Files
deepseek-harness/docs/architecture.md
Tianyi Cui b54381f3e7 fix(agent): commit mutable setup at publication
Agent setup may await while a mutable contribution registry changes. The previous subagent path validated and committed its provisioning batch inside the setup callback. A revocation queued after that callback returned therefore treated the installation as resident and released it, even though AgentLoop had not published the child yet. AgentLoop could then admit and announce a child whose required capability had already disappeared.

Introduce AgentSetupCommit as the optional synchronous result of create and resume setup. AgentLoop now awaits setup, invokes that commit with no intervening asynchronous boundary, and only then enters the Session and Agent registries. A commit failure follows the existing private-transaction rollback, so neither identity is published and the caller can reuse the id.

Keep continuable-subagent installations provisional until this publication commit. Contribution removal still releases every installation immediately, but now marks an unpublished batch invalid so its commit rejects with ACTIVATION_SETUP_REVOKED. Once the commit succeeds, later removal remains ordinary live revocation.

Cover create and resume ordering, resume commit rejection and identity reuse, and an assembled microtask revocation that leaves only the parent Agent and Session. Update the public JSDoc, architecture flow, package contracts, current Agent Notes, Chinese counterparts, pairing records, and generated Cordis API to describe the new boundary.

Validated with the four focused Agent/subagent test files (91 tests), the isolated assembled regression, targeted TypeScript project builds, generated Cordis API freshness, export JSDoc verification, scoped translation pairing, Markdown wrapping, and Mermaid parsing.
2026-08-02 20:09:05 +08:00

19 KiB

DeepSeek Harness Architecture

English | 中文

DeepSeek Harness SDK uses Cordis: everything is a plugin, including the loop.

Overview

Harnesses are Cordis contexts; packages contribute services, typed events, and disposable registrations.

packages/core/ groups the default agent flow; capabilities remain plugins.

Default Services

ctx key Package Role
dsh-scope scoped-context registrations and shared layer storage (library)
ctx.sessions dsh-session in-memory event-sourced sessions
ctx.systemPrompt dsh-system-prompt ordered stable system sections, cache-safe dynamic contexts, tool schemas, and variables
ctx.tools dsh-tools tool registry and execution pipeline
ctx.agents dsh-agent live agents, delegated creation, agent/* events, process-local initiator scope
ctx.agentLoop dsh-agent-loop concrete Agent driver

Capability Services

ctx key Package family Role
ctx.llm llm/ adapter registry, streaming model calls
ctx.tokenMeter llm/token-meter replay-aware request and surface pressure
ctx.bash bash/ foreground/background command execution
ctx.subprocess subprocess/ managed child-process trees for bash, LSP, and ACP subagent backends
ctx.pty pty/ owner-scoped persistent terminal sessions
ctx.sandbox sandbox/ same-world process confinement through argv wrapping and per-call policy
ctx.sandboxPolicy sandbox/ shared sandbox policy home
ctx.codeRuntime code-runtime/ model-written program execution
ctx.fs fs/ filesystem provider primitives and policy events
ctx.lsp lsp/ semantic navigation registry
ctx.skills skill/ skill provider registry, progressive disclosure
ctx.web web/ search/fetch provider registries
ctx.compact, ctx.toolResultPrune compact//compact-tool-result-prune summary compaction, optional model-free result pruning
ctx.subagents subagent/ named delegation providers and Activation-based continuations
ctx.planMode plan/ logged plan collaboration state
ctx.tasks tasks/ background task registry, generic task_* controls
ctx.workflows workflow/ script-driven multi-agent orchestration
ctx.goals goal/ persisted same-session goals
ctx.sessionPersistence session-persistence/ durable session-log storage
ctx.sessionQuery session-query/ live-preferred exact/filter/trace queries over SQLite FTS, workspace-authorized model tools
ctx.sessionTitle session-title/ log-backed fallbacks, one optional asynchronous provider
ctx.settings settings/ per-plugin user-settings namespaces layered over composition entries
ctx.credentials credentials/ named secret references resolved per operation, never inlined in configuration
ctx.directoryPicker host/directory-picker GUI-host directory picking (native/browse interactions)
ctx.typert typert/registry runtime registry for generated package reflection and live Zod schemas
ctx.invariants support/invariants package-name-selected registry of package-owned runtime checks

Event

Events are the service extension API (catalog, producer/consumer map).

Event Domains

  • Session events are durable log facts emitted through session/event.
  • Agent events carry live Agent for status, prompt admission, request shaping, validation, and continuation.
  • Capability events let owning seams attach policy and adapters without a loop import.

Interception Semantics

Waterfalls are around-middleware: listeners delegate with next(); returning without it vetoes or takes over (semantics).

Default Loop Lifecycle

A session is append-only. An ordinary turn claims one queued send() item; injection claims none. A successor awaits its predecessor's checkpoint but may share its running interval (decision). A turn ends when model or plugins stop it; a step is one model request plus tools. Quotes in the sequence below mark durable events.

