiated `ABORTED` result cannot tell durable consumers whether body side effects were possible. Racing a tool promise against cancellation is not a safe fallback because abandoned same-process work continues after the registry reports completion. + +## Decision + +`ToolExecutionInput.signal` is a required readonly `AbortSignal`. `ToolExecution.signal` and `ToolRunContext.signal` are therefore required and readonly as well. Every typed caller supplies the signal it owns; the registry provides no overload, default controller, never-abort sentinel, or convenience execution path. + +`ToolDefinition.execute(args, exec)` keeps its existing signature. `defineTool()` contextually types `exec.signal` as a required `AbortSignal`, so every registered TypeScript tool can observe or forward cancellation without a cast. First-party direct callers and nested Code Mode dispatches pass their current operation signal explicitly. + +The registry trusts this typed same-process contract. It does not perform runtime `AbortSignal` validation or add hostile-input tests for an omitted or malformed signal. Validation remains at parser/config, model/tool JSON, durable/file, worker, process, and wire boundaries; untyped JavaScript that violates the TypeScript interface has no compatibility contract. + +### Mutability follows the pipeline stage + +`ToolDispatchExecution` is identical to `ToolExecution` except that its required `signal` is mutable. Only the `tools/execute` waterfall receives this type. Pre-policy, post-policy, result observers, guards, and tool implementations receive readonly views of a private registry-owned mutable run object. + +An around-dispatch wrapper may replace `exec.signal` for its delegated lifetime but cannot typefully delete it or assign `undefined`. The registry captures the required caller signal outside that mutable object, fuses every wrapper replacement with the caller signal immediately before body invocation, removes dispatch-scoped listeners after settlement, and restores the required upstream signal unconditionally. + +### Cancellation codes record whether dispatch occurred + +`dsh-tools` exports `TOOL_ABORTED = 'ABORTED'` and `TOOL_ABORTED_BEFORE_DISPATCH = 'ABORTED_BEFORE_DISPATCH'`. The registry records body invocation immediately before calling `ToolDefinition.execute()`. + +`ABORTED_BEFORE_DISPATCH` carries `{ name: 'AbortError' }` and model text `Error: tool call aborted before dispatch`. It applies whenever cancellation prevents body invocation, including pre-aborted entry, cancellation during pre-policy or approval, an aborted wrapper signal, a wrapper success overtaken by caller cancellation before delegation, and agent-loop siblings skipped after turn cancellation. + +`ABORTED` carries model text `Error: tool call aborted` and applies only after the body was invoked, including cancellation while an around wrapper or post-policy listener waits after body completion. A denial, wrapper failure, tool failure, or post-policy failure remains more specific than generic cancellation. A timeout owned by timeout-policy remains `TOOL_TIMEOUT`, and contexts deferred before a successful outcome is replaced remain attached. + +### Pre-aborted entry short-circuits after materialization + +The registry first creates the call token and losslessly snapshots and freezes the arguments. A materialization failure wins even when the caller signal is already aborted. After successful materialization, a pre-aborted signal skips `tools/pre-execute`, approval, `tools/execute`, `tools/post-execute`, and the tool body, then publishes exactly one frozen authoritative `tools/result` with `ABORTED_BEFORE_DISPATCH`. + +### Started work still reaches quiescence + +Once a tool body starts, the registry awaits it. Cancellation reaches the body through the fused signal but never races or abandons its promise. A cooperative implementation stops or forwards cancellation and settles after its owned work reaches quiescence; an uncooperative same-process implementation can keep the registry pending indefinitely. Process, worker, network, and provider layers retain responsibility for their own termination mechanisms. + +This decision requires cancellation at the tool invocation seam only. Making signals required on asynchronous capabilities reachable from tool bodies is a separate migration proposed in [Required cancellation through tool-reachable capability seams](../../proposed/architecture/2026-07-19-required-cancellation-through-tool-capability-seams.md). + +## Verification + +[`execution-signal-types.spec.ts`](../../../../packages/core/tools/tests/execution-signal-types.spec.ts) proves the required exact signal types, readonly observer and tool views, mutable-but-required around-dispatch view, and `defineTool()` inference. [`tools.spec.ts`](../../../../packages/core/tools/tests/tools.spec.ts) covers pre-aborted materialization, phase skipping, policy and wrapper races, body invocation classification, caller-signal fusion, error precedence, context retention, and quiescent drainage. [`tool-calls.spec.ts`](../../../../packages/core/agent-loop/tests/tool-calls.spec.ts) and [`contract-regressions.spec.ts`](../../../../packages/core/agent-loop/tests/contract-regressions.spec.ts) cover balanced durable results for undispatched siblings. [`code-mode.spec.ts`](../../../../packages/core/tools/tests/code-mode.spec.ts) and first-party integration suites cover explicit forwarding, while [`timeout-policy.spec.ts`](../../../../packages/timeout/timeout-policy/tests/timeout-policy.spec.ts) preserves timeout ownership. + +No registry test can prove that arbitrary third-party same-process code observes the signal or stops in bounded time. Capability tests continue to prove cancellation and quiescence at the boundary that owns each side effect. + +## Alternatives considered + +**Keep the signal optional and synthesize a fallback.** Rejected because a registry-owned fallback has no caller lifetime to represent and preserves the exact omission the type should prevent. + +**Validate `AbortSignal` at runtime.** Rejected because this is a typed same-process seam, not a serialization boundary. Runtime checks would duplicate the static contract without making cooperative use enforceable. + +**Add `supportsCancellation` metadata, callback-arity checks, or signal-use linting.** Rejected because none proves that asynchronous work observes or correctly forwards cancellation. Availability is a type contract; behavior remains a tool and capability responsibility. + +**Expose one mutable execution type to every stage.** Rejected because observers and tool implementations only borrow the signal. Stage-specific types make replacement possible only where the pipeline owns that operation. + +**Forbid around wrappers from replacing the signal.** Rejected because deadlines and nested operational scopes need lexical derivation. Capturing and fusing the caller signal preserves composition without allowing detachment. + +**Race the tool promise against cancellation.