v0.0.29

Changelog

• 82fe87c Add site generator to present benchmark results (#615)
• d4435ab Migration: Remove k8s-ai-bench from kubectl-ai repo (#625)
• 6393786 Revert "build and run against current (#624)" (#626)
• 002c25a Unify session backends (#616)
• f3eb44a add sensible defaults and match previous behvaior (#621)
• 23faf4a build and run against current (#624)
• 50662f7 feat: don't use context for executor (#622)
• 1e7b9c1 feat: multi-sessions ui (#620)
• 333a38b feat: use-sandbox and use-image flag (#613)
• 38382d1 fix deadlock in gollm.NewClient() (#634)

---

Released by GoReleaser.

v0.0.28

What's Changed

• Update k8s-bench README and run-eval-loop.sh by @noahlwest in #606
• Add guide and script for running local models by @noahlwest in #607
• update k8s-bench references to k8s-ai-bench to prep for new repo by @noahlwest in #608
• (repo split) Update k8s-ai-bench scripts to download & install kubectl-ai instead of building it by @noahlwest in #610
• gollm: tolerate empty streaming parts in Gemini responses by @justinsb in #611
• feat: Add support for JSONL benchmark output format and simplify output paths by @noahlwest in #614

Full Changelog: v0.0.27...v0.0.28

v0.0.27

What's Changed

• feat: Add raw http tracing by @noahlwest in #556
• feat: add eval multi-run dev script by @noahlwest in #563
• Explicitly set default transport to avoid nil transport with journaling by @noahlwest in #565
• evals: Add per-task timeout to k8s-bench by @noahlwest in #566
• Feat OpenAI response api by @droot in #562
• enhance: canary deployment task by @ShubyM in #570
• eval: Add retry to kind cluster creation by @noahlwest in #571
• eval: Generally up timeouts for verify/setup scripts by @noahlwest in #572
• k8sbench: fix eval fix-pending-pod by @droot in #575
• k8s-bench: fix the eval fix-rbac-wrong-resource by @droot in #576
• k8s-bench: made prompt to be more specific for list-images-for-pods eval by @droot in #577
• k8sbench: added cluster-creation-policy support by @droot in #574
• eval: (fix) Avoid race condition in fix-image-pull task by @noahlwest in #579
• eval: (fix) Make debug-app-logs more accurate and consistent by @noahlwest in #578
• eval: (fix) avoid race condition in rolling-update-deployment verification by @noahlwest in #581
• eval: Make create-network-policy verify spec, not test realized state by @noahlwest in #582
• eval: (fix) Have horizontal-pod-autoscaler check spec of hpa, not realized state by @noahlwest in #583
• k8s-bench: fix hpa eval to support autoscaling/v2 apis by @droot in #584
• k8s-bench: made statefulset-cycle eval for be less resource intensive by @droot in #586
• eval: Small fixes by @noahlwest in #585
• eval: More realistic dev cluster prompt & update verification by @noahlwest in #587
• docs(README.md): improve documentation clarity, consistency by @appleboy in #588
• k8s-bench: Up per-task timeout (5m -> 10m) by @noahlwest in #591
• eval: Update create-pod's ns to avoid hinting, check that created pod image is correct by @noahlwest in #592
• feat: Return unexpected kubectl-ai errors in RunOnce mode, display them in k8s-bench by @noahlwest in #594
• eval: remove hints from crashloop and probes by @ShubyM in #593
• docs: update MCP standardize documentation and configuration examples by @appleboy in #590
• feat: add SkipVerify option and update MCP client dependencies by @appleboy in #589
• refactor(transports): remove SSE server mode by @appleboy in #558
• feat: add support for custom HTTP headers in MCP client and docs by @appleboy in #596
• refactor: standardize "ui-type" terminology across codebase by @appleboy in #598
• fix: Add lastErr field to agent and return error stored there from UI by @noahlwest in #595
• k8s-bench: Add support for MCP client mode by @dshnayder in #601
• k8s-bench: Fix for the task "debug-app-logs" by @dshnayder in #602
• feat: added small timer code for tracking tasks in k8s-bench by @ShubyM in #603

New Contributors

• @dshnayder made their first contribution in #601

Full Changelog: v0.0.26...v0.0.27

v0.0.26

What's Changed

• fix: remove comment on krew.yaml by @ShubyM in #537
• feat: Add --show-failures flag to analyze-evals, update scripts by @noahlwest in #542
• feat: streamable-http server support by @ShubyM in #545
• feat: Add timer to eval runs by @noahlwest in #547
• fix: auth directive being ignored in mcp.yaml config by @tdubenirmata in #546
• fix: handle empty LLM response by @droot in #551
• feat: add llm request/response tracing, http client timeout by @noahlwest in #552
• feat: Add ability to create new kind cluster for k8s-bench run by @noahlwest in #553
• Made the system prompt to be human interaction aware by @droot in #555
• fix: improve the logic for looking up hpa resource by @droot in #561
• feat: fix list-images-for-pods issues by @noahlwest in #557
• docs: standardize and clarify Markdown documentation formatting by @appleboy in #560