Creation without an id mints <config-id>-session-<uuid>; sessionId resumes or creates, while resumeSessionId requires history. Resume restores lineage and delegation depth before publication. Setup failures emit agent-loop/config-start-failed; teardown is silent.

Turn Flow

choose declarative identity and fresh/resume path
  -> prepare private session + agent.ctx -> await unpublished setup -> invoke optional synchronous setup commit
  -> enter session + agent -> session/created -> agent/created
  -> enable driving -> agent/session-start(source) -> start driver
forever:
  wait for queued occurrence
  claim (edit/remove end) -> emit agent/status(running) if starting an interval
  open the next-step acceptance window
  -> agent/prompt-submit
    blocked or failed prompt -> close the window without opening a turn
      append a context-only caller batch immediately
      keep steering and context staged beside it pending for a later admitted turn
    allowed prompt:
      'turn/start'
      append prompt + additional contexts as separate 'user/message' events
    STEP loop:
      agent/step
      assemble system prompt and tools
      materialize changed runtime context as sourced 'user/message'
      drain injected context and provisional steering (steering bypasses prompt-submit)
      snapshot the derived messages (the reconstruction boundary)
      'step/start'
      admit the drained steering receipts
      agent/request (config only) -> prepare adapter defaults/provenance + context capacity under turn signal -> log request/header (+ request/context on route change) -> llm/stream (frozen, registration-bound)
      'assistant/chunk'
      'assistant/message'
      schedule tool calls by ctx.tools.executionMode:
        exclusive -> barrier
        parallel -> rolling pool, <= maxParallelToolCalls; reclassify-at-start; scheduler failure -> stop starts, drain dispatches
        start -> 'tool/call' -> ordered tools/pre-execute -> concurrent tools/execute
        model-order result -> ordered tools/post-execute -> 'tool/result'
      drain accepted tool context after all results; keep steering provisional
      'step/end'
      continue for tools or steering unless a result concluded the turn and rejects pending steering
      otherwise agent/turn-stopping -> drain context -> continue only for steering
    close the next-step acceptance window
    'turn/end' -> agent/settled
  start the next waking queued message, or emit agent/status(idle)

idle inject:
  append 'user/message'
  do not open a turn or run the model

Each step assembles ordered stable system sections, cache-safe dynamic contexts, tool schemas, and variables; unknown references fail the turn. dsh-system-prompt owns identity and persona; the loop supplies provider, model, and cwd (prompt ownership).

Admission-time and active-turn inject() stage for the next step; tool-time injection and post-tool additionalContexts settle after results. Steering shares the outbox but remains provisional until a request admits it. steer() returns a message-owned receipt: after agent/step and asynchronous prompt assembly succeed, the loop commits the stable batch, snapshots request history, opens step/start, then resolves its receipts as admitted with the turn and step; later arrivals wait. A turn-concluding tool result, broad cancellation, disposal, or a claimed idle-steering turn that never opens a step rejects affected receipts, while cancel(..., { keepInbox: true }) and non-terminal routing preserve pending delivery. Idle inject() appends immediately without changing turn numbers; persistence drains eagerly.

Pruning precedes summaries; overflow retries require durable progress. agent/request-error may authorize one retry turn between failed-step and turn close; cancellation wins. Adapter-owned retryPolicy makes normal mode bounded; always mode delegates specialized recovery before retrying until success or cancellation (compaction, retry foundation, provider policy).

Failure Boundaries

Adapter failures close their step before agent/request-error receives the exact Error, normalized LlmFailure, and signal. A handled failure closes its turn and opens a retry turn from durable history without an idle notification; exhaustion leaves terminal turn/end. Failed chunks commit neither messages nor tool calls.

Other failures use agent/error. Cancellation and disposal beat recovery. Before request-header commit, the turn signal cancels asynchronous model-capability preparation; undispatched tools get synthetic tool/call/ABORTED_BEFORE_DISPATCH pairs. Effective cancel(cause) emits its cause before queue clearing and abort; observers cannot veto; idle calls emit nothing. Durability records user or parent cancellation as aborted, teardown as disposed; teardown awaits quiescence. The cause affects reporting, not late result-context handling (decision).

Turn and step events are turn-enclosed. Idle user/message and standalone compact/* { turn: null } consume no turn; their lock-time markers may interleave with injection. Reload synthesizes interrupted turn ends; session/end-seed distinguishes stale compaction orphans from live locks. After close, only agent/error reports failures. Each turn has one TurnEndReason.

Agent Handles

ctx.agents owns agents, returning AgentHandle { agent, dispose() }. Plugins use send() or followup(), receipt-bearing steer(), and inject() presets; reserveTurnAdmission() synchronously reserves idle for durable work without changing queued prompt identity. Await a steering receipt when request admission matters; best-effort UI steering may ignore it. cancel() and whenIdle() control lifecycle. Caller, factory, and consumer co-own teardown through one awaited disposer.

