chunkops 0.3.0

The CI/CD Pipeline for RAG - Detect semantic conflicts before they become hallucinations

ChunkOps CLI

The CI/CD Pipeline for RAG. Detect conflicts before they become hallucinations.

Installation

pip install chunkops

Quick Start

1. Initialize ChunkOps

chunkops init

This creates a chunkops.yaml config file in your project root.

2. Scan Your Documents

chunkops scan

Level 1 (Free): Detects exact duplicates and empty files
Level 2 (Cloud): Semantic conflict detection (requires authentication)

3. Enable Cloud Features

chunkops login

Opens browser for authentication. Enables semantic conflict detection.

4. Add to CI/CD

Add to your .github/workflows/rag-test.yml:

- name: Run ChunkOps CI Check
  run: chunkops ci
  env:
    CHUNKOPS_API_KEY: ${{ secrets.CHUNKOPS_API_KEY }}

Commands

chunkops init

Initialize ChunkOps in your project. Creates chunkops.yaml config file.

chunkops init --docs-path ./data --output-path ./reports

chunkops scan

Scan documents for duplicates and conflicts.

chunkops scan                    # Use config defaults
chunkops scan ./docs            # Scan specific directory
chunkops scan -o report.json    # Custom output file
chunkops scan --verbose         # Verbose output

Output:

• ✅ Valid chunks count
• ⚠️ Exact duplicates (with estimated space saved)
• 🟡 Near duplicates
• 🚨 Semantic conflicts (requires authentication)

chunkops login / chunkops auth

Authenticate with ChunkOps Cloud to enable semantic conflict detection.

chunkops login                  # Opens browser
chunkops auth --api-key chk_... # Use API key directly

chunkops ci

CI/CD mode: Runs silently, outputs JSON, exits with error code on failures.

chunkops ci                     # Use config defaults
chunkops ci ./docs              # Scan specific directory
chunkops ci --no-fail-on-critical  # Don't exit on critical conflicts

Exit Codes:

• 0: No critical conflicts
• 1: Critical conflicts detected (blocks PR merge)

Configuration

The chunkops.yaml file:

docs_path: ./data
output_path: ./chunkops-reports
exact_threshold: 1.0
near_threshold: 0.90
enable_cloud: false
api_url: https://console.chunkops.ai

CI/CD Integration

GitHub Actions

name: RAG Quality Check

on:
  pull_request:
    paths:
      - 'docs/**'
      - 'data/**'

jobs:
  chunkops-check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.11'
      - run: pip install chunkops
      - run: chunkops ci
        env:
          CHUNKOPS_API_KEY: ${{ secrets.CHUNKOPS_API_KEY }}

GitLab CI

chunkops-check:
  image: python:3.11
  script:
    - pip install chunkops
    - chunkops ci
  variables:
    CHUNKOPS_API_KEY: $CHUNKOPS_API_KEY

Output Format

Interactive Mode (chunkops scan)

Beautiful terminal output with colors, progress bars, and summary tables.

CI Mode (chunkops ci)

JSON output only:

{
  "scan_date": "2024-01-15T10:30:00Z",
  "total_files": 14,
  "total_chunks": 245,
  "exact_duplicates": 2,
  "near_duplicates": 5,
  "semantic_conflicts": 3,
  "critical_conflicts": 1,
  "summary": {
    "valid_chunks": 240,
    "exact_duplicates": 2,
    "near_duplicates": 5,
    "conflicts": 3
  }
}

The Two-Tier Model

Level 1 (Free/Local):

• Exact duplicate detection (MD5 hash)
• Empty file detection
• Fast, runs entirely locally

Level 2 (Cloud/Upsell):

• Semantic conflict detection
• Policy contradiction detection
• Requires authentication (chunkops login)

License

MIT

Support

Visit chunkops.ai for more information.