Files
deepseek-harness/packages/core/agent-loop
Hypatia May d6da8ca29a fix(compact): decide step-alignment from surface tool-pairing, fire compaction pre-step (CBR-001)
Codex round 1 CBR-001: a head-anchored compaction checkpoint was
mis-classified by the log-position step-alignment scan, so a second
auto-compaction over a checkpoint-headed surface silently failed.

Root cause: `isStepAlignedStart/End` scanned the LOG by seq, but a
`replace` op lands a checkpoint at a high log seq whose SURFACE position
is the head — its log neighbours (the open step's assistant/message) are
not its surface neighbours, so the forward scan wrongly reported mid-step.

Fix, per the agreed direction:
- Replace the two log-position predicates with one surface-anchored
  helper `isToolPairingBalanced(nodes, events, beforeSeq)` in
  `dsh-session` (renamed step-boundary.ts → tool-pairing.ts). A cut is
  balanced when no unanswered tool-call precedes it on the surface; a
  region is collapsible iff both edges are balanced cuts. The open-tail
  and free-node cases fall out of the same counter. It also throws on a
  corrupt surface (a tool/result with no matching call).
- Move compaction off the in-step seam to a new "pre-step" seam fired
  after turn/start and before step/start, so a compaction's log-only
  compact/* records and its replacement node land cleanly OUTSIDE any
  step (the honest structure crash-safety relies on). Renamed the event
  agent/pre-request → agent/pre-step and switched its dispatch from
  parallel → serial (listeners mutate the surface as a side effect;
  serial isolates them so concurrent appends can't interleave). Extended
  the catalog generator to accept @mode serial.

Regression coverage: a real-loop test driving an auto-compaction asserts
the landed checkpoint is a balanced cut on both sides; unit tests pin the
checkpoint case, the mid-step injection case, multi-call steps, and the
corrupt-surface guard. Proven red on the old log-position logic.
2026-06-26 13:51:01 +08:00
..

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

  • 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?, agentOptions? }): AgentHandle — programmatic create on a caller-supplied sessionId (e.g. an ACP-generated id), NOT ${id}-session. Returns an AgentHandle — 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) 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) hold a handle and own per-agent teardown.

Injected services

agents, sessions, llm, tools, systemPrompt — all five interface services.

Configuration (schemastery)

interface Config {
  agents: Array<{
    id: string                 // required
    model?: string
    systemPrompt?: string
  }>
}

Agents listed in config are auto-created at startup.

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:

forever:
  wait for queued messages (idle)
  TURN (error-contained):
    drain queued → 'turn/start' → session('user/message')
    STEP loop:
      drain steering
      assembly = systemPrompt.assemble()
      request = waterfall agent/request
      stream llm.stream(request) → session('assistant/chunk')
      message = waterfall agent/step-result
      session('assistant/message')
      each tool-call: session('tool/call') → tools.execute() → session('tool/result')
      drain steering → session('steering/message')
      cont = waterfall agent/turn-continuation
      if !cont: 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/request, agent/step-result, tools/execute, agent/turn-continuation
  • Compaction: agent/request
  • Sandbox, permission, plan mode: tools/execute
  • Sub-agents: TODO seam on AgentLoop.create()
  • Persistence: session/event + session/flush
  • UI: agent/stream-chunk + agent/* events