Agent Scope

Each agent owns scoped agent.ctx; shared storage overlays its tool, prompt, and command entries on globals while preserving domain views (decision). Scoped listeners filter dispatch; contributions unwind with awaited cleanup. CreateAgentOptions.setup(agentCtx) composes before publication and may return a synchronous commit that the factory invokes immediately before registry entry, after every setup await. Typed resolvers derive carrier checks from merged Events and scopeTarget (semantic gates). Details: agent scope, subagent composition. AgentLoop runs under ctx.agents.withInitiator(); private orchestration derives agent.session, but turn, step, signal, cwd, and authority stay explicit (decision).

State

Session Log

The session log is authoritative. deriveMessages() projects model history; raw assistant/chunk events preserve replay and UI fidelity. Fork, resume, transcript rendering, telemetry, and persistence derive from this stream.

Model-visible ⟺ logged: before step/start, the loop appends the full current runtime-context snapshot as a sourced user/message, then snapshots derived messages. Those messages and the folded request/header reconstruct each request. The header marks adapter defaults so later proposals discard them and re-resolve the route without losing explicit settings. dsh-agent-loop/invariant asserts this through ctx.invariants (reconstructability).

Durability is a plugin concern. Backends eagerly drain synchronous session/event notifications. session/flush barriers precede each request and top-level tool dispatch, then follow turn/end before another queued turn or idle observation. SessionPersistence stores SessionEvent directly and metadata in SessionHeader; JSONL defaults to checksummed Zstandard, while SQLite shares the contract (decision).

Between turns, owners append log-only events through Session, flushing only for durability. session/title needs eager persistence and lifecycle drains; manual compaction flushes its bracket before releasing admission. Title work never delays responses; latest wins with provenance. Title records are inherited fork boundaries (decision).

Model Content

Messages use typed blocks from merge-extensible ContentBlockMap; the pattern also types MessageSource, FinishReason, TurnTrigger, and TurnEndReason. New blocks coordinate adapters, UI, compaction, token metering, and persistence; replay measurements live in token-meter.md.

Streaming uses raw chunks and BlockAssembler. Each LlmAdapter.stream() is one provider attempt; adapters report normalized failure facts, and a handling agent/request-error plugin returns a retry action. The loop logs chunks, successful provenance, and replay state. Remote adapters use per-read idle watchdogs. Replay crosses routes only through a shared adapter instance (contract).

Extension And Composition

Capability Pattern

A swappable capability usually has interface / implementation / consumer layers: service/events, backend, and model-facing tools/prompts. Bash is the reference; the capability graph maps each family.

Exceptions combine LLM interface/consumer, filesystem policy, web registries, and named skill/subagent providers. Subagents spawn fresh, fork a completed-turn prefix, or use ACP children (subagent.md).

dsh-workspace-context injects baseline at the first agent/step and appends ctx.fs-discovered changes through tools/post-execute; its decision records isolation. dsh-paths owns shared paths.

Bundles And Apps

dsh-agent-spine-demo bundles a spine and optional goals. App packages own TUI, CLI, ACP automation, and JSON-RPC front doors (README, acp/, ui/). dsh-jsonrpc-agent boots external cordis.yml; the Python SDK defaults when config is absent (Python SDK). Thin deployments use swappable backends and optional tools (examples/, runnable wirings, graph atlas).

Where New Behavior Goes

New behavior attaches to a documented extension point; a loop change updates this map.

Goal Mechanism
Add a model provider register its adapter on ctx.llm
Add a model-facing capability register on ctx.tools; schemas join prompt assembly
Add shell execution implement and register a ctx.bash backend; the local backend spawns through ctx.subprocess
Add persistent terminal execution register a ctx.pty backend plus dsh-tool-pty
Add a human command register on ctx.commands; adapters discover and dispatch without a model turn
Add background work register on ctx.tasks; generic task_* tools collect or stop it
Add filesystem access or policy implement a ctx.fs provider or listen to fs/* policy events
Confine spawned processes use a ctx.sandbox backend; consumers wrap argv before spawning
Intercept a request, tool, or turn use its agent/* or tools/* event; agent/turn-stopping is the stop boundary
Add model-facing context call agent.inject() to append a sourced user/message without a turn
Add UI or editor integration drive ctx.agents, render from session/event; terminal-only overlays use ctx.tui
Add durable session state extend SessionEventMap; render and replay from the log
Add asynchronous session-title generation register the sole ctx.sessionTitle provider
Manage a same-session objective use ctx.goals; continue through Agent and agent/*
Fork a live session call ctx.sessions.fork(source, boundary?, childSessionId?)
Scope a registration to one agent use its agent.ctx (see Agent Scope)

The extension cookbook has plugin skeletons and the feature-to-seam map; guides cover packages, tools, LLM adapters, and vendored packages.