** Rejected because it reports completion while side effects may remain live, violating the [quiescent-disposal rule](../../../../docs/defensive-patterns.md#dispose-must-reach-quiescence-not-just-request-it). + +## Consequences + +- TypeScript rejects every `ToolExecutionInput` that omits `signal`, every tool or observer mutation of a readonly signal, and every around-dispatch attempt to remove the signal. +- Durable consumers can distinguish calls whose body may have produced side effects (`ABORTED`) from calls that never entered the body (`ABORTED_BEFORE_DISPATCH`). +- The change is intentionally breaking under the repository's pre-release stance; no compatibility overload or runtime fallback remains. +- Cooperative tools stop promptly and reach quiescence; an implementation that ignores its signal remains observable as a pending call. +- Downstream capability interfaces remain unchanged until the linked proposed Agent Note is accepted and implemented. diff --git a/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.zh.md b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.zh.md new file mode 100644 index 0000000000..6af8e57349 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-cooperative-tool-cancellation.zh.md @@ -0,0 +1,73 @@ +# Agent Note: 注册表边界上的协作式工具取消 + +Status: implemented + +[English](2026-07-19-cooperative-tool-cancellation.md) | 中文 + +## 问题 + +每次类型化工具调用都需要一个由调用方持有的取消信号。可选的 `ToolExecutionInput.signal` 允许直接调用方不承担所有权，使每个工具主体中的 `exec.signal` 都成为可选值，也会诱使注册表提供无法表达真实调用方生命周期的后备信号。 + +流水线各阶段对可变性的需求也不同。工具实现、前置策略、后置策略和结果观察者只借用取消状态，而环绕调度包装层必须临时替换信号，以加入截止时间或其他词法取消作用域。单一的可变公开类型要么把修改权限授予过多阶段，要么阻止这种组合。 + +取消可能发生在策略之前、审批期间、环绕调度等待期间、工具主体启动之后，或后置策略等待期间。单一的 `ABORTED` 结果无法让持久化结果的使用方判断工具主体是否可能产生过副作用。让工具 promise 与取消竞速也不是安全的后备方案，因为注册表报告完成后，被丢弃的同进程工作仍会继续运行。 + +## 决策 + +`ToolExecutionInput.signal` 是必填且只读的 `AbortSignal`，因此 `ToolExecution.signal` 和 `ToolRunContext.signal` 也都是必填且只读。每个类型化调用方显式提供自己持有的信号；注册表不提供重载、默认控制器、永不中止哨兵或便捷执行路径。 + +`ToolDefinition.execute(args, exec)` 保持现有签名。`defineTool()` 会把 `exec.signal` 上下文推断为必填的 `AbortSignal`，因此每个已注册的 TypeScript 工具都能在无需类型断言的情况下观察或转发取消。所有第一方直接调用方和 Code Mode 嵌套调度都会显式传入当前操作的信号。 + +注册表信任这份类型化同进程契约。它不在运行时校验 `AbortSignal`，也不为缺失或畸形信号添加敌意输入测试。校验仍位于解析器与配置、队列、模型与工具 JSON、持久化与文件、worker、进程和线协议边界；违反 TypeScript 接口的无类型 JavaScript 不享有兼容性契约。 + +### 可变性由流水线阶段决定 + +`ToolDispatchExecution` 与 `ToolExecution` 相同，唯一差异是其必填 `signal` 可修改。只有 `tools/execute` waterfall（瀑布式事件）接收这个类型。前置策略、后置策略、结果观察者、守卫和工具实现接收注册表私有可变运行对象的只读视图。 + +环绕调度包装层可以在委托期间替换 `exec.signal`，但无法通过类型系统删除它或赋值为 `undefined`。注册表在可变对象之外捕获必填的调用方信号，在工具主体调用前把每次包装层替换与调用方信号融合，在完成后移除仅属于本次调度的监听器，并无条件恢复必填的上游信号。 + +### 取消代码记录是否发生过调度 + +`dsh-tools` 导出 `TOOL_ABORTED = 'ABORTED'` 和 `TOOL_ABORTED_BEFORE_DISPATCH = 'ABORTED_BEFORE_DISPATCH'`。注册表在调用 `ToolDefinition.execute()` 的前一刻记录工具主体已经开始。 + +`ABORTED_BEFORE_DISPATCH` 携带 `{ name: 'AbortError' }` 和模型可见文本 `Error: tool call aborted before dispatch`。凡取消阻止工具主体调用时都使用该结果，包括进入时已中止、前置策略或审批期间取消、包装层信号已中止、包装层在委托前返回的成功结果被调用方取消抢先，以及轮次取消后 agent loop 跳过的同批调用。 + +`ABORTED` 携带模型可见文本 `Error: tool call aborted`，并且只在工具主体已经调用后使用，包括工具主体完成后环绕包装层或后置策略监听器等待期间发生的取消。拒绝、包装层失败、工具失败或后置策略失败比通用取消更具体。timeout-policy 自身拥有的超时仍为 `TOOL_TIMEOUT`，成功结果被取消替换前延后附加的上下文仍会保留。 + +### 进入时已中止会在物化后短路 + +注册表先创建调用 token，并对参数进行无损快照和冻结。即使调用方信号已经中止，参数物化失败仍优先返回。物化成功后，进入时已中止的信号会跳过 `tools/pre-execute`、审批、`tools/execute`、`tools/post-execute` 和工具主体，然后发布且只发布一次冻结的权威 `tools/result`，其代码为 `ABORTED_BEFORE_DISPATCH`。 + +### 已启动工作仍必须完全停稳 + +工具主体一旦启动，注册表就会等待它完成。取消通过融合信号到达工具主体，但注册表不会与其 promise 竞速或丢弃该 promise。协作式实现会停止自身工作或继续转发取消，并在所持有的工作完全停稳后完成；不协作的同进程实现可能让注册表无限期保持等待。进程、worker、网络和提供方层仍负责各自的终止机制。 + +这项决策只要求工具调用接缝携带取消信号。让工具主体可达的异步能力也必须接收信号，属于另一项迁移，见提议中的[工具可达能力接缝中的必填取消](../../proposed/architecture/2026-07-19-required-cancellation-through-tool-capability-seams.md)。 + +## 验证 + +[`execution-signal-types.spec.ts`](../../../../packages/core/tools/tests/execution-signal-types.spec.ts) 证明必填的精确信号类型、观察者与工具的只读视图、环绕调度可替换但不可删除的视图，以及 `defineTool()` 推断。[`tools.spec.ts`](../../../../packages/core/tools/tests/tools.spec.ts) 覆盖进入时已中止的物化与阶段跳过、策略和包装层竞态、工具主体调用分类、调用方信号融合、错误优先级、上下文保留和完全停稳。[`tool-calls.spec.ts`](../../../../packages/core/agent-loop/tests/tool-calls.spec.ts) 与 [`contract-regressions.spec.ts`](../../../../packages/core/agent-loop/tests/contract-regressions.spec.ts) 覆盖未调度同批调用的持久化配对结果。[`code-mode.spec.ts`](../../../../packages/core/tools/tests/code-mode.spec.ts) 和第一方集成测试覆盖显式转发，[`timeout-policy.spec.ts`](../../../../packages/timeout/timeout-policy/tests/timeout-policy.spec.ts) 保持超时归属。 + +任何注册表测试都无法证明任意第三方同进程代码会观察信号或在有界时间内停止。各能力的测试仍需在拥有相应副作用的边界证明取消与完全停稳。 + +## 考虑过的替代方案 + +**保留可选信号并生成后备值。** 不予采纳，因为注册表持有的后备信号不代表任何调用方生命周期，也会保留类型系统本应阻止的缺失情况。 + +**在运行时校验 `AbortSignal`。** 不予采纳，因为这是类型化同进程接缝，不是序列化边界。运行时检查只会重复静态契约，仍无法强制实现协作式使用信号。 + +**添加 `supportsCancellation` 元数据、回调参数数量检查或信号使用 lint。** 不予采纳，因为这些方法都无法证明异步工作会观察或正确转发取消。信号可用性属于类型契约；具体行为仍由工具和能力负责。 + +**向所有阶段公开同一个可变执行类型。** 不予采纳，因为观察者和工具实现只需要借用信号。按阶段划分类型可以把替换权限限制在流水线拥有该操作的位置。 + +**禁止环绕包装层替换信号。** 不予采纳，因为截止时间和嵌套运行时作用域需要词法派生信号。捕获并融合调用方信号既保留组合能力，也不允许切断调用方取消。 + +**让工具 promise 与取消竞速。** 不予采纳，因为这种方式会在副作用仍可能存活时报告完成，违反[资源释放必须完全停稳的规则](../../../../docs/defensive-patterns.md#dispose-must-reach-quiescence-not-just-request-it)。 + +## 后果 + +- TypeScript 会拒绝所有缺少 `signal` 的 `ToolExecutionInput`、工具或观察者对只读信号的修改，以及环绕调度删除信号的尝试。 +- 持久化结果的使用方可以区分工具主体可能产生过副作用的调用（`ABORTED`）和从未进入工具主体的调用（`ABORTED_BEFORE_DISPATCH`）。 +- 根据仓库的预发布原则，这项变更刻意保持破坏性；不保留兼容重载或运行时后备行为。 +- 协作式工具会及时停止并完全停稳；忽略信号的实现会表现为仍在等待的调用。 +- 下游能力接口保持不变，直到关联的提议 Agent Note 被接受并实现。 diff --git a/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.i18n.yaml new file mode 100644 index 0000000000..d3e5608c22 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.i18n.yaml @@ -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 +2026-07-19-gui-layering-and-rpc-protocol.md: ebe21a6060ec69ba9807ab9fbf9906ae24b07823 +2026-07-19-gui-layering-and-rpc-protocol.zh.md: 0c256b60ce44a8e16ec6edfba146c776c4ae2129 diff --git a/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.md b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.md new file mode 100644 index 0000000000..ebe21a6060 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.md @@ -0,0 +1,253 @@ +# Agent Note: GUI layering and the RPC protocol — host/client layering by capability provider, the four-quadrant message model, and the fetch carrier + +Status: implemented + +English | [中文](2026-07-19-gui-layering-and-rpc-protocol.zh.md) + +> Division of labor: this document = the layering model + the channel-independent RPC protocol; the protocol's Web implementation (HTTP+SSE) is in the [web client architecture RFC](2026-07-19-gui-web-client-architecture.md). + +## Problem + +We need a UI integration layer. Beyond the existing ACP/stdio baseline, more product UI shapes are coming — Web (server), Electron, and others. We call these shapes Clients, uniformly, and want the following capabilities: + +- One `dsh` process supporting both `dsh web` (serve) and `dsh -p` (headless) — one process, two modes (a design reservation) +- Launching inside Electron with the same Web technology shape as `dsh web` + +That demands a stable layered responsibility model in the engineering codebase, so future client shapes plug in cleanly. + +At the same time the physical channels differ per consumer (HTTP/SSE, in-process direct calls, IPC later), so we also need a channel-independent message model and a single contract source of truth — "adding a method" and "swapping a carrier" must not entangle each other, and every message on the wire must be type-validatable, observable, and reconcilable. + +## Decision + +### Layering + +Directories layer as follows: + +- `packages/host/*`: packages provide host-side capability only (representing the Node.js engineering core built on the existing harness plugin system), and additionally + - the unified backend protocol (fetch, HTTP, streaming interfaces…) — definitions and support, see the "Message protocol" sections below +- `packages/client/*`: packages provide client-side capability only; every package stays single-sided. Two kinds live here: + - **Pure libraries** (`ui-slots`, `web-react`, `ui-primitives`): ordinary root-index packages, statically bundled into the shell and seeded into the browser plugin loader's module table. + - **dshClient plugin packages** (`connection`, `runtime`, `ui-theme`, `i18n`, `ui-layout`, `ui-sidebar`, `ui-conversation`, `ui-trajectory`): dual-entry — the root index is the node half (an empty `apply`, existing so the host Loader governs lifecycle and the web plugin registry discovers the package.json `dshClient` declaration); the entire implementation and its types live under `src/client/`, shipped as the `./client` subpath (a tsdown closure-factory bundle), and cross-package consumption imports the `/client` form. `runtime` additionally exports `./loader` (the shell-held browser bundle loader — a loader cannot load itself). +- `apps/` holds the externally exported application shapes, assembled from Client / Host mixtures. + - `apps/web` (`dsh-frontend`) is the vite application: a thin `main.ts` over the shell surface exported by `dsh-client-web`. + - `apps/cli` (`@deepseek-ai/dsh`) dispatches shapes: `dsh web` = startHost + webserver + the built `dsh-frontend` dist; `dsh -p` = headless in-process calls, zero HTTP. + - A future Electron shape reuses the same web client packages over an IPC fetch carrier. + +``` +apps/* (application shapes: apps/web = vite app, apps/cli = bin dispatch) + │ consume + ▼ +packages/host/* packages/client/* + apiproxy front layer: protocol pure libs: ui-slots / web-react / ui-primitives + runtime assembly / host entity dshClient plugins ×8 (node half = empty apply, + webserver web-shape HTTP carriage client half = src/client/) + │ ctx.plugin(...) ▲ import only apiproxy's /api /client subpaths + ▼ │ (type-only + the client base class) +harness core packages ──────────────────┘ (types reach the browser via import type) +``` + +Direction discipline (every rule auditable from package deps): + +- `runtime → apiproxy` is one-way; apiproxy depends only on type definitions. +- Client-side packages **never import** host-side package runtime (they consume only the two browser-safe subpaths `/api` and `/client`). +- `webserver` does not depend on `runtime`: it provides a `{ fetch }`-shaped implementation — "webserver ← runtime" is a runtime injection relationship, not a package dependency. +- Cross-package client imports use the `/client` subpath for plugin packages (a bare package name would inline a second runtime instance into a browser bundle; the tsdown purity gate rewrites or rejects it). + +TypeScript checks in **two aggregate programs** (`tsconfig.json` = host side + tests, excluding `packages/client`; `tsconfig.client.json` = client packages and their tests): both sides merge the cordis `Context` interface under the same keys (`sessions`, `loader`) with different services, so one program would see both declaration merges and report a collision. Shared leaves (session/llm/tools/apiproxy…) build once and are referenced by both programs. + +On the protocol side: TS interfaces (`packages/host/apiproxy/src/api/`, zero Node dependencies, browser-importable); wire messages unify under a **bidirectional model** — each logical message is shaped by "who initiates × request/response" (two axes, four cells, called the four quadrants below), decoupled from the physical channel; clients all inherit `AbstractApiClient` (protocol invariants live entirely in the base class, platform differences are just the `doFetch` transport aspect). + +#### Layer roles + +| Layer | Package | Responsibility | Key discipline | +|---|---|---|---| +| Front layer | `dsh-host-apiproxy` | TS/zod definitions (api/) + the fetch abstraction (fetch/: handler + client base class) | Keep it simple — every consumer needs it; importable from Node and browser alike; protocol content in the "Message protocol" sections below; clients must not bypass api through ctx | +| Assembly layer | `dsh-host-runtime` | Plugin composition + ApiProxy integration + the web UI plugin mount (in-memory Loader tree over the eight dshClient packages); home of host-level configuration (defaults/persistenceRoot, future user profile) | Which plugins mount and with what defaults is decided only here; shells must not alter the assembly | +| Carrier layer | `dsh-host-webserver` | Web-shape HTTP: static serving + `/api/*`→handler forwarding + SSE write-out + close semantics; plugin bundle endpoint + `__DSH_BOOT__` manifest injection (fed by the web plugin registry) | Web (browser access) only; zero workspace dependencies (the registry arrives by structural injection); Electron does not reuse it | +| Client libraries | `dsh-client-ui-slots` / `dsh-client-web-react` / `dsh-client-ui-primitives` | Slot registry core / ctx↔React glue / pure React atoms | Zero cordis runtime dependency in components; seeded into the loader module table by the shell | +| Client plugins | `dsh-client-connection` / `dsh-client-runtime` / `dsh-client-ui-theme` / `dsh-client-i18n` / `dsh-client-ui-layout` / `dsh-client-ui-sidebar` / `dsh-client-ui-conversation` / `dsh-client-ui-trajectory` | Browser-side cordis plugin tree (wire consumer, core services, theme, i18n, layout, sidebar, conversation, trajectory) — see the web client architecture RFC | Dual entry (node half = empty apply; implementation in `src/client/`); the consumption face goes exclusively through ApiProxy | +| Application shape | `@deepseek-ai/dsh` (apps/cli) + `dsh-frontend` (apps/web, the vite application) | Coarse bin dispatch + one assembly module per shape (web.ts / headless.ts); the vite app is a thin main over the `dsh-client-web` shell surface | Shapes dynamic-import so they never load each other; workspace knowledge like dist location stays in the app | + +#### Naming rule + +Packages under `packages/host/*` and `packages/client/*` **must carry the directory-group prefix in the package name**: host/runtime → `dsh-host-runtime`, client/runtime → `dsh-client-runtime`. The directory name does not repeat the group prefix (host/ already expresses it). The package-name tail therefore ≠ the directory name, so the `dsh-*` wildcard in tsconfig.base.json (which resolves by directory name) misses them — **each package in these two groups needs an explicit paths entry**, including separate entries for the plugin packages' `/client` (and runtime's `/loader`) subpaths so source-level resolution matches the exports map. + +#### How to integrate a new shape (operational checklist) + +1. **Pick a fetch impersonation**: browser same-origin HTTP / in-process `host.handler.fetch` injection / your own transport-aspect subclass (e.g. future Electron IPC, see the "Subclass table" below). +2. **Write an assembly module under `apps/`**: `startHost()` + a client subclass + the shape's private signal/print/exit semantics; a mixture never becomes a package — assembly is written in the app. +3. **Import `dsh-host-webserver` only if you need HTTP carriage**, otherwise zero ports. + +The two existing shapes are the template: `apps/cli/src/web.ts` (startHost + dist location + startWebServer + signal shutdown) and `headless.ts` (startHost + InProcessApiClient isomorphic direct calls, zero HTTP zero ports). ACP-class protocol bridges do not follow this checklist: they expose core to the external ecosystem, mount via `ctx.plugin(front-door plugin)` directly, and wear no fetch. + +## Message protocol + +The sections from here down are the protocol body carried by the front layer (`dsh-host-apiproxy`). The wire has exactly four message kinds (the four quadrants) — the Web carriage in the right column is only an example; swapping the carrier (in-process/IPC) leaves the quadrants unchanged: + +``` + client 发起 server 发起 + request ① ClientRequest ③ ServerRequest + （POST /api/ body） （SSE 帧：session 事件、审批/问答 requested） + response ② ServerResponse ④ ClientResponse + （该 POST 的 HTTP 应答体） （POST /api/respond body，回填 ③ 的 rpcId） +``` + +### Wire full forms: a four-member named discriminated union (`api/rpc.ts`) + +| Type | Discriminant tag | Fields | rpcId ownership | Web carriage | +|---|---|---|---|---| +| `ClientRequest` | `'client-request'` | `rpcId` `method` `payload` | client mints | `POST /api/` body | +| `ServerResponse` | `'server-response'` | `rpcId` `result` | echoes ① | that POST's response body (always HTTP 200) | +| `ServerRequest` | `'server-request'` | `rpcId` `method` `payload` | server mints | SSE `data:` line | +| `ClientResponse` | `'client-response'` | `rpcId` `result` | echoes ③ | `POST /api/respond` body | + +`RpcMessage = ClientRequest | ServerResponse | ServerRequest | ClientResponse`, narrowed via `switch (message.type)`. + +**rpcId discipline** (`RpcId` is a branded string with constructor `RpcId()`): + +- Whoever initiates mints; a response always echoes the corresponding request's rpcId and **never mints a new id**. +- server-requests split into two kinds, distinguished statically by `method` (= the frame type), with **no third kind**: answerable frames (`approval/requested`, `question/requested`) carry a stable logical request id (minted once on acceptance, reused verbatim on baseline replay, echoed by the client's answer); pure-push frames (`session/event` etc.) carry an rpcId identifying that one push (freshly minted each time). +- Business code never mints: unary minting funnels into the client base class `callUnary`, frame minting funnels into the host side. + +### Signature narrow forms and carrier completion + +Domain interface signatures perceive only the narrow forms: `RpcRequest

= { rpcId, payload }`, `RpcResponse = { rpcId, result: RpcResult }`. The carrier layer completes narrow forms into full forms (adding the `type` tag and `method`); direction is never inferred from the channel. `RpcResult = { ok: true; value } | { ok: false; error: RpcError }` — methods do not throw business errors. + +### RpcReceipt: the carrier receipt + +The HTTP response body of a `ClientResponse` is `RpcReceipt = { accepted: true } | { accepted: false; reason: 'not-pending' | 'bad-response' }` — a carrier-layer receipt, **not** an RpcMessage (a response has no response); late/duplicate answers get `not-pending`, and the logical convergence surface is the `*/resolved` frames. + +## The type system: signatures are the source of truth + +### RpcMethodMap and derived generics (`api/rpc-map.ts`) + +Method parameter/return structures **live only in the interface method signatures**; the map registers the methods themselves; every other position (handler, client, store, tests) references the derived generics — copying literals or introducing flat named types is banned: + +```ts ignore-check +export interface RpcMethodMap { + 'session.list': SessionsApi['list'] // map key 即 wire 路径段 + // …其余方法同形登记，全集见 api/rpc-map.ts +} +// 派生泛型（穿透窄形取业务类型；实际声明带 K extends keyof RpcMethodMap 约束） +export type RequestPayload = Parameters[0]['payload'] +export type ResponseValue = + Awaited> extends RpcResponse ? T : never +``` + +Stream methods (`events.mux`/`events.host`) stay out of the map (not unary); `respond` stays out of the map (it is a client-response, not a method call). + +### The error model (`RpcErrorDetailsMap`) + +One example row of an error code: + +| code | details | when | +|---|---|---| +| `bad-request` | `{ issues: ZodIssue[] }` | wire/payload zod validation failed | + +The full code set is `RpcErrorDetailsMap` in `api/rpc.ts`. `RpcError` is the distributive union expanded from the map: `code` discriminates, `details` narrows automatically after a `switch`; **details is required** — a new code = one map row + one error-schema branch, and omission is a compile error. Transport failures (network down, host not up) are thrown by the carrier as exceptions; the two layers never mix. + +### Bidirectional zod validation and anchoring + +- **Two-level parse**: the full-form schema once (type/rpcId/method structure + the handler checking path==method) → the business payload dispatched by method/frame type for a second parse; rejection = `bad-request`. +- **Anchoring**: schemas uniformly `satisfies z.ZodType>` (`api/rpc.schema.ts`). `Wire` is a deep "| undefined" widening — the repo enables `exactOptionalPropertyTypes` while zod `.optional()` outputs `T | undefined`, so anchoring the original type is unusable across the board; on the JSON wire, absence and undefined are indistinguishable, so the widening loses no validation semantics. Passthrough wide branches (`SessionEvent`/`ContentBlock`/frame unions/`RpcError`) and brand-id schemas use explicit casts with comments. +- Brand casts have one point each: every schema file funnels its id cast into one place (`rpcIdSchema` is the only cast point in rpc.schema.ts). + +## The contract face (ApiProxy) + +The root interface is `ApiProxy = { sessions, host, events, respond }` (`api/index.ts`). A new client-request domain = one new file pair (`.ts` + `.schema.ts`) + one root-interface field + one map row. + +### The unary method table + +One example row (the table structure is the reading key): + +| method key | request payload | return value | semantics | +|---|---|---|---| +| `session.list` | `{ cursor?: string }` (cursor is a reserved seat, unimplemented) | `{ items: SessionSummary[] }` | persisted sessions, updatedAt descending; v1 builds no index | + +The remaining methods (`session.create`/`session.history`/`session.prompt`/`session.cancel`/`host.describe`) are not re-copied here — signatures are the source of truth; see `api/sessions.ts`, `api/host.ts`, and `RpcMethodMap`. + +### Frames (server→client, named unions) + +Two SSE streams: the mux stream (`GET /api/events.mux`, all-session aggregate) and the host stream (`GET /api/events.host`, host-level events). One example frame row: + +| frame type | payload | when | +|---|---|---| +| `session/event` | `{ sessionId; event: SessionEvent }` | core passthrough: core events pass verbatim, `assistant/chunk` IS the token stream, no separate delta frame | + +The remaining frame types are not re-copied here; the full unions are `MuxFrame`/`HostFrame` in `api/events.ts`. Three semantic points to know: `session/subscribed` carries lastSeq for history seam-race detection; the `approval/question` requested frames are answerable (stable rpcId) and the resolved frames are the convergence surface; `host/agent-error` is the only outlet for live failures with no turn position. + +**Passthrough discipline**: events/messages/content blocks on the wire ARE the core types (`SessionEvent`/`ContentBlock`) — no second DTO set; types reach the browser through the `import type` dependency chain. `SessionEventMap` is merge-extensible: the client applies its documented default (ignore) to unknown types, and the event schema keeps a "valid envelope + unknown type" branch — the envelope stays strict; this is not field-level passthrough. + +### Session semantics (impl-side commitments) + +- **History = event replay**: one fold (client side); history pagination and live increments share one code path; the server maintains no second materialized-snapshot system. History **page boundaries align to