council-gate

Automated quality gate using multi-model consensus for CI/CD pipelines.

When to Use

• Add AI-powered quality checks to GitHub Actions
• Automate PR approval workflows
• Gate deployments on multi-model verification
• Enforce quality standards in pipelines

Exit Codes

| Code | Verdict | CI/CD Behavior | |------|---------|----------------| | 0 | PASS | Pipeline continues | | 1 | FAIL | Pipeline fails | | 2 | UNCLEAR | Pipeline pauses for human review |

Transcript Location

All deliberations are saved for audit:

.council/logs/{timestamp}-{hash}/
├── request.json      # Input snapshot
├── stage1.json       # Model responses
├── stage2.json       # Peer reviews
├── stage3.json       # Synthesis
└── result.json       # Final verdict

GitHub Actions Integration

name: Council Quality Gate

on:
  pull_request:
    branches: [main, master]

jobs:
  council-gate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: Install LLM Council
        run: pip install 'llm-council-core[http]'

      - name: Run Council Gate
        env:
          OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
        run: |
          llm-council gate \
            --snapshot ${{ github.sha }} \
            --rubric-focus Security \
            --confidence-threshold 0.8

      - name: Upload Transcript
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: council-transcript
          path: .council/logs/

Configuration

| Parameter | Default | Description | |-----------|---------|-------------| | confidence_threshold | 0.7 | Minimum confidence for PASS | | rubric_focus | null | Focus area (Security, Performance) | | timeout | 300s | Maximum execution time | | tier | balanced | Council tier (quick, balanced, high) |

Output Schema

{
  "verdict": "pass",
  "confidence": 0.85,
  "blocking_issues": [],
  "rationale": "All models agreed...",
  "exit_code": 0,
  "transcript_path": ".council/logs/2025-12-31T...",
  "partial": false,
  "timeout_fired": false,
  "completed_stages": ["stage1", "stage2", "stage3"]
}

Under LLM_COUNCIL_STRUCTURED_FINDINGS=true (ADR-051) the response also carries a typed findings array (all severities) and the verdict is computed as fail iff any critical finding exists. blocking_issues is then the critical subset. Gate on the verdict / exit_code, not on blocking_issues == [] (empty even on real FAILs on the default prose-scrape path).

Timeout Handling (ADR-040)

If timeout_fired: true, the gate timed out before completing all stages. This returns exit code 2 (UNCLEAR), pausing the pipeline for human review. Check completed_stages to see how far it got. Consider using --tier quick for faster gate checks (the tier sets the timeout budget; there is no separate --timeout flag).

Example Usage

# Basic gate check
llm-council gate --snapshot $(git rev-parse HEAD)

# Security-focused gate
llm-council gate --snapshot $(git rev-parse HEAD) --rubric-focus Security --confidence-threshold 0.9

# Quick tier for faster feedback
llm-council gate --snapshot $(git rev-parse HEAD) --tier quick

Progressive Disclosure

• Level 1: This metadata (~200 tokens)
• Level 2: Full instructions above (~800 tokens)
• Level 3: See references/ci-cd-rubric.md for CI/CD-specific scoring

Related Skills

• council-verify: General verification
• council-review: Code review with feedback