Skip to content

feat(config): raise timeout ceilings, add pendingTtlMs - #67

Merged
alpertarhan merged 1 commit into
mainfrom
fix/issue-61-timeout-ceilings
Sep 17, 2026
Merged

alpertarhan merged 1 commit into
mainfrom
fix/issue-61-timeout-ceilings

Conversation

@alpertarhan

Copy link
Copy Markdown
Owner

Implements the config portion of #61. Review context: #61 (comment)

Changes

Raised ceilings (defaults unchanged)

  • codexMaxCallMs: max 300_000 → 3_600_000 (1 h per call)
  • maxLatencyMs: max 600_000 → 7_200_000 (2 h pipeline)

The old caps structurally excluded slow local models — the #61 reporter measured a successful 17.5-minute synthesis, and the run ceiling couldn't be configured past 10 minutes. 0 semantics are unchanged (codexMaxCallMs: 0 = heuristic watchdog, maxLatencyMs: 0 = unlimited) and are now spelled out in the README table, including the provider timeout multiplier note.

New pendingTtlMs setting (1000–3600000, default 300000)
The staged-summary TTL was a hardcoded 5 min inside the commit store. Now a first-class config key, wired into createCompactionCommitStore, exposed in the settings UI, and documented. Default behavior identical.

Verification

  • typecheck, 981 tests (new: ceiling boundaries accept/reject, pendingTtlMs validation + default, settings coverage), bench, build, release audit — all green
  • CI: verify job

- codexMaxCallMs max 300000 → 3600000, maxLatencyMs max 600000 →
  7200000: slow local models (15-20 min syntheses) were structurally
  excluded; defaults unchanged (0 = heuristic / unlimited)
- pendingTtlMs (1000-3600000, default 300000): the staged-summary TTL
  was hardcoded 5 min; now configurable and wired into the commit store
- settings UI + README document the raised ranges and 0 semantics
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