Files
deepseek-harness/packages/core/agent-loop/README.md
T
Tianyi Cui 2489402610 Merge origin/master: scope-aware fusion of the tools/execute seam, session-prefix, and tool-cordis
Master brought 50 commits (the tool-cordis group, dsh-code-runtime + worker,
the tools/execute around-dispatch seam + timeout-policy, repeat-tool-guard,
agent/session-prefix, the ui reorganization). Beyond the ten textual
conflicts, the merge reconciles master's new seams with this branch's
scoped-registration world:

- tools/execute (new waterfall around core dispatch): dispatched with the
  SAME exec.agent carrier as the pre/post waterfalls — an agent.ctx wrapper
  times/retries only its own agent's calls — and its base thunk resolves the
  tool through the caller's visible view (get(exec.name, exec.agent)), so a
  scoped/shadowed tool dispatches and a restricted-away global stays
  UNKNOWN_TOOL. Declared this: Scoped<ToolRegistry> with the scope-filtered
  doc sentence; invariants table + verify-scoped-dispatch pin it (21 events).
- agent/session-prefix (new waterfall, once per loop instance): composed via
  the fused agentEvents dispatcher (scope-filtered like every agent-subject
  event), declared this: Scoped<Agent>, table-pinned. agent/pre-step keeps
  master's new sessionPrefix parameter with this branch's Scoped this.
- timeout-policy reads the budget through the caller's visible view
  (get(exec.name, exec.agent)): a scoped tool's own timeoutMs governs its
  calls; a global name-twin's budget is never misapplied to a shadowing
  per-agent variant.
- tool-cordis: cordis_inspect's tools section lists the CALLING agent's view
  (its description promises "what you can call"); the sandbox tool façade's
  reads resolve through the mount's own scope, mirroring where its register
  lands writes; sandboxRegisterTool's return type carries the exact-disposer
  union honestly. dsh-scope declared as peer+dev with the project reference.
- doc-sync chain unions master's verify-cordis-api with this branch's
  verify-scoped-dispatch; the generated catalogs, event matrix (the
  zero-dispatcher guard passes over master's new events), module graph, and
  the cordis api-catalog are regenerated on the merged surface.

Full gate sequence green on the merged tree: typecheck, lint, per-file 100%
coverage (2668 tests), snapshots (38), doc-sync, module graph, build,
hygiene, demo smoke.
2026-07-09 23:24:42 +08:00

