Skip to content

release: 0.7.0 — one run per conversation and could-not-process; agent-core 0.7.0, manual 2.4.0 - #23

Merged
sanctrl merged 3 commits into
mainfrom
feat/could-not-process
Nov 20, 2026
Merged

sanctrl merged 3 commits into
mainfrom
feat/could-not-process

Conversation

@sanctrl

@sanctrl sanctrl commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

What changes

Docs for the engine's one run per conversation and could not process (agentchatme/agentchat-agent-core#27), which this integration's daemon takes as is: the runner inherits the Codex preset's runtime check, and status / doctor are the engine's. One code change: uninstall now clears the daemon's status files and the could-not-process list (below); otherwise a comment in daemon-main.ts.

  • README: the diagram and how a wake becomes runs (one per conversation, one at a time, the scoped agentchat_wait for a live exchange, invites in a run of their own, an older server's one run for the tray); Could not process in plain words: what happens on the first and second failure, the 24 h pause that ends on its own, the runtime check (a tiny codex exec call with the fixed prompt Reply with the single word OK., no peer content, no AgentChat tools, no key, read-only and ephemeral, in $CODEX_HOME/agentchat-runtime-check), and what status and doctor show; the live-fire steps' log lines (run start kind=conversation conv=…, two peers, two runs).
  • A test of the runtime check this runner inherits (its exact argv, folder, no key and no headers helper, ok and not ok), skipped until the engine pin has checkRuntime.
  • uninstall leaves no daemon status behind. It never called the engine's clearIntegrationStatus(identityHome), so integration.status.json (and the file an older daemon wrote) stayed, and from the engine release with could not process could-not-process.json would too. It now calls it once the service is gone. A test runs uninstall over all three files (the ledger's part runs once the engine pin knows it).
  • The runtime-check test follows the engine's current check (agent-core#27 c7e1f1a): the fake answers a JSON result when asked, and argv and settings come from the engine's builders.
  • CHANGELOG (Unreleased). No version bump; the pin moves at the release step.

Checked whether this package duplicates anything the engine changed: status, doctor and daemon status render through the engine (createIdentityCommands, formatIntegrationStatus), and the runner extends the engine's CodexRunner (its env() override, which drops AGENTCHAT_API_KEY, also feeds the check), so nothing here pins old engine text; no test needed changing.

Note: unlike Claude Code's, Codex's runtime check is not tool-free. It runs with --sandbox read-only and the agentchat server switched off, but Codex keeps its shell and the user's other MCP servers (the engine's flags). The README says exactly that rather than "no tools".

Verification

  • Tests: with the engine and SDK linked from a build of agentchat-agent-core#27 head c7e1f1a and of agentchatme 1.5.0 (built from its release commit d6b74a3) (symlinks in node_modules, not committed): pnpm type-check ok, pnpm build ok, pnpm test 7 files, 122 passed, 1 skipped (123), bash scripts/release-smoke.sh ok. On the current pins (agent-core 0.6.0): type-check and build ok, pnpm test 121 passed, 2 skipped (the runtime-check test skips; the uninstall test runs, without its ledger part), so CI should stay green. gitleaks on the commits: no leaks.
  • Review: self-review of the diff only; no independent review yet (draft).
  • Live check: n/a until release: lab live-fire after the engine release.

Risk and rollback

Docs, tests and the one-line uninstall fix. Merge with or after the engine release so the README matches the shipped daemon. Rollback: revert.

The engine's daemon starts one run per conversation that owes a reply
(with the server's per-conversation prompt) and reports a conversation
its runs keep failing on, while the runtime otherwise works, as could
not process. Nothing in this package's code changes for it beyond a
comment: the runner inherits the engine preset's runtime check, and
status and doctor are the engine's.

- README: the diagram and how a wake becomes runs (one per
  conversation, one at a time, the scoped wait for a live exchange,
  invites in a run of their own), could not process in plain words
  (what the owner sees in status, the 24 h pause that ends on its own,
  the runtime check: a fixed "Reply with the single word OK." model
  call with no peer content), the live-fire log lines.
- A test of the runtime check this runner inherits (its flags, its
  folder, no key in it), skipped until the engine pin has it.
- CHANGELOG (Unreleased).
uninstall never called the engine's clearIntegrationStatus, so the
daemon's status file (and the one an older daemon wrote) stayed in the
identity home; from the engine release with could not process, so would
could-not-process.json. It now clears them once the service is gone.

A test runs uninstall over all three files (the ledger's part once the
engine pin knows it). The runtime-check test follows the engine's
current check: the fake answers a JSON result when one is asked for,
and the argv and settings come from the engine's own builders.
@sanctrl
sanctrl marked this pull request as ready for review November 20, 2026 00:56
@sanctrl sanctrl changed the title Docs: one run per conversation, and could not process release: 0.7.0 — one run per conversation and could-not-process; agent-core 0.7.0, manual 2.4.0 Nov 20, 2026
@sanctrl
sanctrl enabled auto-merge (squash) November 20, 2026 00:56
@sanctrl
sanctrl merged commit 7cf0601 into main Nov 20, 2026
5 of 6 checks passed
@sanctrl
sanctrl deleted the feat/could-not-process branch November 20, 2026 01:00
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant