• feat: add diagnose_ci tool for CI failure diagnosis #192 👈 (View in Graphite)
• chore: update copier template to 0.29.3 and pin fastmcp>=3.0.0 #191
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

Greptile Summary

Adds a new diagnose_ci MCP tool that consolidates CI failure diagnosis into a single call, replacing the typical 3-5 sequential gh commands agents currently need to run. The tool finds failed workflow runs, identifies failed jobs and steps, and extracts actionable error lines from logs while filtering out CI noise like timestamps and cleanup messages.

Key changes:

• New business logic in src/codereviewbuddy/tools/ci.py with helper functions for log parsing, error extraction, and run discovery
• CIJobFailure and CIDiagnosisResult models added to models.py for structured output
• Tool registration in server.py with prerequisite validation that correctly requires repo when run_id is provided without workspace detection
• Comprehensive test coverage with 23 tests covering unit helpers and integration scenarios

Issue found:

• Line 257 in ci.py has error matching logic that may duplicate all error lines across multiple failed jobs when no job-specific matches are found (the or error_lines fallback). This could make the diagnosis output confusing when multiple jobs fail with generic error messages that don't contain job names.

Confidence Score: 4/5

• Safe to merge with minor logic improvement recommended
• The implementation is well-structured with comprehensive test coverage and follows existing patterns. The bug fix for prerequisite validation is correct. The only issue is the error line duplication logic which could cause confusion but won't break functionality.
• Pay close attention to src/codereviewbuddy/tools/ci.py:257 - the error matching fallback logic

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[diagnose_ci tool call] --> B{run_id provided?}
    B -->|Yes| C[Use specified run_id]
    B -->|No| D[Find latest failed run]
    
    D --> E{pr_number provided?}
    E -->|Yes| F[Get branch from PR]
    E -->|No| G[Use current branch]
    
    F --> H[gh run list --branch]
    G --> H
    H --> I[Filter for failed runs]
    
    C --> J[Get run details]
    I --> J
    
    J --> K[Extract failed jobs]
    K --> L{Failed jobs found?}
    
    L -->|No| M[Return error: no jobs failed]
    L -->|Yes| N[Get failed logs]
    
    N --> O[Clean log lines]
    O --> P[Extract error lines with context]
    P --> Q{Multiple failed jobs?}
    
    Q -->|Yes| R[Match errors to job names]
    Q -->|No| S[Assign all errors to job]
    
    R --> T[Build CIJobFailure objects]
    S --> T
    T --> U[Return CIDiagnosisResult]

_{Last reviewed commit: 47bd4c3}

Merge activity

• Feb 19, 12:46 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Feb 19, 12:47 PM UTC: Graphite rebased this pull request as part of a merge.
• Feb 19, 12:48 PM UTC: @ichoosetoaccept merged this pull request with Graphite.

Found 1 test failure on Blacksmith runners:

Failure