Fleet: eval sets at cloud scale, plus two new ways to author a pack

The Fleet epic landed in fifteen pull requests — case-level fan-out, Kubernetes sandboxes, per-account LLM throttling, eval-set manifests, a live matrix dashboard, post-hoc scanners, budgets with emergency stop, OTel metrics, and a Helm chart. Alongside it: one-command local setup for contributors, custom OpenAI-compatible endpoints, pack drafts generated from a tryout or a plain-English description, and a machine-readable public site.

  • Fleet eval sets
  • Self-host and scale
  • Pack authoring
  • Onboarding

What shipped

Added

Improved

  • OpenTelemetry metrics, stall detection, and a ready-to-import alert pack for run and eval-set health.
  • Oversized run-event payloads now spill to object storage, and outbound LLM calls are throttled per account across workers.

Fixed

Merged pull requests

31 PRs
  1. #1256feat: make AgentClash agent-readable
  2. #1255feat: support custom OpenAI-compatible endpoints
  3. #1247chore(main): release 2.1.2
  4. #1246feat: generate an eval pack draft from a plain-English description
  5. #1245feat: promote agent tryouts into editable pack drafts
  6. #1233chore(main): release 2.1.1
  7. #1232fix: human labels and clearer Fleet matrix UI
  8. #1231docs: add Fleet eval-set guides to agentclash.dev/docs
  9. #1230chore(main): release 2.1.0
  10. #1229fix: onboarding funnel quick wins — dead CTAs, jargon, hidden prerequisites
  11. #1228feat(api): /healthz/ready readiness probe
  12. #1227feat(deploy): Helm chart, KEDA Temporal scaling, GHCR images (#1209)
  13. #1226feat(observability): OTel metrics, stall detection, alert pack (#1211)
  14. #1225feat(scanners): post-hoc eval-set scanner framework (#1208)
  15. #1224feat(evalset): per-set budgets and emergency stop (#1210)
  16. #1223feat(web): live eval-set matrix dashboard (#1207)
  17. #1222feat(evalset): case_results warehouse search, report, and export
  18. #1221feat(cli): add evalset submit, status, logs, report, and cancel
  19. #1220feat(evalset): persist eval sets and orchestrate child sessions
  20. #1219feat(evalset): parse manifests and dry-run combination expansion
  21. #1218feat(runevents): spill oversized event payloads to object storage
  22. #1217feat(provider): distributed outbound LLM throttle per account
  23. #1216feat(sandbox): Fleet 4 — Kubernetes provider + conformance suite
  24. #1215feat(sandbox): Fleet 3 — capacity budget, warm pool, docker provider
  25. #1214feat(workflow): Fleet 2 — bounded fan-out, worker knobs, queue partitioning
  26. #1213feat(workflow): Fleet 1 — case-level fan-out for native run-agents
  27. #1197fix(deploy): build backend images from repo root so ../runtime resolves
  28. #1180chore: community-health files and structured issue intake (PR B)
  29. #1179docs: README landing page and adoption assets (PR F)
  30. #1178feat: frictionless local setup and contributor onboarding (PR A + E)
  31. #1048fix(web): reorder benchmarks hub to lead with proof, drop empty scoreboard columns
All releases