You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
F16 — TUI streaming-watchdog wiring:
- Adds `cli.tui.streamingWatchdogMs` config field (zod-validated, optional).
- `runTui` now reads this value and passes it to `createEventHandlers`,
closing the gap where `context.streamingWatchdogMs` was a consumer-side
knob with no producer-side wiring.
- Raises the in-code default from 30s to 600000ms (10 minutes). The 30s
default tripped on long-reasoning models, slow tool runs, and SSE
keepalive gaps; the 10m value matches the empirically validated
downstream patch.
- Adds `resolveTuiStreamingWatchdogMs(config)` helper + unit tests for
config-shape edge cases (NaN, negative, Infinity, fractional, string).
- Adds tui-event-handlers test verifying the new default fires at 10m
(not 30s) when no override is supplied.
F18 — CLI watchdog defaults:
- Raises `CLI_FRESH_WATCHDOG_DEFAULTS.maxMs` from 600000 (10m) to
3600000 (60m).
- Raises `CLI_RESUME_WATCHDOG_DEFAULTS.maxMs` from 180000 (3m) to
3600000 (60m).
- Per-CLI-backend overrides at `cli.watchdog.fresh.maxMs` /
`.resume.maxMs` continue to apply unchanged; this commit only moves
the fallback ceiling so users without a per-backend override don't
see false aborts on long but legitimate runs.
Backward compatibility:
- `cli.tui.streamingWatchdogMs` is optional; existing configs without it
pick up the new 10m default rather than the prior 30s.
- The CLI watchdog defaults change only affects callers that did NOT
set their own per-backend ratio/min/max, where previously the
computed `maxMs` ceiling was 600000/180000.
- All existing tests for `resolveCliNoOutputTimeoutMs` continue to pass
because their assertions (e.g. 480000 = 600000 * 0.8) sit below the
new ceiling.
Docs:
- `docs/gateway/configuration-reference.md` documents the new
`cli.tui.streamingWatchdogMs` field with default + disable semantics.
Co-Authored-By: Forge <[email protected]>
- TUI/CLI runner: expose `cli.tui.streamingWatchdogMs` config field, raise the in-code streaming watchdog default from 30s to 10m, and lift `CLI_FRESH_WATCHDOG_DEFAULTS.maxMs` (10m → 60m) and `CLI_RESUME_WATCHDOG_DEFAULTS.maxMs` (3m → 60m) so long-reasoning models, slow tool runs, and SSE keepalive gaps stop tripping false aborts; per-backend overrides at `cli.watchdog.fresh.maxMs`/`.resume.maxMs` continue to apply unchanged.
9
10
- Chat commands: add `/think default` and `/fast default` to clear session overrides and inherit configured/provider defaults. (#79385) Thanks @VACInc.
10
11
- Dependencies: refresh workspace dependency pins and lockfile, including `@openai/codex` `0.130.0`, `acpx` `0.7.0`, AWS SDK `3.1044.0`, OpenTelemetry `0.217.0`, `typebox` `1.1.38`, `vite` `8.0.11`, `oxfmt` `0.48.0`, and `oxlint` `1.63.0`, and update the Codex harness model snapshot for the new bundled app-server catalog.
11
12
- Plugins/install: add guarded plugin install overrides so onboarding and repair tests can route specific plugins to registry specs or local `npm pack` artifacts via environment variables.
0 commit comments