message boundaries** (never cut mid-message; chunks group with their finalized message), and the tail page includes the in-flight partial's chunks. +- **Prompt correlation**: the prompt's rpcId rides MessageSource (`'user-rpc'`) into the `user/message` event; the client uses it to promote the optimistic echo. +- **Reconnect = rebuild**: no resume cursor (`mux`'s `since` signature is a reserved seat, ignored if passed); on disconnect reopen the stream + refetch history; compare `subscribed.lastSeq` with the history tail seq and backfill once if there is a seam. +- **Cold sessions resume implicitly**: when `history`/`prompt` hits an unattached session the impl auto-resumes, deduplicating concurrent triggers with an in-flight table; attachment status is not exposed to clients (`running` already covers it). +- **Approvals/questions**: the requested frame mints a stable rpcId on acceptance; first answer wins, and the host's in-memory pending table (keyed by rpcId) is the only referee; after a mux reopen, still-pending requested frames replay after the subscribed frame (rpcId reused verbatim — refresh recovery). The audit events `approval/asked`/`decided` continue through the durable log — frames = the live control plane, events = the durable audit. **Status**: the contract and frame types are shipped; the host-side pending table/wire answerer is unimplemented (`respond` in `api-proxy.ts` is a stub, always `not-pending`); PendingCard v1 is display-only. +- **No protocol version**: client and host release bound together; `host.describe` has no protocolVersion field; introduce one when an independently released client appears. +- **Reserved-seam discipline**: the map holds only implemented methods; an unknown method fails loud at envelope parse (`bad-request`) — no not-implemented fallback code. The reservation list (implementing = copy the signature into the domain interface + add the map row + add the schema pair): `session.fork`, `prompt.mode` gaining `'inject'`, `task.list`, `host.listModels`, describe gaining `hostInstanceId`. + +## The client carrier: the AbstractApiClient class family (`fetch/client.ts`) + +**Protocol invariants live in the base class; platform differences are two aspects**: the abstract method `doFetch(url, init)` (transport) + the overridable `onEnvelope` (observation). + +### IApiClient: the caller view + +The same domain tree as `ApiProxy`, but unary methods **take the business payload directly** — the carrier mints the rpcId and wraps the envelope; business code never mints, and code needing this call's rpcId reads it from the returned `RpcResponse` echo. `ApiProxy` is the narrow-form signature contract the impl side implements; `IApiClient` is the payload-direct view clients consume; `AbstractApiClient` bridges the two. Methods derive per key from `RpcMethodMap` — a map row addition updates them mechanically. + +### Protocol paths held by the base class + +| Path | Content | +|---|---| +| `callUnary` | mint → tap → POST full form → `serverResponseSchema` parse → **rpcId echo check** (mismatch throws) → tap → emit narrow form | +| `readSse` | streaming fetch (not EventSource), `\n\n` framing, `data:` concatenation, ServerRequest full-form parse, tap, emit narrow `RpcRequest` | +| `respond` | client-response passthrough (rpcId is an echo — never minted here); response body parsed by `rpcReceiptSchema` | +| unary timeout | `AbortSignal.timeout` (default 30s, constructor-tunable); streams have no timeout (long-lived by nature) | +| `resolveBase` | browser = same-origin origin; no-location environment (Node) = the `http://dsh.internal` fake authority | + +### The instance-level envelope observation aspect + +All four quadrant full forms pass through `onEnvelope`; the base implementation is an **instance-owned microtask-batched buffer** (frame storms must not disturb consumers per frame; module-level state would leak across instances/tests, hence instance-owned). Observers subscribe via `subscribeEnvelopes(listener)` (receiving whole batches as `readonly RpcMessage[]`, returning an unsubscribe function); a listener throw is isolated (observation must never bite the carrier). With no subscribers the buffering costs nothing. No shipped consumer subscribes today — the aspect is the designated seat for wire diagnostics (the retired RPC debug panel was its first consumer, and a future one plugs in without touching the carrier). + +### The subclass table (transport carriage) + +| Subclass | Package | doFetch | Purpose | +|---|---|---|---| +| `InProcessApiClient` | apiproxy itself | the injected `{ fetch }` handler | **The isomorphic point**: `new InProcessApiClient(toFetchHandler(api))` never touches the network yet runs the real wire serialization/zod/SSE framing — `dsh -p` headless is the protocol's second real consumer | +| `WebApiClient` | dsh-client-connection | `globalThis.fetch` (same-origin `/api/*`) | the browser shape; HTTP+SSE carriage details in the web client architecture RFC | +| `FixtureApiClient` | dsh-client-connection | unused (protocol-layer override) | serverless UI development (`?fixture`): overrides the `callUnary`/`openMux`/`openHost`/`respond` virtuals and is itself the fake server (frame rpcIds minted by it, semantics self-consistent) | +| (future) IPC bridge subclass | apps/electron | IPC serialization round trip | swaps only doFetch; contract and base class unchanged | + +## How to extend (operational checklists) + +**Add a unary method (5 steps)**: ① add the method signature to the domain interface (parameters/return inline — this is the single source of truth); ② add one `RpcMethodMap` row; ③ add the request/value schema pair in `.schema.ts` (anchored `Wire>`); ④ add one handler `UNARY_ROUTES` row (the handler's Web carriage is in the web client architecture RFC); ⑤ implement in the impl (echo `request.rpcId`). On the client side, add the passthrough row to the `IApiClient`/`AbstractApiClient` domain method tables. + +**Add a frame type (3 steps)**: ① add a branch to the `MuxFrame`/`HostFrame` union (answerable frames must note the stable-rpcId semantics); ② add a frame-schema branch; ③ the consumers' fold/routing documented-default already covers unknown types — add an explicit branch as needed. + +**Add an error code (2 steps)**: ① add one `RpcErrorDetailsMap` row (details required); ② add one `rpcErrorSchema` discriminatedUnion branch. + +**Plug in a new carrier**: subclass `AbstractApiClient` implementing only `doFetch`; to intercept at the protocol layer (like the fixture), override the `callUnary`/`openMux`/`openHost` virtuals instead. Contract and base class stay unchanged. + +**Promote a reserved seam**: copy the reserved signature into the domain interface → add the map row → add the schema pair → add the UNARY_ROUTES row → implement. + +## Consequences + +Every client shape consumes one contract: adding a unary method is a five-step mechanical change radiating from a single signature, swapping a carrier touches only a `doFetch` subclass, and every wire message is zod-validated, observable through the envelope tap, and reconcilable by rpcId. The accepted costs: two groups of packages need explicit tsconfig paths entries, and the reserved seams (fork/inject/task.list/listModels/hostInstanceId) stay dormant until a real consumer arrives. + +## Alternatives considered + +| Rejected | One-line reason | +|---|---| +| Packaging by "product shape" (a web family, an electron family) | What shapes share is host/client capability, not the shape itself; capability-provider