New Contributors

• @tdubenirmata made their first contribution in #546

Full Changelog: v0.0.25...v0.0.26

v0.0.25

What's Changed

• fix: eval hangs by @ShubyM in #529
• feat: e2e test for clear by @ShubyM in #513
• Fix/ci periodic by @zvdy in #530
• update k8s-bench results by @noahlwest in #533
• feat: add krew releaser step by @ShubyM in #532

Full Changelog: v0.0.24...v0.0.25

v0.0.24

What's Changed

• feat: meta query tests by @ShubyM in #506
• Bedrock streaming minimal by @augustintsang in #496
• feat: make installation directory configurable via INSTALL_DIR env var by @appleboy in #502
• feat: add ci job for verifying mocks by @ShubyM in #514
• feat: Fix and refactor setup-dev-cluster verification script by @noahlwest in #522
• fix: GetLatestSession() does not get latest session by @noahlwest in #523
• fix: address statefulset-lifecycle and list-images-for-pods consistent failures by @noahlwest in #525
• fix: make multi-container-pod-communication eval less flaky by @noahlwest in #524
• feat: always ask for approval for non-readonly commands by @droot in #510
• fix(mcp): handle object schemas without properties field by @bharats97 in #521
• [Feature]: Support Ctrl+D to exit interactive mode by @prasad89 in #526
• fix: install mockgen before releasing by @ShubyM in #527
• fix: go releaser by @ShubyM in #528

New Contributors

• @bharats97 made their first contribution in #521

Full Changelog: v0.0.23...v0.0.24

v0.0.23

What's Changed

• docs: update README with latest benchmarks and new sections by @droot in #489
• fix: don't return error on failed initialize by @noahlwest in #492
• Fix Var reference for UI configuration for ui Type by @iomarcovalente in #493
• ci: make periodic eval run sequential by @droot in #499
• feat: Add resume-session to metaquery handling by @noahlwest in #480
• eval: Add canary deployment eval by @noahlwest in #472
• docs: enhance and clarify MCP and Bedrock documentation by @appleboy in #503
• feat: Add save-session metaquery and ask to save on exit by @noahlwest in #484
• do not ask uses to save session on exit by @droot in #509
• Adding DeepWiki badge to Github page by @nisranjan in #505
• feat(web): Add support for multiline input by @hglzm in #491

New Contributors

• @iomarcovalente made their first contribution in #493
• @hglzm made their first contribution in #491

Full Changelog: v0.0.22...v0.0.23

v0.0.22

What's Changed

• fix: vertexai dropping kubectl by @ShubyM in #450
• Add AWS Bedrock provider with single-file consolidation by @augustintsang in #435
• eval: Add blue/green traffic switch eval by @noahlwest in #455
• feat: show tool call output in terminal ui by @ShubyM in #451
• Add session persistence for Gemini by @noahlwest in #443
• fix: close terminal after 'exit' or 'quit' by @noahlwest in #453
• Doc and sample files for Issue 257 by @nisranjan in #457
• feat: add insecure option and robust handling for GitHub downloads by @appleboy in #467
• feat: Remove goodbye message, make sessions metaquery use table format by @noahlwest in #475
• fix: update terminal tool call output to 250 by @ShubyM in #477
• feat: Added --show-tool-output to display tool's output by @droot in #478
• chore: ignore .air.toml by @droot in #479
• agent: ignore empty query by @droot in #485

New Contributors

• @augustintsang made their first contribution in #435
• @nisranjan made their first contribution in #457

Full Changelog: v0.0.20...v0.0.22

v0.0.21

What's Changed

• fix: vertexai dropping kubectl by @ShubyM in #450
• Add AWS Bedrock provider with single-file consolidation by @augustintsang in #435
• eval: Add blue/green traffic switch eval by @noahlwest in #455
• feat: show tool call output in terminal ui by @ShubyM in #451
• Add session persistence for Gemini by @noahlwest in #443
• fix: close terminal after 'exit' or 'quit' by @noahlwest in #453
• Doc and sample files for Issue 257 by @nisranjan in #457
• feat: add insecure option and robust handling for GitHub downloads by @appleboy in #467
• feat: Remove goodbye message, make sessions metaquery use table format by @noahlwest in #475
• fix: update terminal tool call output to 250 by @ShubyM in #477
• feat: Added --show-tool-output to display tool's output by @droot in #478

New Contributors

• @augustintsang made their first contribution in #435
• @nisranjan made their first contribution in #457

Full Changelog: v0.0.20...v0.0.21

v0.0.20

What's Changed

• fix: surface error from mcp tool calls by @ShubyM in #444

Full Changelog: v0.0.19...v0.0.20