ci(kai): add KaiBench evaluation workflow to CI pipeline [AI-2588]#399
Open
jordanrburger wants to merge 8 commits intomainfrom
Open
ci(kai): add KaiBench evaluation workflow to CI pipeline [AI-2588]#399jordanrburger wants to merge 8 commits intomainfrom
jordanrburger wants to merge 8 commits intomainfrom
Conversation
Add kaibench.yml reusable workflow that builds MCP server from the PR branch, starts the full stack (MCP server + kai-assistant + Postgres + Redis), and runs KaiBench evaluations. Triggered as a non-blocking job in ci.yml after the build passes (same-repo pushes only). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
Restricts GITHUB_TOKEN to read-only contents access to satisfy CodeQL security policy. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
cli.py imports requests (for requests.JSONDecodeError handler) but it was not declared in pyproject.toml. This caused test collection to fail on Python 3.11 in CI where no transitive dependency pulls it in. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The repo was renamed from keboola/keboola-mcp-server to keboola/mcp-server, causing the if condition to evaluate false. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Always use pre-built kai-assistant Docker image tag instead of building from the UI repo source. This removes the need for UI repo access and TURBO_TOKEN. The image tag is provided via vars.KAI_ASSISTANT_IMAGE_TAG. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Instead of requiring a static image tag, the workflow now queries Docker Hub for the latest production-kai-assi-* tag when none is provided. This ensures CI always tests against the newest kai-assistant without manual variable updates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove silent curl flags to surface errors when Docker Hub API calls fail during kai-assistant tag resolution. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Instead of running the full evaluation locally (which needs UI repo access, TURBO_TOKEN, and registry credentials), dispatch the eval to the KaiBench repo where all secrets are centralized. Results are posted back as a commit status on the MCP server repo. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Linear: AI-2588
Change Type
Summary
Adds automated KaiBench evaluations to the CI pipeline so MCP server changes are tested against the full AI agent stack before merging. This catches regressions where tool description changes, argument schema modifications, or behavior changes break the agent's ability to answer questions correctly.
How it works:
kaibench.ymlreusable workflow builds the MCP server Docker image from the PR branch, starts the full stack (MCP server + kai-assistant + Postgres + Redis), clones KaiBench, and runs evaluations against 3 question types (Data Analysis Query, Configuration Reasoning, Storage Object Reasoning)ci.ymlcalls this workflow as a non-blocking job after the build passes (same-repo pushes only)workflow_dispatchfor manual runs with configurable question typesBefore merging — secrets required:
KAIBENCH_REPO_TOKENkeboola-rnd/KaiBenchKAIBENCH_STATIC_TOKENKAIBENCH_MANAGEMENT_TOKENKAIBENCH_API_URLhttps://connection.canary-orion.keboola.devDOCKERHUB_TOKENKAI_GOOGLE_VERTEX_CREDENTIALSKAI_GOOGLE_VERTEX_PROJECTKAI_GOOGLE_VERTEX_LOCATIONTURBO_TOKENTesting
Streamable-HTTPtransports)N/A — this is a CI-only change (GitHub Actions workflows). Testing plan:
kaibench.ymlviaworkflow_dispatchto validate the full stack starts and evals runsummary.json+results.jsonlare presentChecklist
🤖 Generated with Claude Code