The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	May 26, 2026 11:43pm

PR Summary

Medium Risk
Migration CI path and statement_timeout changes affect how schema deploys run; execution logging is observability-only with a guarded helper.

Overview
Improves workflow execution failure logs by walking Drizzle’s wrapped error.cause chain and attaching driver/Postgres fields (code, severity, detail, etc.) so intermittent DB errors are easier to classify than a bare “Failed query” message.

CI database migrations now run via bun run ./scripts/migrate.ts instead of drizzle-kit migrate, and that script SET statement_timeout = 0 before applying migrations so long-running DDL is not cut off by a server timeout.

^{Reviewed by Cursor Bugbot for commit 63e3927. Bugbot is set up for automated code reviews on this repo. Configure here.}

Greptile Summary

This PR improves observability for workflow execution failures by surfacing the underlying Postgres/driver error that Drizzle wraps in error.cause, and switches the CI migration runner to a custom script with richer diagnostics and a session-level SET statement_timeout = 0 guard.

• describeErrorCause walks the cause chain (depth-bounded to 10, wrapped in try/catch) and extracts the first error carrying a code, then passes it as a structured cause field to logger.error — enabling CloudWatch Log Insights to distinguish connection drops, pool exhaustion, and statement timeouts behind the previously opaque "Failed query" message.
• packages/db/scripts/migrate.ts is now the CI migration runner via .github/workflows/migrations.yml; it disables statement_timeout for the session and prints comprehensive Postgres diagnostic fields on failure via printMigrationError.

Confidence Score: 4/5

Safe to merge; changes are purely additive diagnostics with no behavioral impact on the execution path.

The core execution path is unchanged — describeErrorCause can never throw or mask the original error thanks to its try/catch. The migration script improvement is well-guarded. The only rough edge is the unbounded recursion in printMigrationError, which is harmless in the real failure modes this code targets but worth hardening before it becomes a pattern.

packages/db/scripts/migrate.ts — the recursive printMigrationError helper has no depth guard; worth adding one consistent with the approach used in describeErrorCause.

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[executeWorkflowCore throws] --> B[catch block]
    B --> C[logger.error with error + describeErrorCause]
    C --> D{describeErrorCause}
    D --> E{error instanceof Error?}
    E -- No --> F[return undefined]
    E -- Yes --> G[Walk cause chain up to depth 10]
    G --> H{candidate.code defined?}
    H -- Yes --> I[driver = candidate, break]
    H -- No --> J[advance to candidate.cause]
    J --> G
    G -- no Error found --> K[driver = first Error or undefined]
    I --> L[filterUndefined on driver fields]
    K --> L
    L --> M{driver exists?}
    M -- No --> F
    M -- Yes --> N[Return: name/message/code/severity/detail/routine/errno/syscall]
    N --> O[logger merges as cause into structured JSON log entry]
    F --> P[cause field omitted from log entry]

_{Reviews (1): Last reviewed commit: "log(db): Add db failure cause log messag..." | Re-trigger Greptile}

Superseded by #4749, which isolates the logging change cleanly off staging. The two DB/CI commits on this branch (fix(db): disable statement_timeout, fix(ci): route migration workflow) aren't in staging/main yet and will get their own PR.