Repository navigation
ci(bot): enable distillation and cross-review - #1067
Conversation
Signed-off-by: Vu Anh Phung <vu.phung@databricks.com>
There was a problem hiding this comment.
🔀 Cross-reviewed: two independent reviews (Claude Opus 5.5 · GPT-6.1 Terra), checked by a judge model (Claude Sonnet 5.5).
Verdict: 1 Nit
Looks good, with no blocking issues and one nit. The reviewer now runs the base-branch workflow code, checks out the PR head as a separate read-only tree and rejects closed or fork PRs. I spot-checked lesson provenance against the frozen archive and it matches.
Other findings
- ⚪ Nit — In lessons.md, three section headings (
## Results and lifecycle,## Logging and concurrency,## Verification and bot workflow) directly follow a list item with no blank line. Add a blank line before each, unless the engine regenerates the file in this layout. (Found by Claude Opus 5.5 · confirmed by the judge)
Signed-off-by: Vu Anh Phung <vu.phung@databricks.com>
There was a problem hiding this comment.
Verdict: 1 Low
Looks good overall, with one low concern: some seeded lessons don't list all of their archive sources. The reviewer-workflow change is sound. It runs trusted code from the base SHA, rejects fork and closed PRs on dispatch, keeps the PR head in a separate review-content checkout without credentials, and runs python -m from the trusted workspace root. I couldn't check whether engine 6014478 still needs author.knowledge_log now that it's removed from the config, because the engine source isn't available here.
Signed-off-by: Vu Anh Phung <vu.phung@databricks.com>
What type of PR is this?
Description
Apply kernel #476's onboarding: enable distill mode with learning-PR auto-merge off, and pin the bot engine to 6014478 for two independent reviewers plus a judge. Manual reviews use trusted workflow code and a separate resolved PR checkout.
Preserve all 72 rules from #1018 at 257decd as a frozen archive and seed 47 lessons. New learning uses dated journal files and a fresh rolling branch.
The required
Kernel E2Eresult now has its own Actions job that checks the existing test outcomes. This prevents the result from disappearing when GitHub supersedes an unrelated suite containing a manually posted check.How is this tested?
98 focused engine tests passed. Config, archive preservation, lesson routing/source checks, seven reviewer-input cases, Bash syntax and actionlint passed. Live cross-review approved the current head with no findings. Kernel gate: actionlint, nine outcome scenarios, and the live required job passed.
Related Tickets & Documents
https://github.com/databricks/databricks-sql-kernel/pull/476
After merge, run learning once and inspect its first journal/lessons PR. Check #1018 for newer entries before retiring that legacy PR.