Tianyi Cui
f393043b03
The first OUT-OF-PROCESS subagent backend, proving the seam generalizes past the in-process backends. @deepseek-ai/dsh-subagent-acp runs each child agent in a spawned subprocess, driven over the Agent Client Protocol as the CLIENT — the direction-inverted twin of the dsh-acp server bridge. Point the configured command at the acp-agent example and the harness talks to its own process. - Fresh process per run: start spawns, runs one ACP session (initialize → newSession → prompt), dispose kills the subprocess and awaits its exit. - Minimal client stub: advertises no fs/terminal; accumulates agent_message_chunk text as the result output; auto-answers session/request_permission by a configured policy (reject default / allow). No start-time capabilities (an out-of-process child can't enforce the parent's depth/tool-filter); ignores request.parent; injects only `subagents`. - StopReason mapping (end_turn→completed, cancelled→aborted, …); result resolves error/aborted on a child failure, never rejects (seam contract). - Security: credential-shaped ambient env vars are scrubbed; the child's own key is forwarded only via explicit config.env. A spawn-level error (ENOENT) is captured and raced against the ACP drive so a bad command settles error rather than crashing the parent. Testing designed at every tier: keyless integration drives a scripted mock ACP server subprocess (cancellation incl. the pre-newSession race and a torn-pipe-after-cancel, permission auto-answer, non-message updates, spawn failure, HMR, export shape) at 100% coverage; a with-key e2e drives the REAL acp-agent example process (PONG + real file write, verified on disk) — the harness driving itself. Snapshot coverage of an ACP child is deferred as TODO(acp-subagent-replay) (each child is its own process with its own replay). Stayed on @agentclientprotocol/sdk 0.25.1: the proposed 0.28.x bump only deprecates the stable ClientSideConnection/AgentSideConnection API this layer uses (33 sites incl. the server bridge), turning no-deprecated red across code this PR shouldn't rewrite — that fluent-API migration is its own follow-up. The backend needs nothing 0.28.x adds. This completes the subagent seam stack (PR1 interface → PR2 in-process → PR2.5 snapshot infra → PR3 ACP); the seam RFC moves to implemented/, amended.
2.0 KiB
@deepseek-ai/dsh-tool-subagent
The model-facing subagent tool: delegate a self-contained task to a child agent and return its final output. Pure schema + lifecycle shaping over the ctx.subagents provider registry — an in-process, ACP, or future A2A backend swaps in without changing what the model sees.
Provider selection is config, not model-facing
This plugin binds to exactly one provider (Config.provider). The model sees only { description, prompt } — there is no provider/type parameter in the schema. To expose more than one transport, load the plugin more than once, each bound to a different provider and a distinct toolName (the tool registry rejects a duplicate name, so a second load that kept the default subagent name would throw). Keeping selection in config (not the schema) is the deliberate split: the service holds a multi-provider registry; the tool picks one.
| Config key | Meaning |
|---|---|
provider (required) |
The ctx.subagents provider name to start runs on (spawn, fork, acp, …). |
toolName |
The model-facing tool name to register (default subagent). Set a distinct value per load when exposing multiple providers, e.g. subagent + subagent_acp. |
agentOptions |
Default per-child { model?, systemPrompt? } applied to every spawned child. |
Lifecycle (synchronous collect)
execute starts a run on the configured provider and awaits run.result inside a try/finally that always dispose()s the run — the owned child agent/session is torn down on every path (success, error, abort), never leaked. The tool's abort signal (exec.signal) is bridged to run.cancel(). A non-completed stop reason (aborted/error/max-tokens/refusal) maps to an isError tool result rather than returning partial output as success.
Background / poll collection is deferred (see the RFC); this cut blocks the parent turn until the child finishes.