98 lines
8.2 KiB
Markdown

# dsh-agent-loop
THE concrete agent plugin: `ReactLoopAgent` and the loop driver. Implements the `Agent` interface and drives the session/turn/step lifecycle.
This is the only package in the harness that contains concrete loop logic. Everything else is an abstract service or a plugin against extension seams — new behavior goes into plugins, not here.
## Service: `AgentLoop` (ctx key: `agentLoop`)
### Public API
Lifecycle (scoped): the composite creation effect mints the agent's scope (`agent.ctx`), enters the session through it (the session's dispatch carrier), registers the agent, runs `CreateAgentOptions.setup`, emits `agent/session-start`, then starts the loop; teardown runs stop/drain → unregister → detach session → unwind scope, keeping store/registry rollback synchronous on every failure path. All `agent/*` dispatches go through `agentEvents(ctx, agent)`; per-step assembly through `assembleContextFor(agent)`; the turn-end durability checkpoint through `ctx.sessions.flush(session)`.
- `ctx.agentLoop.create(id: string, options?: AgentOptions): ReactLoopAgent` — config-driven create: an agent on a fresh per-run session id `${id}-session-<uuid>` (no cwd). Used for `cordis.yml`-configured agents. The per-run uuid avoids colliding with the on-disk log a prior run materialized once a durable persistence backend is loaded; each run is a new session (a deliberate demo simplification — a real resume-or-create policy is a TODO). Disposed with the calling fiber.
`AgentLoop` also implements the `AgentFactory` seam and registers itself via `ctx.agents.setFactory(this)`, so plugins create/resume agents through `ctx.agents` (the interface):
- `ctx.agents.create({ agentId, sessionId, meta?, seed?, agentOptions? }): AgentHandle` — programmatic create on a caller-supplied `sessionId` (e.g. an ACP-generated id), NOT `${id}-session`; `meta` carries cwd/lineage/seed-boundary metadata and `seed` reconstructs a forked child prefix. Returns an [`AgentHandle`](../agent/README.md) — the owner disposes it to tear down exactly this agent (stop loop + await quiescence + unregister + remove session).
- `ctx.agents.resume({ agentId, resumeSessionId, agentOptions? }): Promise<AgentHandle>` — load a persisted session via `ctx.sessionPersistence` ([session persistence](../../../docs/rfc/implemented/architecture/2026-06-14-session-persistence.md)) and resume an agent on it. The live session id is the resumed id; turn numbering and derived history continue from the loaded log. Requires a session-persistence backend (NOT hard-injected — non-persistent demos still work; `resume` rejects with a clear error when persistence is absent). Returns an `AgentHandle`.
The config-driven `ctx.agentLoop.create()` path keeps its agent owned by the loop fiber (it discards the handle) — only the programmatic factory callers (the ACP bridge and in-process subagent backends) hold a handle and own per-agent teardown.
### Injected services
`agents`, `sessions`, `llm`, `tools`, `systemPrompt` — all five interface services.
### Configuration (schemastery)
```ts
interface Config {
agents: Array<{
id: string // required
model?: string
}>
}
```
Agents listed in config are auto-created at startup. (There is no per-agent persona: the deployment persona is `dsh-system-prompt`'s own `persona` config, shared by every agent in the context.) The plugin registers the built-in `model`/`cwd` prompt variables on `ctx.systemPrompt`, resolved per step from the `assemble({ agent })` context — runtime facts of the agents THIS loop drives, unlike the `harness:identity`/`deployment:persona` sections, which live on `dsh-system-prompt` so they survive a swapped loop plugin.
### Classes
- `ReactLoopAgent` — the concrete `Agent` implementation. Owns the inbox (`Inbox`), the per-step `AbortController`, and the loop driver. Everything observable happens through session events and the `agent/*` event taxonomy.
- `Inbox` — per-agent queued + steering FIFOs (`enqueue`, `steer`, `drainQueued`, `drainSteering`, `waitForQueued`).
### Loop lifecycle (`loop.ts`)
One invocation of `runLoop()` drives one agent for its whole lifetime:
```
create agent → emit agent/session-start(source) ⟵ once, before turn 1
forever:
wait for queued messages (idle)
TURN (error-contained):
'turn/start'
each queued: waterfall agent/prompt-submit → allow (→ session('user/message'),
inject additionalContext) | block (→ session('prompt/blocked'), drop)
if every prompt blocked: 'turn/end'(rejected), no step ⟵ zero-step turn
STEP loop:
drain steering
assembly = systemPrompt.assemble({agent}) ⟵ renderPrompt(assembly) IS the full prompt
prefix ??= waterfall agent/session-prefix ⟵ once per instance (first step): frozen
session prefix; on the header, never history
await serial agent/pre-step(…, prefix) ⟵ surface mutation (compaction) outside the step;
pressure gates see the prefix the request carries
boundary = session.deriveMessages() ⟵ reconstruction boundary: same sync frame,
session('step/start') strictly before step/start
config = waterfall agent/request ⟵ frozen seed; return a replacement to switch
session('request/header'[-delta]) ⟵ the header event this request owes the log
stream llm.stream(freeze({header..., messages: prefix+boundary})) → session('assistant/chunk')
message = waterfall agent/step-result
session('assistant/message')
each tool-call: session('tool/call')
→ tools.execute() [waterfall tools/pre-execute → dispatch → tools/post-execute]
→ session('tool/result')
append buffered post-execute additionalContext as session('context/message')(s)
drain steering → session('steering/message')
cont = waterfall agent/turn-continuation → ContinuationDecision
({action:'continue', reason?} records reason as next-step steering)
if action==stop (and no pending steering): break
session('turn/end')
await session/flush
re-enqueue leftover steering as queued
idle unless more queued
```
Error containment: a throwing plugin ends the **turn**, never the loop. Dispose mid-turn emits `agent/status('disposed')` and ends with reason `disposed`. A step that hits the model's output-token ceiling makes the turn end `max-tokens` (the rule: any `max-tokens` step in the turn surfaces as `max-tokens`; `disposed`/`aborted`/`error` still take precedence) — distinct from a clean `completed` stop.
Cancellation: `agent.cancel()` is the single public stop primitive — it clears the queued + steering FIFOs, aborts the in-flight step, and drives a turn-scoped marker the driver checks at every point a turn could start or continue (right after the idle wait, after the `running` flip, before each step, and at the continuation gate) so a turn about to start is dropped. A cancelled turn ends `aborted`; a queued-but-not-started prompt never runs and cannot be batched into the cancelled turn. The marker is reset once per loop iteration, so a cancel governs exactly one turn and never leaks onto a later prompt. (The loop still aborts its own per-step `AbortController` directly on disposal and from `cancel()`; that controller is loop-internal, not a public verb.)
### What is NOT here
Everything that goes beyond "call the model, run the tools, repeat" belongs to plugins listening on the event taxonomy:
- Hooks: `agent/session-start`, `agent/prompt-submit`, `agent/pre-step`, `agent/request`, `agent/session-prefix`, `agent/step-result`, `tools/pre-execute`, `tools/post-execute`, `agent/turn-continuation`
- Compaction: `agent/pre-step`
- Sandbox, permission, plan mode: `tools/pre-execute` (deny/ask gate), `tools/post-execute`
- Sub-agents: implemented outside the loop as `ctx.subagents` providers; in-process providers use `ctx.agents.create()` and owned `AgentHandle` teardown, while child streaming/progress and background/poll collection remain deferred.
- Persistence: `session/event` + `session/flush`
- UI: `session/event` (assistant token stream, boundaries, tool activity) + `agent/*` control events (`agent/status`, `agent/created`/`agent/disposed`)