layering means a new shape needs zero new packages | +| A package per mixture (e.g. a standalone headless package) | A mixture has exactly one consumer (its own app); packaging it is ownerless abstraction, while assembly in the app is readable and disposable | +| Consuming clients connecting to ctx directly (skipping the apiproxy layer) | A second command plane bypasses the contract, losing wire validation/observability/multi-client consistency; ctx keeps exactly two formal uses — front doors and headless event subscription | +| webserver depending on runtime (saving the handler injection) | Structural-typing injection keeps webserver reusable by sidecars/tests with zero workspace deps; a package dependency would drag assembly knowledge into the carrier layer | +| Package names without the group prefix (continuing dsh-) | `dsh-runtime`/`dsh-web-ui` lose their belonging in the flat npm namespace; the cost is one explicit paths entry per package | +| Reusing the in-repo JSON-RPC 2.0 (dsh-jsonrpc) | Numeric error codes degrade to a single fallback code, contracts get aligned by hand in two copies, and naming drifts without a convention | +| A three-envelope model (Request/Response/Frame envelopes, signatures direction-blind) | rpcId correlation is logical-layer; frame and response direction semantics inferred from the channel break the moment the carrier changes | +| Named Request/Response type pairs as the source of truth (map registering type pairs) | Flat named types are a second name for the same fact; signature inference makes adding a method a one-place change | +| REST-style paths | The consumer is our own client with no third-party REST expectations; RPC mapping straight onto the method table is more mechanical | +| A DTO layer (a second wire-only structure set) | Core types reach the browser type-only at zero cost; a DTO is a permanent two-way synchronization tax | +| Cursor resumption (implementing mux since) | Reconnect = rebuild (opencode-style) covers all v1 needs; the signature keeps the seat, implementation waits for a real consumer | +| A createApiClient factory function (the original implementation) | Platform differences (transport/observation) are inheritance aspects, not parameters; the class family lets the fixture substitute at the protocol layer instead of wrapping a fake envelope | diff --git a/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.zh.md b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.zh.md new file mode 100644 index 0000000000..0c256b60ce --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-gui-layering-and-rpc-protocol.zh.md @@ -0,0 +1,251 @@ +# RFC: GUI 分层与 RPC 协议——host/client 按能力支持方分层、四象限消息模型与 fetch 载体 + +Status: implemented + +[English](2026-07-19-gui-layering-and-rpc-protocol.md) | 中文 + +> 分工线：本篇 = 分层模型 + 通道无关的 RPC 协议；协议的 Web 实现（HTTP+SSE）见 [Web 客户端架构 RFC](2026-07-19-gui-web-client-architecture.md)。 + +## Problem + +需要提供 UI 对接层，除已有 ACP/stdio基础版本外，还需要 Web(server) 、 Electron 、等其他产品 UI 形态。我们把这些形态统一称为 Client。希望有如下能力支持: +- 以 `dsh` 进程，同时支持 `dsh web`(启动) 和 `dsh -p`(headless) ，一个进程两种模式（设计预留） +- 以与 `dsh web` 同构的 Web 技术形态，在 Electron 中启动 + +那么当前的工程代码需要稳定的分层职责模型，便于以后接入各类 client 形态。 + +同时各消费端的物理通道不同（HTTP/SSE、进程内直调、将来 IPC），还需要一个通道无关的消息模型和单一契约事实源，让「加一个方法」「换一种载体」互不牵连，且 wire 上的每条消息可类型校验、可观测、可对账。 + +## Decision + +### 分层 + +目录按照如下分层: +- `packages/host/*`: 包只提供 Host 侧能力（代表了以现在 Harness 实体插件系统为主体的 Node.js 代码核心工程），除此之外，还包含 + - 统一后端协议（fetch、HTTP、流式接口等）定义和支持，见本篇「消息协议」起各节 +- `packages/client/*`：包只提供 Client 侧能力，每包单边不混。这里住两类包： + - **纯库**（`ui-slots`、`web-react`、`ui-primitives`）：普通根入口包，静态打包进壳，并播种进浏览器插件 loader 的模块表。 + - **dshClient 插件包**（`connection`、`runtime`、`ui-theme`、`i18n`、`ui-layout`、`ui-sidebar`、`ui-conversation`、`ui-trajectory`）：双入口——根入口是 node 半边（空 `apply`，其存在是为了让 host Loader 管辖生命周期、让 web 插件注册表发现 package.json 的 `dshClient` 声明）；实现与类型全部住在 `src/client/` 下，经 `./client` 子路径发布（tsdown 闭包工厂 bundle），跨包消费一律 import `/client` 形式。`runtime` 额外导出 `./loader`（壳持有的浏览器 bundle loader——loader 加载不了自己）。 +- `apps/` 作为对外导出的应用形态入口，可以由 Client / Host 混合组装。 + - `apps/web`（`dsh-frontend`）是 vite 应用：`dsh-client-web` 导出的壳表面之上的一层薄 `main.ts`。 + - `apps/cli`（`@deepseek-ai/dsh`）做形态分发：`dsh web` = startHost + webserver + 构建出的 `dsh-frontend` dist；`dsh -p` = headless 进程内直调，零 HTTP。 + - 将来的 Electron 形态经由 IPC fetch 载体复用同一套 web client 包。 + +``` +apps/* (application shapes: apps/web = vite app, apps/cli = bin dispatch) + │ consume + ▼ +packages/host/* packages/client/* + apiproxy front layer: protocol pure libs: ui-slots / web-react / ui-primitives + runtime assembly / host entity dshClient plugins ×8 (node half = empty apply, + webserver web-shape HTTP carriage client half = src/client/) + │ ctx.plugin(...) ▲ import only apiproxy's /api /client subpaths + ▼ │ (type-only + the client base class) +harness core packages ──────────────────┘ (types reach the browser via import type) +``` + +方向纪律（每条都由包 deps 可核）： + +- `runtime → apiproxy` 单向；apiproxy 仅依赖类型定义。 +- client 侧包**永不 import** host 侧包的运行时（只吃 `/api`、`/client` 两个浏览器安全子路径）。 +- `webserver` 不依赖 `runtime`：它提供 `{ fetch }` 特定实现 ——「webserver ← runtime」只是运行时注入关系，不是包依赖。 +- client 侧跨包 import 插件包一律走 `/client` 子路径（裸包名会把第二份运行时实例内联进浏览器 bundle；tsdown 纯度门禁会改写或拒收）。 + +TypeScript 以**两个聚合 program** 检查（`tsconfig.json` = host 侧 + 测试，排除 `packages/client`；`tsconfig.client.json` = client 各包及其测试）：两侧在相同键（`sessions`、`loader`）下以不同服务合并 cordis `Context` 接口，单一 program 会同时看到两份声明合并而报冲突。共享叶子包（session/llm/tools/apiproxy 等）只构建一次，由两个 program 共同引用。 + +协议侧：TS interface（`packages/host/apiproxy/src/api/`，零 Node 依赖，浏览器可 import）；wire 消息统一为**双向模型**——每条逻辑消息由「谁发起 × request/response」定形（两轴四格，后文称四象限），与物理通道解耦；客户端统一继承 `AbstractApiClient`（协议不变量全在基类，平台差异只是 `doFetch` 传输切面）。 + +#### 分层角色 + +| 层 | 包 | 职责 | 关键纪律 | +|---|---|---|---| +| 前置层 | `dsh-host-apiproxy` | TS/zod 定义 (api/)+ fetch 抽象 (fetch/：handler + 客户端基类) | 做简单、所有接入方都要；Node/浏览器皆可 import；协议内容见下文「消息协议」起各节；client 不得经 ctx 绕开 api | +| 装配层 | `dsh-host-runtime` | 插件组合 + ApiProxy 集成 + web UI 插件挂载（覆盖八个 dshClient 包的内存 Loader 树）；host 级配置归属地（defaults/persistenceRoot，将来用户 profile） | 装什么插件、给什么默认值只在这里定；壳不得改装配 | +| 承载层 | `dsh-host-webserver` | Web 形态 HTTP：静态服务 + `/api/*`→handler 转发 + SSE 写出 + close 语义；插件 bundle 端点 + `__DSH_BOOT__` manifest（元数据清单）注入（由 web 插件注册表供给） | Web（浏览器访问）专用；零 workspace 依赖（注册表经结构注入到达）；Electron 不复用它 | +| client 库 | `dsh-client-ui-slots` / `dsh-client-web-react` / `dsh-client-ui-primitives` | slot 注册表核心 / ctx↔React 胶合 / 纯 React 原子组件 | 组件零 cordis 运行时依赖；由壳播种进 loader 模块表 | +| client 插件 | `dsh-client-connection` / `dsh-client-runtime` / `dsh-client-ui-theme` / `dsh-client-i18n` / `dsh-client-ui-layout` / `dsh-client-ui-sidebar` / `dsh-client-ui-conversation` / `dsh-client-ui-trajectory` | 浏览器侧 cordis 插件树（wire 消费者、核心服务、主题、i18n、布局、侧栏、对话、轨迹）——见 Web 客户端架构 RFC | 双入口（node 半边=空 apply；实现在 `src/client/`）；消费面唯一经 ApiProxy | +| 应用态 | `@deepseek-ai/dsh`（apps/cli）+ `dsh-frontend`（apps/web，vite 应用） | bin 粗分发 + 每形态一个拼装模块（web.ts / headless.ts）；vite 应用是 `dsh-client-web` 壳表面之上的薄 main | 形态间动态 import 互不加载；dist 定位等 workspace 知识留在 app | + +#### 命名规则 + +`packages/host/*` 与 `packages/client/*` 下的包名**必须含目录组前缀**：host/runtime → `dsh-host-runtime`、client/runtime → `dsh-client-runtime`。目录名不重复组前缀（host/ 已表达）。因此包名尾段 ≠ 目录名，tsconfig.base.json 的 `dsh-*` 通配（按目录名解析）命不中——**这两组的每包需显式 paths 条目**，且插件包的 `/client`（以及 runtime 的 `/loader`）子路径要单列条目，使源码级解析与 exports map 一致。 + +#### 怎么接入一个新形态（操作清单） + +1. **选 fetch 伪造方式**：浏览器同源 HTTP / 进程内 `host.handler.fetch` 注入 / 自写传输切面子类（如将来 Electron IPC，见下文「子类表」）。 +2. **在 `apps/` 下写拼装模块**：`startHost()` + 客户端子类 + 该形态私有的信号/打印/退出语义；混合体不建包，拼装写在 app 里。 +3. **需要 HTTP 承载才 import `dsh-host-webserver`**，否则零端口。 + +现有两形态即模板：`apps/cli/src/web.ts`（startHost + dist 定位 + startWebServer + 信号停机）与 `headless.ts`（startHost + InProcessApiClient 同构直调，零 HTTP 零端口）。ACP 类协议桥不走本清单：它把 core 暴露给外部生态，直接 `ctx.plugin(前门插件)` 挂载、不套 fetch。 + +## 消息协议 + +以下各节是前置层（`dsh-host-apiproxy`）承载的协议本体。wire 上只有四种消息（四象限）——右列的 Web 承载只是示例，换载体（进程内/IPC）时四象限不变： + +``` + client 发起 server 发起 + request ① ClientRequest ③ ServerRequest + （POST /api/ body） （SSE 帧：session 事件、审批/问答 requested） + response ② ServerResponse ④ ClientResponse + （该 POST 的 HTTP 应答体） （POST /api/respond body，回填 ③ 的 rpcId） +``` + +### wire 全形：四具名判别 union（`api/rpc.ts`） + +| 类型 | 判别 tag | 字段 | rpcId 归属 | Web 承载 | +|---|---|---|---|---| +| `ClientRequest` | `'client-request'` | `rpcId` `method` `payload` | client mint | `POST /api/` body | +| `ServerResponse` | `'server-response'` | `rpcId` `result` | 回填 ① | 该 POST 的应答体（恒 HTTP 200） | +| `ServerRequest` | `'server-request'` | `rpcId` `method` `payload` | server mint | SSE `data:` 行 | +| `ClientResponse` | `'client-response'` | `rpcId` `result` | 回填 ③ | `POST /api/respond` body | + +`RpcMessage = ClientRequest | ServerResponse | ServerRequest | ClientResponse`，`switch (message.type)` 窄化。 + +**rpcId 纪律**（`RpcId` 是 branded string，构造函数 `RpcId()`）： + +- 谁发起谁 mint；应答一律回填对应 request 的 rpcId，**绝不 mint 新 id**。 +- server-request 分两类，静态按 `method`（=帧 type）区分，**不设第三种 kind**：可应答帧（`approval/requested`、`question/requested`）的 rpcId 是稳定逻辑请求 id（受理时 mint 一次、基线重放原样复用、client 以它回填应答）；纯推送帧（`session/event` 等）的 rpcId 标识该次推送（每次新 mint）。 +- 业务代码不 mint：unary 的 mint 收口在客户端基类 `callUnary`，帧的 mint 收口在 host 侧。 + +### 签名窄形与载体补全 + +域接口签名只感知窄形：`RpcRequest

= { rpcId, payload }`、`RpcResponse = { rpcId, result: RpcResult }`。载体层把窄形补全为全形（补 `type` tag 与 `method`），方向不靠通道推断。`RpcResult = { ok: true; value } | { ok: false; error: RpcError }`——方法不 throw 业务错误。 + +### RpcReceipt：载体回执 + +`ClientResponse` 的 HTTP 应答体是 `RpcReceipt = { accepted: true } | { accepted: false; reason: 'not-pending' | 'bad-response' }`——载体层回执，**不是** RpcMessage（response 不再有 response）；迟到/重复应答收 `not-pending`，逻辑收敛面是 `*/resolved` 帧。 + +## 类型体系：函数签名即事实源 + +### RpcMethodMap 与派生泛型（`api/rpc-map.ts`） + +方法的参数/返回结构**只住在接口方法签名里**；map 登记方法本身；其余一切位置（handler、client、store、测试）引用派生泛型，禁止复写字面量或另起平铺具名类型： + +```ts ignore-check +export interface RpcMethodMap { + 'session.list': SessionsApi['list'] // map key 即 wire 路径段 + // …其余方法同形登记，全集见 api/rpc-map.ts +} +// 派生泛型（穿透窄形取业务类型；实际声明带 K extends keyof RpcMethodMap 约束） +export type RequestPayload = Parameters[0]['payload'] +export type ResponseValue = + Awaited> extends RpcResponse ? T : never +``` + +流方法（`events.mux`/`events.host`）不进 map（不是 unary）；`respond` 不进 map（是 client-response 不是方法调用）。 + +### 错误模型（`RpcErrorDetailsMap`） + +错误码示例一行： + +| code | details | 何时 | +|---|---|---| +| `bad-request` | `{ issues: ZodIssue[] }` | wire/payload zod 校验失败 | + +码全集见 `api/rpc.ts` 的 `RpcErrorDetailsMap`。`RpcError` 是 map 展开的分布式 union：`code` 判别、`switch` 后 `details` 自动窄化；**details 必填**——新码=map 加一行+错误 schema 加一支，漏填是编译错误。transport 故障（断网、host 没起）由载体抛异常，与业务错误两层不混。 + +### zod 双向校验与锚定 + +- **两级 parse**：全形 schema 一次（type/rpcId/method 结构 + handler 校验 path==method）→ 业务 payload 按 method/帧型分派二次 parse；拒收 = `bad-request`。 +- **锚定**：schema 统一 `satisfies z.ZodType>`（`api/rpc.schema.ts`）。`Wire` 是深度「| undefined」宽化——仓库开 `exactOptionalPropertyTypes` 而 zod `.optional()` 输出 `T | undefined`，直接锚原类型全线不可用；JSON wire 上缺席与 undefined 同形，宽化不损失校验语义。透传宽分支（`SessionEvent`/`ContentBlock`/帧 union/`RpcError`）与 brand id schema 用显式 cast + 注释。 +- brand cast 单点：每个 schema 文件的 id cast 收口一处（`rpcIdSchema` 是 rpc.schema.ts 唯一 cast 点）。 + +## 契约面（ApiProxy） + +根接口 `ApiProxy = { sessions, host, events, respond }`（`api/index.ts`）。新 client-request 域 = 新的一对文件（`<域>.ts` + `<域>.schema.ts`）+ 根接口一个字段 + map 加行。 + +### unary 方法表 + +方法示例一行（表结构即读法）： + +| method key | 请求 payload | 返回 value | 语义 | +|---|---|---|---| +| `session.list` | `{ cursor?: string }`（cursor 留座不实现） | `{ items: SessionSummary[] }` | 已持久化 session，updatedAt 倒序；v1 不建索引 | + +其余方法（`session.create`/`session.history`/`session.prompt`/`session.cancel`/`host.describe`）的参数与返回不在此复写——签名即事实源，见 `api/sessions.ts`、`api/host.ts` 与 `RpcMethodMap`。 + +### 帧（server→client，具名 union） + +两条 SSE 流：mux 流（`GET /api/events.mux`，全 session 聚合）与 host 流（`GET /api/events.host`，host 级事件）。帧示例一行： + +| 帧 type | 载荷 | 何时发 | +|---|---|---| +| `session/event` | `{ sessionId; event: SessionEvent }` | 核心透传：core 事件原样过，`assistant/chunk` 即 token 流，无独立 delta 帧 | + +其余帧型不在此复写，union 全集见 `api/events.ts` 的 `MuxFrame`/`HostFrame`。语义上须知三点：`session/subscribed` 的 lastSeq 供 history 补缝竞态检测；`approval/question` 的 requested 帧可应答（rpcId 稳定）、resolved 帧是收敛面；`host/agent-error` 是无 turn 位置 live 失败的唯一出口。 + +**透传纪律**：wire 上的事件/消息/内容块就是 core 类型（`SessionEvent`/`ContentBlock`），不造第二套 DTO；类型经 `import type` 依赖链直达浏览器。`SessionEventMap` merge-extensible：client 对未知 type documented-default（忽略），事件 schema 留「合法信封+未知类型」分支——信封仍严格，不是字段级 passthrough。 + +### 会话语义（impl 侧承诺） + +- **历史 = 事件重放**：一套 fold（client 侧），历史分页与 live 增量同一条代码路径；server 不做物化快照第二套。history **页边界对齐消息边界**（绝不从消息中间截断；chunk 随定稿消息归组），尾页含进行中 partial 的 chunk。 +- **prompt 关联**：prompt 的 rpcId 经 MessageSource（`'user-rpc'`）透传进 `user/message` 事件，client 以此把乐观回显转正。 +- **重连 = 重建**：不做续传 cursor（`mux` 的 `since` 签名留座、传了忽略）；断线重开流 + 重拉 history；`subscribed.lastSeq` 与 history 尾 seq 比对，有缝再补拉一次。 +- **冷 session 隐式 resume**：`history`/`prompt` 命中未 attach 的 session 时 impl 自动 resume，并发触发用在途表去重；attach 与否不对客暴露（`running` 已覆盖）。 +- **审批/问答**：requested 帧受理时 mint 稳定 rpcId；先到先赢，host 内存 pending 表（keyed by rpcId）是唯一裁判；mux 重开后在 subscribed 帧后重放仍 pending 的 requested 帧（rpcId 原样复用，刷新恢复）。审计事件 `approval/asked`/`decided` 照旧走 durable 日志——帧=live 控制面，事件=durable 审计。**现状**：契约与帧类型已 shipped，host 侧 pending 表/wire answerer 未实现（`api-proxy.ts` 的 `respond` 是 stub，恒回 `not-pending`）；PendingCard v1 只展示。 +- **不设协议版本**：client 与 host 绑定发布，`host.describe` 无 protocolVersion 字段；出现独立发布的 client 时再引入。 +- **预留接缝纪律**：map 只含已实现方法，未知 method 在信封 parse 即 fail loud（`bad-request`），不设 not-implemented 兜底码。预留清单（实现时把签名抄进域接口+map 加行+schema 加对即升格）：`session.fork`、`prompt.mode` 加 `'inject'`、`task.list`、`host.listModels`、describe 加 `hostInstanceId`。 + +## 客户端载体：AbstractApiClient 类体系（`fetch/client.ts`） + +**协议不变量住基类，平台差异是两个切面**：抽象方法 `doFetch(url, init)`（传输）+ 可覆写 `onEnvelope`（观测）。 + +### IApiClient：caller 视图 + +与 `ApiProxy` 同域树，但 unary 方法**收业务 payload 直传**——载体 mint rpcId 并包信封，业务代码永不 mint；需要本次调用 rpcId 的从返回的 `RpcResponse` 回显里读。`ApiProxy` 是 impl 侧实现的窄形签名契约，`IApiClient` 是 client 侧消费的 payload 直传视图，`AbstractApiClient` 桥接两者。方法逐 key 从 `RpcMethodMap` 派生——map 加行即机械更新。 + +### 基类持有的协议路径 + +| 路径 | 内容 | +|---|---| +| `callUnary` | mint → tap → POST 全形 → `serverResponseSchema` parse → **rpcId 回显校验**（不符即 throw）→ tap → 吐窄形 | +| `readSse` | streaming fetch（非 EventSource）、`\n\n` 分帧、`data:` 拼接、ServerRequest 全形 parse、tap、吐窄形 `RpcRequest<帧>` | +| `respond` | client-response 透传（rpcId 是回填，此处不 mint）；应答体 `rpcReceiptSchema` parse | +| unary 超时 | `AbortSignal.timeout`（默认 30s，构造参数可调）；流不设超时（长连接本性） | +| `resolveBase` | 浏览器=同源 origin；无 location 环境（Node）=`http://dsh.internal` 假 authority | + +### 实例级 envelope 观测切面 + +四象限全形均过 `onEnvelope`；基类实现是**实例持有的微任务合批缓冲**（帧风暴不逐帧惊扰消费者；模块级状态会跨实例/测试泄漏，故实例持有）。观测者经 `subscribeEnvelopes(listener)` 订阅（收整批 `readonly RpcMessage[]`，返回退订函数）；listener 抛异常被隔离（观测不得反噬载体）。无订阅者时零缓冲成本。当前没有任何现役消费者订阅——该切面是 wire 诊断的预留位（已退役的 RPC 调试面板是它的首个消费者，将来的诊断消费者接入时不动载体）。 + +### 子类表（传输承载） + +| 子类 | 所在包 | doFetch | 用途 | +|---|---|---|---| +| `InProcessApiClient` | apiproxy 本包 | 注入的 `{ fetch }` handler | **同构点**：`new InProcessApiClient(toFetchHandler(api))` 全程不过网络但真跑 wire 序列化/zod/SSE 帧——`dsh -p` headless 即协议第二真实消费者 | +| `WebApiClient` | dsh-client-connection | `globalThis.fetch`（同源 `/api/*`） | 浏览器形态；HTTP+SSE 承载落地见 Web 客户端架构 RFC | +| `FixtureApiClient` | dsh-client-connection | 不用（协议层覆写） | 无 server 的 UI 开发（`?fixture`）：覆写 `callUnary`/`openMux`/`openHost`/`respond` 虚方法，自己就是假 server（帧 rpcId 由它 mint，语义自洽） | +| （将来）IPC 桥子类 | apps/electron | IPC 序列化往返 | 仅换 doFetch，契约/基类零改 | + +## 怎么扩展（操作清单） + +**加一个 unary 方法（5 步）**：①域接口加方法签名（参数/返回内联，这是唯一事实源）；②`RpcMethodMap` 加一行；③`<域>.schema.ts` 加 request/value schema 对（锚 `Wire>`）；④handler `UNARY_ROUTES` 加一行（handler 的 Web 承载见 Web 客户端架构 RFC）；⑤impl 实现（回显 `request.rpcId`）。client 侧 `IApiClient`/`AbstractApiClient` 的域方法表同步加一行透传。 + +**加一个帧型（3 步）**：①`MuxFrame`/`HostFrame` union 加一支（可应答帧须注明 rpcId 稳定语义）；②帧 schema 加一支；③消费端 fold/路由的 documented-default 已兜底未知型，按需加显式分支。 + +**加一个错误码（2 步）**：①`RpcErrorDetailsMap` 加一行（details 必填）；②`rpcErrorSchema` discriminatedUnion 加一支。 + +**接一种新载体**：继承 `AbstractApiClient` 只实现 `doFetch`；需要拦截协议层（如 fixture）再覆写 `callUnary`/`openMux`/`openHost` 虚方法。契约与基类零改。 + +**升格一个预留接缝**：把预留签名抄进域接口 → map 加行 → schema 加对 → UNARY_ROUTES 加行 → impl 实现。 + +## Consequences + +所有 client 形态消费同一契约：加一个 unary 方法是从单一签名辐射的五步机械改动，换载体只动一个 `doFetch` 子类，wire 上每条消息可 zod 校验、可经 envelope tap 观测、可按 rpcId 对账。接受的代价：两组包需要显式 tsconfig paths 条目；预留接缝（fork/inject/task.list/listModels/hostInstanceId）在真实消费者出现前保持休眠。 + +## Alternatives considered + +| 放弃项 | 一句话理由 | +|---|---| +| 按「产品形态」分包（web 一族、electron 一族） | 形态间共享的是 host/client 两侧能力而非形态本身；能力支持方分层让新形态零新包 | +| 混合体建包（如 headless 独立包） | 混合体只有一个消费者（它自己的 app），建包是无主抽象；拼装写在 app 里可读可弃 | +| 消费型 client 直连 ctx（省 apiproxy 一层） | 第二命令面绕开契约，wire 校验/观测/多端一致性全失；ctx 只留给前门与 headless 事件订阅两个正式用途 | +| webserver 依赖 runtime（省 handler 注入） | 结构 typing 注入让 webserver 可被 sidecar/测试复用且零 workspace 依赖；包依赖会把装配知识拖进承载层 | +| 包名不带组前缀（沿用 dsh-<尾段>） | `dsh-runtime`/`dsh-web-ui` 在扁平 npm 命名空间里失去归属信息；代价只是每包一条显式 paths | +| 复用仓内 JSON-RPC 2.0（dsh-jsonrpc） | 数字错误码退化成单码兜底、契约双份人肉对齐、命名无 convention 自然漂移 | +| 三信封模型（Request/Response/Frame 各一信封，签名不感知方向） | rpcId 是逻辑层关联，帧与应答的方向语义靠通道推断在换载体时即失效 | +| 具名 Request/Response 类型对为事实源（map 登记类型对） | 平铺具名类型是同一事实的第二个名字；签名 infer 反推让加方法只改一处 | +| REST 风格路径 | 消费者是自家 client，无第三方 REST 体验诉求；RPC 直映方法表更机械 | +| DTO 层（wire 专用第二套结构） | core 类型 type-only 直达浏览器零成本；DTO 是永久的双向同步税 | +| cursor 续传（mux since 实装） | 重连=重建（opencode 同款）覆盖 v1 全部需求；签名留座，实装等真实消费者 | +| createApiClient 工厂函数（原实现） | 平台差异（传输/观测）是继承切面不是参数；类体系让 fixture 在协议层替换而不是包一层假信封 | diff --git a/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.i18n.yaml b/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.i18n.yaml new file mode 100644 index 0000000000..91abd32a3e --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.i18n.yaml @@ -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 +2026-07-19-gui-web-client-architecture.md: 58320570f752d4259004172d3b4527172c2cc646 +2026-07-19-gui-web-client-architecture.zh.md: 744fdaa4b89a01e2710f85b177228713189e3025 diff --git a/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.md b/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.md new file mode 100644 index 0000000000..58320570f7 --- /dev/null +++ b/.agents/notes/implemented/architecture/2026-07-19-gui-web-client-architecture.md @@ -0,0 +1,148 @@ +# Agent Note: Web client architecture — the client cordis plugin tree, the slot system, and the React-free object layer + +Status: implemented + +English | [中文](2026-07-19-gui-web-client-architecture.zh.md) + +> Division of labor: the channel-independent layering model and RPC protocol (message model / type system / contract face / client base class) are in the [layering and RPC protocol RFC](2026-07-19-gui-layering-and-rpc-protocol.md); this document = the browser side: how the client cordis tree loads, how UI plugins compose through slots and services, and how the React-free object layer feeds React through immutable snapshots. + +## Problem + +Two forces shape the browser client. First, streaming: in an event-driven conversation UI, if business state (the event window, streaming accumulation, pending interactions, the connection state machine) scatters across React components and a global store, every token chunk shakes the render tree, and swapping the UI library means rewriting the business logic. Second, modularity: UI features (layout, sidebar, conversation, theme, locale) must be independently loadable plugins — composed at runtime from a host-served manifest, not compiled into one bundle — without giving up compile-time type safety across plugin boundaries. + +## Decision + +Both ends run cordis. The host is a cordis plugin tree; the browser runs a second, client-side cordis tree whose every UI capability is a plugin loaded dynamically by a shell-held loader. Inside that tree, cordis ctx hosts all runtime facts (services, stores, session scopes) and React is pure projection: components import nothing from the framework, receive everything through props, and subscribe to immutable snapshots via `useSyncExternalStore` (uSES below). + +``` +┌─ Host ─────────────────────────┐ ┌─ Browser ─────────────────────────────────────────┐ +│ sessions/agents/SessionLog │ │ client cordis root ctx │ +│ apiproxy: RPC + mux/host 双流 │◀─▶│ ├ loader（壳静态持有，不能经自己装载） │ +│ webserver: │ │ ├ immediately 先行组: connection/runtime/ │ +│ ├ GET /plugins//client.js │ │ │ ui-theme/i18n（动态 bundle，并行先装） │ +│ └ GET / 注入 __DSH_BOOT__ │ │ ├ 后续组: layout/sidebar/conversation/trajectory │ +└────────────────────────────────┘ │ └ session scope ×N（观看驱动，惰性建） │ + │ React: loading 页 → settled → 整 UI 一次成型 │ + └────────────────────────────────────────────────────┘ +``` + +## The client cordis tree and the loading chain + +Every UI plugin is simultaneously a host plugin (dual-entry package): the node half sits in the host's plugin tree so the host Loader governs its lifecycle, and the browser half is a tsdown closure bundle under the package's `exports["./client"]`. The host webserver derives the boot manifest from loaded plugins carrying a `dshClient` manifest field and injects it into the page as `window.__DSH_BOOT__` — the HTML alone tells the browser everything to fetch, zero extra round trips. + +The loading chain, end to end: + +1. `GET /` → the shell boots, mounts `ctx.loader` (the loader mechanism is held statically by the shell — a loader cannot load itself; its code home is `packages/client/runtime/src/client/loader/`, imported through the `./loader` subpath so the shell bundle does not swallow the rest of the runtime package), seeds the require module table with the pure-library instances (react, react-dom, cordis, ui-slots, web-react, ui-primitives), and renders a plugin-independent loading page. +2. `loader.start()` reads `__DSH_BOOT__`. Entries flagged `immediately` form the early-load group (connection, runtime, ui-theme, i18n): fetched in parallel, applied in intra-group `inject` topological order, and **the whole group must land before anything else loads**. Remaining plugins then load in inject order. +3. Each bundle executes `window.DSHClientProxy.loadPlugin({ id, factory })`. The loader calls `factory(require)` — bundles are closure factories whose external dependencies arrive through the injected `require`, resolved against the module table (no globals, no import maps; an unresolvable specifier fails loud). The factory returns its module export surface (including the cordis `apply`); the loader runs `ctx.plugin(apply)`, then **registers that export surface into the module table under the package name**, so inject topology guarantees later plugins can `require` earlier ones. Plugin CSS is inlined in the bundle and injected as `