All pull requests

Pull requests

• chore(deps): bump the github-actions group across 1 directory with 3 updates

  dependenciesgithub_actions
  #730·dependabot[bot] opened Nov 2, 2026Bot·

  ·

• chore(deps): bump the cargo-minor-patch group across 1 directory with 3 updates

  dependenciesrust
  #729·dependabot[bot] opened Oct 26, 2026Bot·

  ·

• feat(live): self-consistent rules + code; release zerobench (native)

  #707·yoavkatz opened Oct 18, 2026Collaborator·

  ·
  1

• fix(claude-code): disable background tasks

  #649·elronbandel opened Oct 8, 2026Contributor·

  ·
  1

• feat(benchmarks): add promptfoo-redteam, a coding-agent red team

  #619·elronbandel opened Sep 22, 2026Contributor·15 of 17 tasks·

  ·
  2

• feat(delivery): registry retention census (#578)

  #594·yoavkatz opened Sep 15, 2026Collaborator·10 tasks done·

  ·
  1

  2

• fix(release): a channel run keeps per-task combos whose :latest lags

  #592·elronbandel opened Sep 15, 2026Contributor·11 of 17 tasks·

  ·
  1

• feat(runtime): start the gateway and the collector only when asked [blocked on per-task :latest stitch]

  #588·elronbandel opened Sep 15, 2026Contributor·

  ·
  4

• rfc(rules): the gateway and the collector become opt-in

  #587·elronbandel opened Sep 15, 2026Contributor·

  ·

• fix(release): let a dispatch publish the :latest channel

  #539·yoavkatz opened Sep 10, 2026Collaborator·

  ·

• Store replay fixtures zstd-compressed

  #510·yoavkatz opened Sep 9, 2026Collaborator·6 tasks done·

  ·

• fix(kind): cap Helm release names at 53 chars; add EVAL_HELM_SET escape hatch

  #487·yoavkatz opened Sep 7, 2026Collaborator·

  ·
  2

• fix(kind): route around Docker Hub's anonymous pull-rate limit

  #486·yoavkatz opened Sep 7, 2026Collaborator·

  ·

• feat(kind): add --eval-model-api to force the gateway upstream wire

  #410·yoavkatz opened Aug 25, 2026Collaborator·

  ·

• fix(kind): route docker.io through mirror.gcr.io to dodge Docker Hub 429s

  #409·yoavkatz opened Aug 25, 2026Collaborator·

  ·

• fix(build): wire per-task benchmarks as OCI-layout contexts, not registry refs

  #408·gitMichal opened Aug 25, 2026Collaborator·

  ·
  1

• fix(tau-bench): use an accessible model and make images usable on kind

  #400·yoavkatz opened Aug 19, 2026Collaborator·

  ·

• Benchmark-declared MCP for all 16 capable agents

  #396·elronbandel opened Aug 18, 2026Contributor·

  ·

• fix(appworld): grade via persistent session + isolate bridge.py's ground_truth access

  #395·elronbandel opened Aug 18, 2026Contributor·

  ·

• fix(chart): RFC-1123 Job names, and refuse a dataset Job that needs helpers

  #393·elronbandel opened Aug 18, 2026Contributor·

  ·

• feat(cli): --platform flag on eval-containers build

  #386·drbwa opened Aug 18, 2026Contributor·

  ·
  1

• fix(core): GOSU_IMAGE default doesn't require core/gosu on the target registry

  #377·drbwa opened Aug 17, 2026Contributor·

  ·
  1

• fix(swe-agent): point sweagent at /testbed instead of a fake /workspace/repo

  #323·cdoron opened Aug 10, 2026Collaborator·

  ·

• fix(compose): bind output volume to host path and pre-create it

  #312·cdoron opened Aug 10, 2026Collaborator·

  ·

• fix(gateways): strip trailing /v1 from OPENAI_API_BASE in bifrost/start

  #309·cdoron opened Aug 10, 2026Collaborator·

  ·