@deepseek-ai/dsh-agent

dsh-agent

English | 中文

Agent interface, registry, process-local initiator scope, and agent/* event vocabulary. Every plugin (UI, hooks, orchestrators) programs against the Agent handle defined here — it has zero loop dependency, so the loop is swappable.

The optional @deepseek-ai/dsh-agent/invariant companion registers this package's agent-status transition checks with ctx.invariants. The root agent service does not load diagnostics implicitly.

Service: AgentRegistry (ctx key: agents)

Tracks live agents and carries the initiating Agent through asynchronous driver work without importing the concrete loop package.

Public API

The scoped-registration surface: Agent.ctx is the agent's scope context (dsh-scope, key = the agent) — register tools/sections/variables/listeners through it for that agent alone, all unwound on disposal. agentEvents(ctx, agent) is the fused dispatcher for ordinary agent-subject operations (carrier + injected subject in one move); its notification mode invokes every listener and contains both synchronous throws and returned-promise rejections. The registry lifecycle pair reuses one stable routing carrier. assembleContextFor(agent) builds the per-agent assembly context (agent + scope together). installModelSelection(agentCtx, selection) snapshots a mutable provider/model/reasoning-effort selection during prompt assembly, applies its provider and model to prompt variables, and applies the complete selection to request routing for one step; an absent selected effort clears an inherited effort so adapter/provider defaults apply. CreateAgentOptions.setup(agentCtx) and ResumeAgentOptions.setup(agentCtx) compose a fresh or resumed agent's scoped world while both objects remain unpublished. Setup is trusted, composition-only same-process code: drive the agent only after creation resolves.

AgentOptions supplies the initial provider/model route and an optional positive maxTokens output cap. The concrete loop resolves any exact-model adapter default, records the effective cap in the request header, and applies it to each conversation-model request; an explicit Agent option wins, while omission leaves the adapter or provider route default in control.

• ctx.agents.register(agent: Agent): () => void — record an already-constructed agent. Disposed with the calling fiber.
• Advanced ordered lifecycle: enter(agent, owner): () => void enforces agent.id === agent.session.id, performs the authoritative ID collision check, and inserts without announcing; owner explicitly records the live creator-agent relation (or undefined for a root), independently of durable session lineage. announce(agent) emits agent/created exactly once. A detach requested synchronously by a creation listener is deferred until that dispatch unwinds, and every detach checks the captured entry object, so a stale capability cannot delete a later same-ID replacement. The async factory uses this split; ordinary plugins use register().
• ctx.agents.get(id: SessionId): Agent | undefined
• ctx.agents.isOwnedBy(id: SessionId, owner: Agent): boolean — whether the exact live entry was created through that parent agent's scoped context; runtime ownership is independent of durable session lineage.
• ctx.agents.list(): Agent[]
• ctx.agents.roots(): Agent[] — live agents created without an owning agent context; a resumed lineage-bearing session can still be a runtime root.

Initiating Agent scope

AgentLoop runs each concrete driver's complete lifetime inside an initiator boundary. Concurrent drivers remain isolated: a child driver's continuations carry the child, while the parent continuation regains the parent as soon as withInitiator() returns; drain tracking continues until the child driver's Promise settles. Creation, persistence load, and unpublished setup remain outside the child's boundary, so setup initiated by a parent inherits the parent while agentCtx.agent identifies the child explicitly.

• ctx.agents.currentInitiator(): Agent | undefined — read the inherited initiator without requiring one.
• ctx.agents.requireInitiator(): Agent — read it or throw no initiating agent is active.
• ctx.agents.withInitiator(agent, operation) — run with one exact Agent and preserve the operation's exact synchronous value or Promise.
• ctx.agents.withoutInitiator(operation) — hide an inherited initiator for unrelated process-local work.

The scope carries the Agent itself and is process-local. Ambient presence is neither liveness proof nor authorization; explicit Agent fields remain authoritative at service, worker, process, persistence, and wire boundaries. Teardown rejects new boundaries, lets injected dependents and returned-Promise boundaries drain, then disables the underlying AsyncLocalStorage; unreturned work remains owned by the subsystem that detached it. If a boundary's inherited async chain starts an owning Cordis fiber's unload, that nested boundary chain is released from the drain so the unload cannot wait on itself; its continuations observe the disposed service after teardown. The initiator-scope decision owns the detailed boundary and teardown contract.

