Harden CI image build and add AI code-review job - #128
Merged
Conversation
Add unattended-upgrades removal and apt-daily/apt-daily-upgrade timer disable to the silk CI dependencies component before its first apt-get, so stock Ubuntu background upgrades cannot hold the dpkg lock during the image build or at runner boot. Bump image_recipe_version to 1.0.12 so a fresh AMI is baked with the changed component content.
|
Workflow [PR], commit [6257717] Code ReviewResult: ✅ No issues found What changed: Hardens CI image creation against background apt lock contention, bumps the image recipe, and adds a dedicated Bedrock-enabled ARM runner pool. It also adds a non-blocking AI code-review job and repository-specific review guidance to the pull-request workflow. No issues found. The previously raised apt-service race is addressed by explicitly stopping both active apt services before purging |
Wire praktika's native AI code-review job into the PR workflow as CODE_REVIEW_JOB, running praktika review against an OpenAI model on Bedrock with allow_failure so a review hiccup never blocks merge and gh auth so it can post its review. Add a dedicated arm-small-bedrock runner pool carrying a scoped bedrock:InvokeModel grant via ext[iam_statements], its SMALL_ARM_BEDROCK label, and a silk-specific review prompt so only this pool's role can reach the Bedrock model API.
maxknv
force-pushed
the
ci-image-hardening-ai-review
branch
from
August 21, 2026 16:24
4047e27 to
6257717
Compare
vadimskipin
approved these changes
Aug 22, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Mirrors two upstream changes into silk's CI, adapted to silk's conventions.
Image-build hardening (mirrors ClickHouse/praktika#141)
unattended-upgradesremoval and disable theapt-daily/apt-daily-upgrade/unattended-upgradesunits in the silk CI dependencies component, before its firstapt-get. This stops stock Ubuntu background upgrades from holding the dpkg lock during the image build and, once the AMI is baked, at runner boot. Both commands carry|| trueso a missing package or transient apt failure never aborts the build.image_recipe_version1.0.11->1.0.12so a fresh AMI is actually baked with the changed component content.Note: praktika is consumed here as a pinned wheel, so the upstream fixes to praktika's own
gh.py(Pages-root clean) and base_ubuntu_setup_componentare not vendored in silk. This applies the same hardening to silk's own Ubuntu setup component instead.AI code-review job
CODE_REVIEW_JOBrunningpraktika reviewagainst an OpenAI model on Bedrock, withallow_failure=True(a review hiccup never blocks merge) andenable_gh_auth=True(so it can post its review), wired into thePRworkflow.RunnerLabels.SMALL_ARM_BEDROCKlabel and a dedicatedarm-small-bedrockrunner pool. Based on silk's existingarm-smallpool (keeping the locked-downallow_all_*=Falseposture andsystem_logs), plus a scopedbedrock:InvokeModelIAM statement merged intoext["iam_statements"], so only this pool's role can reach the Bedrock model API. General job runners stay Bedrock-less.ci/prompts/code_review.md: fiber-scheduler concurrency and lock-free correctness, hot-path performance discipline, and the errno-only error model.To confirm before merge
global.openai.gpt-5.6-solBedrock inference profile is reachable fromeu-north-1(Bedrock model availability is region-specific).