Add AI-enhanced webhook mode for intelligent pipeline analysis

Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.

Close this issue

Summary

GitLab webhooks currently provide limited information about pipeline failures, requiring AI-powered analysis tools to make multiple API calls to gather necessary context. This creates latency and complexity for intelligent CI/CD monitoring tools.

Proposal

Introduce an "AI Analysis Mode" for webhooks that includes:

Failed job logs - Direct inclusion of logs from failed jobs
Code context - Information about recent commits and changed files
Structured error data - Pre-categorized error summary for AI consumption
Runner metadata - Infrastructure context for better analysis

Motivation

The rise of AI-powered DevOps tools (like GitHub Copilot for CI/CD, CircleCI's AI insights, and custom solutions) requires richer webhook payloads for effective analysis.

Current Pain Points

Tools must make 3-5 additional API calls per failed pipeline
Increased latency (2-5 seconds) before analysis can begin
Rate limiting issues when monitoring many projects
Incomplete context leads to less accurate AI analysis

Use Cases

Automated Error Resolution: AI tools that automatically create fix MRs
Intelligent Retry Logic: Systems that retry only transient failures
Pattern Recognition: Tools that learn from error patterns across projects
Smart Notifications: AI that knows when to alert vs auto-fix

Example Enhanced Payload

{
  "object_kind": "pipeline",
  "object_attributes": { "status": "failed" },
  "failed_jobs_logs": [{
    "job_id": 12345,
    "job_name": "test",
    "logs": "ModuleNotFoundError: No module named 'pandas'"
  }],
  "ai_metadata": {
    "error_summary": {
      "dependency_errors": 1,
      "syntax_errors": 0,
      "test_failures": 0
    }
  }
}

Benefits

Performance: Reduce API calls by 80% for AI tools
Accuracy: Better context = better AI analysis
Adoption: Easier integration for AI tool developers
Innovation: Enable new types of intelligent automation

Implementation

MR: !192569 (closed) - Initial backend implementation

Hackathon Context

This feature was developed as part of the Google Cloud + GitLab Hackathon 2025, where we built AI Pipeline Guardian - an AI-powered tool that demonstrates the need for this enhancement.