Skip to main content

AgentXchain v2.46.1

v2.46.1 is a truth patch for the timeout dashboard surface. The recently shipped timeout views showed phase/run pressure correctly, but they silently dropped live turn-scope pressure. That was not a docs problem. It was a product defect.

Per-turn timeout pressure now renders truthfully​

Both dashboard timeout surfaces now evaluate active turns individually:

  • Repo-local Timeouts evaluates phase/run once, then evaluates each active turn for turn-scope pressure
  • Coordinator Timeouts does the same for each child repo
  • Live turn-scope rows now include turn_id and role identity, so operators can see which turn is actually over budget

The evaluator contract stayed explicit. We did not make evaluateTimeouts() infer active turns implicitly. The dashboard now owns the aggregation layer instead of hiding behavior inside the evaluator.

Repo timeout endpoint proof is now real​

The repo-local timeout endpoint had render tests and wiring assertions, but no direct server-module proof. That is why a broken context-loader path survived. v2.46.1 closes that gap.

Added proof​

  • direct readTimeoutStatus() tests for:
    • configured vs not-configured state
    • active run with per-turn live pressure
    • blocked run returning empty live arrays while preserving timeout history
    • missing governed state
  • coordinator timeout tests now prove turn-scope live pressure appears in repo snapshots and rendered cards
  • the existing dashboard bridge and full CLI suites remain green

Docs and operator surface alignment​

  • Timeouts now states that dashboard live-pressure tables preserve turn identity for turn-scope entries
  • CLI Reference now describes per-turn identity in both timeout dashboard views
  • release metadata, sidebar, llms.txt, and sitemap.xml are aligned to this patch release

Evidence​

  • 3430 tests / 735 suites / 0 failures
  • node --test cli/test/dashboard-timeout-status.test.js → 18 tests / 3 suites / 0 failures
  • node --test cli/test/dashboard-coordinator-timeout-status.test.js → 8 tests / 3 suites / 0 failures
  • node --test cli/test/dashboard-bridge.test.js → 46 tests / 11 suites / 0 failures
  • cd website-v2 && npm run build → clean production build