Files
deepseek-harness/packages/ui/acp-agent/README.md
Tianyi Cui 66e56bd395 fix review findings: stale coding-agent README, export-shape guards, overclaim wording
Codex review of PR #88 found three issues in the example-app extraction:

A1 — examples/coding-agent/README.md's plugin table still listed the OLD
direct-wired leaf entries (agent-loop, session-persistence, src/stdio-chat.ts —
the whole src/ dir is gone). Rewrite it to the four real leaf entries the
current cordis.yml loads (hmr, llm-deepseek, bash, stdio-agent), noting that
tool-bash/persistence/agent/loop now live inside the agent-core + stdio-agent
bundles.

A2 — the three new app/spine packages (agent-core, stdio-agent, acp-agent)
export NO `inject`, so a stray `export default apply` would let unwrapExports
collapse the module and silently DROP name/Config WITHOUT crashing — the
real-load-path smokes would stay green. agent-core is never Loader-unwrapped at
all. Add an explicit export-shape guard per package: assert no `default` export
and that the real Loader.unwrapExports leaves name/Config/apply intact. Verified
each fails when `export default apply` is added.

B — soften "structurally unreachable / cannot wire a stdout logger" overclaims
in the acp-agent/agent-core READMEs and the implemented RFC: a leaf CAN still add
a sibling logger entry; the accurate claim is the app omits one so the default
leaf has nothing to get wrong. Keep the safety directive (never add a stdout
logger to an ACP leaf).
2026-06-21 12:49:59 +08:00

2.6 KiB

@deepseek-ai/dsh-acp-agent

The ACP server app: a Cordis app plugin that composes the providerless agent spine (@deepseek-ai/dsh-agent-core) with the front-door cluster an Agent Client Protocol server needs, and a bin that boots a leaf cordis.yml speaking ACP JSON-RPC on stdio.

It is the structured counterpart to @deepseek-ai/dsh-stdio-agent: both consume the same spine, but this one bakes in the OPPOSITE front-door cluster.

What it bakes in — and what it deliberately omits

stdout is the ACP JSON-RPC channel, so the cluster is defined as much by what it LEAVES OUT as what it includes:

Plugin Why
@deepseek-ai/dsh-agent-core the spine, pre-creating no agents (ACP session/new creates them on demand)
@deepseek-ai/dsh-session-persistence-jsonl durable JSONL session log (the bridge advertises loadSession)
@deepseek-ai/dsh-acp the bridge that owns stdout for JSON-RPC
console logger omitted — it writes to stdout and would corrupt the protocol frames (the stdout-purity footgun)
hmr omitted — the editor owns the subprocess

Because the package wires no logger entry, an ACP leaf has nothing to get wrong by default: it only picks backends, so the common mistake — copying a console-logger entry from the stdio config — has no place here. (A leaf author technically can still add @cordisjs/plugin-logger-console as a sibling entry; the package can't forbid that. So the rule stands: never add a stdout logger to an ACP leaf — stdout is the JSON-RPC channel. Use a stderr exporter if you need logs.)

Config

Key Default Routed to
model (required) the per-session agent template the bridge creates agents from
systemPrompt (required) the per-session agent's system prompt
persistenceRoot ./.sessions the JSONL backend's root directory

The leaf supplies the swappable backends: an LLM adapter (llm-deepseek for the real model, llm-replay for keyless snapshot replay) and a bash executor (bash-local).

The bin

dsh-acp-agent [path-to-cordis.yml] (default ./cordis.yml):

  • loads a gitignored .env from the cwd — skipped in snapshot REPLAY so a stray key can never trigger a live call;
  • honors DSH_SNAPSHOT=replay by booting the sibling cordis.snapshot.yml (the keyless replay tree, llm-replay in place of llm-deepseek);
  • in a snapshot run, disposes the context on stdin EOF so the session log is fully flushed before exit.

All diagnostics go to stderr — stdout is the protocol.