Factory API (creation)

Agent creation is provided by the plugin implementing AgentFactory (dsh-agent-loop), registered via setFactory. This keeps creation on the dsh-agent interface so consumers (UI, the ACP bridge) program against ctx.agents without depending on the concrete loop package. The registry canonicalizes an already traced Service to its concrete target and re-traces each call through the caller's context; this avoids nested Cordis shadows while passing an explicit caller-bound ownerCtx to plain factories.

• ctx.agents.setFactory(factory: AgentFactory): () => void — register the creation factory (the loop calls this on construction). Throws on a second factory; the slot clears on dispose.
• ctx.agents.create(options: CreateAgentOptions): Promise<AgentHandle> — create a session and agent, await optional setup while unpublished, then publish through final SessionStore.enter() and AgentRegistry.enter() checks. Concurrent same-ID creation is unsupported: more than one operation may prepare, but only one can enter; every loser rolls its private scope/session/driver back. An optional creation-only signal cancels unpublished setup and is detached before the handle is returned; later cancellation uses handle.dispose() or agent.cancel(). Publication is rollback-covered and every delivered creation edge is paired during rollback. Rejects if no factory is registered.
• ctx.agents.resume(options: ResumeAgentOptions): Promise<AgentHandle> — load a persisted session (session persistence), mint a fresh unpublished agent scope, await optional setup, and use the same final-entry publication sequence. Its optional signal is likewise creation-only. Rejects if no factory is registered or session persistence is unconfigured.

AgentHandle = { agent: Agent; dispose(): Promise<void> }. The disposer is a consumer capability — no observer holding the bare registry entry can tear the agent down. The caller fiber and the registered factory provider are structural co-owners: caller unload enforces structured ownership, while factory unload must stop old instances because their scoped dependency surface belongs to that provider. dispose() from any owner reaches one memoized quiescence boundary: it stops the loop, awaits its exit, unregisters the agent, removes its session from the store, and finally unwinds its scoped world. ctx.agents.get(id) still returns a bare Agent; the ACP bridge and in-process subagent backends hold consumer handles, while config-created agents are already owned by the loop fiber.

Live events

dsh-agent declares the live agent/* coordination vocabulary so plugins do not depend on the concrete loop. Exact signatures, dispatch modes, scope-filtering rules, and payload contracts live in the generated region of core.md; the architecture turn flow shows their order relative to durable session events.

The lifecycle edges have two important local caveats. agent/created runs after scoped setup and after both session and agent registry entries exist. Setup is trusted composition-only code; the immediately following non-vetoing agent/session-start notification is the first supported startup injection point. agent/disposed always means the exact agent has left the registry. AgentLoop emits it after its driver is quiescent, while ordered teardown may still be detaching the session and unwinding the scope; custom agents registered directly own any stronger driver-ordering contract themselves.

Most interception points are cooperative waterfalls. agent/pre-step receives a payload carrying the subject agent, the exclusive claimed UserMessage[], and the proposed turn, step, and cancellation signal; its batch may be empty when tools already require another request. Agent-scoped turn extension points carry their explicit AbortSignal in the payload; the remaining turn-scoped extension points receive it through their request value. Listeners may cooperate with a signal but must not retain it as authority over another turn. agent/request-error is the failed-model-request recovery waterfall: it receives request coordinates, normalized failure facts, the serving registration's retry policy when available, and the signal. A listener returns { kind: 'retry' } without calling next() when it owns recovery. agent/turn-stopping runs before an otherwise completed turn closes. The explicit-cancellation decision owns signal lifetime; the agent-scope runtime-design Agent Note owns scoped dispatch and terminal settlement.

PreStepDecision is either { kind: 'reject' } or { kind: 'enter', messages }. The enter branch is the complete identified, frozen batch for the proposed step. A listener that wraps downstream entry preserves that batch unless it intentionally replaces it; additions follow the waterfall's natural return order. Claiming already removed the offered messages from the inbox, so rejection does not retain them. Messages inserted after the claim remain pending for a later boundary.

Inbox live notifications are deliberately per-message and minimal: agent/inbox/inserted { message }, agent/inbox/claimed { message, turn }, and agent/inbox/discarded { message }. They complement the durable agent/inbox/spliced projection without adding another lifecycle envelope.

Turn and step boundaries and the model token stream are durable session/event facts rather than mirrored agent/* notifications. Consumers read turn/*, step/*, and assistant/chunk from the session feed; tool policy and outcome observation belong to the complete pipeline documented by dsh-tools.

foldConsumedWork(events) reads that feed back for the one question the turn sequence cannot answer alone: what became of the work a log consumed. It returns the latest turn/end that accounts for consumed work — a turn that entered a model step, or one that claimed inbox input and then failed, was stopped, or was rejected before reaching one — plus whether accepted work was later cancelled out of the inbox unrun. Both facts come from the log, so a cancellation reads the same whichever owner issued it. A no-step turn that took nothing, or emptied its claim and completed, describes no work and is skipped; a blocked end over claimed input is an account, because rejection discarded that input.

