CI for AI-generated code · Apache 2.0

eslint + code review, but for AI.

Every AI-generated PR scored on production readiness. Not just "do tests pass" — "would you merge it?" Add to your pipeline in 2 minutes.

Add to CI How it works

2 minutes to quality-gated PRs

Add QualBench to your GitHub Actions. Every AI-generated PR gets scored and reviewed automatically.

# .github/workflows/qualbench.yml
name: QualBench
on: [pull_request]
jobs:
  quality-check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: semcod/qualbench-action@v1
        with:
          tool: prollama
          fail_on_score: 70
🧠 QualBench Review
Quality Score: 78/100

  ❌ Complexity increased (+12%)
  ⚠ Security: 1 new finding
  ✔ Tests pass, no regressions
Verdict: needs_review

60 seconds to your first score

# Install and run
pip install qualbench
qualbench quickstart

# Score current diff
qualbench run --tool prollama

# Machine-readable output
qualbench run --json

Six dimensions of production readiness

25%
Correctness
Tests pass. No regressions. Verified via pytest.
25%
Mergeability
Would a senior dev approve? Blind review, 1–5 scale.
15%
Security
No new vulnerabilities. Bandit delta scoring.
15%
Code quality
Complexity should not increase. No dead code.
10%
Iterations
Fewer attempts to converge = better engineering.
10%
Cost efficiency
USD per successful patch. The metric CFOs care about.

Leaderboard

Real-time rankings by quality score and cost efficiency. Refresh

Loading...

One format everywhere

CLI, API, GitHub Action — same portable JSON schema. No translation, no surprises.

{
  "quality_score": 78,
  "verdict": "needs_review",
  "dimensions": { ... },
  "top_issues": ["complexity_increase"]
}
Schema docs