[dprince]

Select psycopg2 explicitly for PostgreSQL database engines

SQLAlchemy 2.1 changed the default driver for postgresql:// URLs from
psycopg2 to psycopg (version 3). LCS depends on psycopg2-binary, so the
implicit driver selection now fails with ModuleNotFoundError for psycopg
when initializing a PostgreSQL-backed runtime database.

Use postgresql+psycopg2:// to select the existing dependency explicitly
and update the database unit test to expect that URL.

This fixes startup for PostgreSQL-backed deployments, including those
managed by the OpenStack Lightspeed operator, which configures the LCS
runtime database to use PostgreSQL. SQLite-backed deployments do not
exercise this path.

Jira: LCORE-5372

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: lightspeed-core/lightspeed-stack
• Branch: release/0.7
• File: .coderabbit.yaml
• Commit: 3fc133c
• Layers: Repository YAML
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: a15c3c8a-073a-431b-885e-4865bfff9318

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

[tisnik]

LGTM

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 5:52 PM UTC · Completed 6:03 PM UTC

Commit: faef1b0 · View workflow run →

Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $4.47

[fullsend-ai-retro]

Retro: PR #3116 — LCORE-5372: Select psycopg2 explicitly for PostgreSQL database engines (backport to release/0.7)

Workflow quality: Clean — nothing to improve. This PR is a straightforward backport of PR #3107 (already merged to main) to the release/0.7 branch. It changes the PostgreSQL connection URL from postgresql:// to postgresql+psycopg2:// to accommodate SQLAlchemy 2.1's default driver change, with a matching test update. One human reviewer (tisnik) approved. No agent involvement — no code, review, or fix agents ran. Merged ~26 minutes after creation.

No proposals filed — all findings map to existing open issues. This retro is a near-duplicate of the retro already completed on the parent PR #3107 (retro comment). The same pattern recurs: retro dispatched on a human-only PR merge despite the repo's .fullsend/harness/retro.yaml restricting retro to explicit /fs-retro commands. The root causes and existing coverage:

• Extend retro skip logic to merged PRs with zero fullsend agent involvement fullsend-ai/fullsend#6297 ("Extend retro skip logic to merged PRs with zero fullsend agent involvement") — directly covers this: no agents participated, yet retro consumed ~$2.88 analyzing a human-only workflow. Multiple duplicates exist: Skip retro dispatch for merged PRs with zero meaningful agent involvement fullsend-ai/fullsend#5399, Skip retro dispatch when no prior agent activity occurred on the PR fullsend-ai/fullsend#7803, Skip retro dispatch when no agents participated in the PR workflow fullsend-ai/fullsend#8084, Skip retro dispatch for human-authored PRs where no fullsend agent stage ran fullsend-ai/fullsend#8673.
• feat(dispatch)!: make harness-dispatch primary and remove bash routing fullsend-ai/fullsend#2902 ("Make harness-dispatch primary and remove bash routing") — the bash routing hardcodes STAGE="retro" for closed PRs, bypassing the CEL trigger evaluation that would respect the harness override. CEL config/route outcomes were both skipped in this run.
• Allow config.yaml to override harness CEL triggers fullsend-ai/fullsend#7719 ("Allow config.yaml to override harness CEL triggers") — the harness trigger override in .fullsend/harness/retro.yaml was not evaluated during dispatch routing.

Repo-local PRs #3112 and #3070 (both still open) aim to work around this by moving the trigger restriction inline into config.yaml, but the platform-level root cause remains unresolved.

Additional data point for fullsend-ai/fullsend#5376 ("Skip retro dispatch for automated cherry-pick PRs that reference an already-analyzed original PR"): PR #3116 is a manual backport of PR #3107, which was already retro-analyzed. The backport introduces no new code — identical 2-line diff — yet triggered a full retro analysis. While #5376 targets automated cherry-picks specifically, this case shows the same waste pattern for manual backports.

Observation — pre-existing E2E failures on release/0.7: Three E2E suites failed (library/ci/default, server/ci/default, server/ci/other), but these are pre-existing and unrelated to this PR's change. Root causes: (1) a hardcoded version assertion expecting 0.7.1 that was not updated when PR #3045 bumped the version to 0.7.2, and (2) CI runner disk space exhaustion during Docker builds. These failures are consistent across recent release/0.7 PRs. The PR was merged despite them.