Agent interface (types.ts)

The handle every plugin programs against:

• agent.inbox — the agent-owned projection of durable agent/inbox/spliced events. nextTurn and nextStep expose pending UserMessage values. append, prepend, replace, remove, clear, splice, and claim mutate them; replace(messageId, newMessage) and remove(messageId) locate the pending message across both lists. Replacement may change identity and publishes the old message as discarded followed by the new message as inserted. Ordinary removals and clear() are durable cancellations and emit agent/inbox/discarded. claim(target) removes the next proposed batch with pure deletion splices; the loop then emits agent/inbox/claimed. MessageId is the only occurrence identity and must remain unique while pending.
• agent.followup(message) — queue an ordinary next-turn message and wake the driver. It returns no completion handle; the message id identifies inbox insertion, claim, and discard facts, not a later output or turn/end.
• agent.steer(message) — queue waking next-step input. An idle agent starts a turn synchronously; a running driver consumes later steering at its next step boundary.
• agent.inject(message) — queue non-waking next-step context. A running driver claims it at the nearest later pre-step boundary; an idle driver leaves it pending until followup() or steer() wakes the driver. It may miss a request whose pre-step already claimed its batch.
• agent.cancel(cause, options?) — cancel the active driver and, unless options.keepInbox, durably cancel all pending inbox work. Idle cancellation is a no-op.
• agent.whenIdle() — observe whole-agent quiescence, including replacement work scheduled before the current driver retires. It does not settle any particular message.
• agent.session, agent.status, agent.options, agent.id, agent.ctx

running describes a driver-wide drain interval, not proof that a turn is still open; it can cover turn close, the durability checkpoint, and consecutive queued turns. Only a caller that owns a complete interval may summarize it as a run result (decision).

Extension points

• Agent creation: AgentLoop.create() is the concrete config-path implementation (in dsh-agent-loop), while programmatic consumers create/resume owned agents through ctx.agents.create() / ctx.agents.resume(). Replace the loop by implementing Agent and registering via ctx.agents.register().
• Event listeners: all agent/* events are declared here — no dependency on the loop package needed.
• Subagent delegation is not an Agent method; providers create or drive ordinary handles through the factory API, so delegation transports stay outside the core agent interface.

Model Experience

User, steering, and injected messages

What the model sees

send, steer, and inject feed the owning session. agent/pre-step and other declared events let plugins reject a proposed step or add durable request material; this interface contributes no fixed prose itself.

Token effect

Accepted content becomes retained history or a repeated session prefix; blocked content contributes no request tokens. Size is caller- and plugin-dependent.

KV Cache effect

Accepted history and steering are append-only; a blocked submission sends no request. A session prefix remains stable within its loop instance, while a new or resumed instance may establish a different prefix.

Agent-scoped request composition

What the model sees

Registrations through agent.ctx can shadow prompt sections or tools and can install agent-only interceptors during unpublished setup.

Token effect

The package adds zero tokens itself; scoped contributions affect only that agent and disappear on disposal.

KV Cache effect

Prefix-stable while an agent's scoped registrations are unchanged. Setup or reload that changes prompt sections, tool definitions, or request listeners may invalidate reuse from the first affected request token.

Known Limitations and Deferred Work

• Initiator scope is process-local — workers, child processes, HTTP, durable queues, and restarts materialize any required identity explicitly.
• Ambient identity may outlive liveness — consumers still check agent.status, cancellation, and the owning capability contract before lifecycle-sensitive work.
• Inter-agent channels beyond delegation — shared state, streaming child output, and background/poll semantics remain outside the current synchronous ctx.subagents seam.
• agent/session-start cannot gate startup — it remains a synchronous, veto-less notification; async composition that must finish before publication belongs in the factory's setup(agentCtx) transaction instead.
• cancel() clears the inbox by default — it aborts the in-flight turn plus queued and steering work; cancel(cause, { keepInbox: true }) aborts only the turn and preserves pending items. There is still no step-only abort that keeps the in-flight turn running (stop API Agent Note).
• Each additional UserMessage carries exactly one MessageSource — contributions from several plugins merged onto one tool call collapse under one source, so the message cannot name several producers.
• SessionStartSource reserves 'clear'/'compact' with no emitter yet — only 'startup'/'resume' occur until the driving subsystems land (TODO(compaction)).