Buzo scans your vector database for data quality issues — stale content, truncated data, embedding model mismatches, duplicates, and contradictions — and blocks deploys when quality drops below your threshold.

Quick Start

npx buzo-cli scan \
  --collection <collection-id> \
  --api-key <your-buzo-api-key> \
  --api-url https://api.buzo.ai \
  --threshold 70

What it does

1. Connects to your Buzo account
2. Scans your vector collection through 6 detection layers
3. Auto-detects your AI stack from package.json / requirements.txt
4. Returns PASS/FAIL based on quality threshold
5. Exits with code 1 on failure (blocks your pipeline)

Usage in CI/CD

GitHub Actions

- name: Buzo Quality Gate
  run: |
    npx buzo-cli scan \
      --collection ${{ secrets.BUZO_COLLECTION_ID }} \
      --api-key ${{ secrets.BUZO_API_KEY }} \
      --api-url https://api.buzo.ai \
      --threshold 70

GitLab CI

quality-gate:
  stage: test
  script:
    - npx buzo-cli scan --collection $BUZO_COLLECTION_ID --api-key $BUZO_API_KEY --threshold 70

Options

| Flag | Short | Default | Description | |------|-------|---------|-------------| | --collection | -c | required | Collection UUID to scan | | --api-key | -k | $BUZO_API_KEY | Buzo API key | | --api-url | -u | $BUZO_API_URL | Buzo API URL | | --threshold | -t | 70 | Minimum quality score (0-100) | | --fail-on-critical | | true | Fail if critical issues found | | --scan-mode | | FULL | FULL, WATCH (delta), DEEP | | --format | | summary | summary, json | | --warn-only | | false | Report but don't fail pipeline | | --no-detect-stack | | | Skip auto-detection of AI stack |

Stack Detection

When run inside a project, buzo-cli automatically reads package.json, requirements.txt, and .env.example to detect your AI stack:

• Frameworks: LangChain, LlamaIndex, CrewAI, AutoGen
• LLM Providers: OpenAI, Anthropic, Google Gemini, Cohere
• Vector DBs: Pinecone, Qdrant, ChromaDB, pgvector, Weaviate, Milvus
• Databases: Supabase, MongoDB, Redis
• Tools: SerpAPI, Tavily, Slack, HubSpot

The detected stack is sent to Buzo and visualized in the infrastructure graph.

Exit Codes

| Code | Meaning | |------|---------| | 0 | Quality gate passed | | 1 | Quality gate failed | | 2 | Error (connection, auth, timeout) |

Example Output

  🤿 Buzo — Vector Quality Gate

  Collection:  prod-kb (Pinecone, 142K vectors)
  Scan mode:   FULL

  Pipeline results:
  ✓ Metadata & Provenance         0 issues
  ✗ Content Patterns              3 stale
  ✗ Relevance Analysis            2 model mismatch
  ✓ Duplicate Detection           0 collisions
  ✓ Contradiction Analysis        0

  ─────────────────────────────────────
  Quality Score:    85 / 100
  Threshold:        70
  Status:           PASS ✓
  ─────────────────────────────────────

  Detected Stack:
    Framework:  LangChain
    LLM:        OpenAI
    Embeddings: OpenAI Embeddings
    Vector DB:  Pinecone

Links

• Website
• Documentation

Zero Dependencies

This CLI has zero runtime dependencies. It uses only Node.js built-in modules.