[lihongyuan99]

Summary

Four docs describe present-tense code but point at module paths that no longer exist. Both targets moved out from under them:

GlobalProcessor / chat router

• docs/agent-call-agent.md:44-45 ("相关文件") — infra/gateway/processor.py, api/routers/chat.py
• docs/agent-delegation.md:25 ("Octop 侧保留代码") — infra/gateway/processor.py
• docs/adr/001-single-process-model.md:35 ("that seam is already partially visible in …") — infra/gateway/processor.py

GlobalProcessor is at src/octop/infra/gateway/process/processor.py:127 and is still the class these lines describe: compose_followup at :454, on_reply at :463, the # -- TeamProcessor (delegates to TeamManager) banner at :441, and slash really is dispatched there (:621 cmd = parse_slash(msg.text)) — so the "@ 与 slash" claim stays true. The chat HTTP surface is now the package src/octop/api/routers/chat/ (routes.py, models.py, turn.py, ws.py, …); the @mention request field is chat/models.py:66.

the gateway/media/ split

• docs/agent-backend-file-io.md:131 — infra/gateway/backend_files.py
• docs/agent-backend-file-io.md:140 — infra/gateway/media.py AgentBackedMediaBackend

backend_files.py now sits under infra/gateway/media/, and AgentBackedMediaBackend is at infra/gateway/media/ingress.py:14. The "现状" cell beside the first row claims a Path.read_bytes fallback on preview failure, and that is still what the code does (media/backend_files.py:406, documented at :13 and :422) — so only the path literal was stale, not the description. The neighbouring infra/backup/system_archive.py / workspace_archive.py rows already resolve and were left alone.

Only path literals change; no sentence is reworded and no new claim is added.

Target branch

• Base is develop (feature / fix — default)
• Base is main (release/* or hotfix/* only)

Type of change

• Bug fix
• New feature
• Breaking change
• Documentation
• Refactor / chore
• Release / hotfix

Test plan

Nothing in the test suite covers path literals in prose, so this is verified by an executable check instead — a script that pulls every backticked repo-relative path out of these four files and resolves each against both <root>/<p> and <root>/src/octop/<p>.

Before (run against the blobs at base 02a6e28d):

BEFORE (base 02a6e28d) — checked 4 files; MISSING refs: 7
  docs/agent-call-agent.md:44  ->  infra/gateway/processor.py
  docs/agent-call-agent.md:45  ->  api/routers/chat.py
  docs/agent-call-agent.md:45  ->  infra/gateway/processor.py
  docs/agent-delegation.md:25  ->  infra/gateway/processor.py
  docs/adr/001-single-process-model.md:35  ->  infra/gateway/processor.py
  docs/agent-backend-file-io.md:131  ->  infra/gateway/backend_files.py
  docs/agent-backend-file-io.md:140  ->  infra/gateway/media.py

After:

checked 4 files; MISSING refs: 0

Gates, run on this branch to confirm nothing else moved:

• uv run ruff check src tests → All checks passed!
• uv run ruff format --check src tests → 1060 files already formatted

No Python is touched (git diff --stat: 4 files, +6/-6), so mypy and pytest are unaffected by this diff and were not re-run.

Checklist

• Updated CHANGELOG.md (if user-facing) — no runtime behaviour changes; docs-only path repairs carry no entry in this file's history
• README / docs updated (if needed)
• Base branch is develop

Deliberately left out

• AGENTS.md §5 and §9 ("How is a message processed?") hold the same two stale infra/gateway/processor.py / api/routers/chat.py references, but docs(agents): repair module paths that no longer exist #908 already rewrites that file — no point landing a second edit on it.
• docs/agent-interop-mailbox.md:225/298/301 also say infra/gateway/processor.py, yet its §7 and §9 tables are an implementation plan whose sibling rows (infra/agents/agent_call.py, tests/test_peer_agent.py, tests/test_inbox.py) never landed either. Repairing one row of a stale plan would make the table read as current, so that is a maintainer call rather than a path fix.
• docs/configuration.md:246 references infra/connectors/credential_crypto.py, which does not resolve; that file is already touched by docs(cli): stop naming a command the CLI does not have #965 plus four other open PRs, so it is left alone to avoid a needless conflict pile-up.