raglint 0.1.0

Production-ready RAG pipeline evaluation and observability platform

🔍 RAGLint

The Honest RAG Evaluation Platform

RAGLint is the all-in-one evaluation platform for Retrieval-Augmented Generation (RAG) systems. Built with production in mind, it provides comprehensive metrics, a beautiful dashboard, and enterprise-grade security.

✨ Why RAGLint?

Feature	RAGLint	Ragas	Arize Phoenix	DeepEval
Plugin Marketplace	✅ Unique	❌	❌	❌
Local LLM (Ollama)	✅ Native	⚠️ Partial	⚠️ Partial	⚠️ Partial
Dashboard	✅ Excellent	⚠️ Basic	✅ Excellent	⚠️ Good
Real-time Analytics	✅ WebSockets	❌	✅	❌
Security	✅ High	⚠️ Moderate	✅ High	⚠️ Moderate
CI/CD Integration	✅ Native Action	⚠️ Manual	✅	⚠️ Manual
Open Source	✅ MIT	✅ Apache	⚠️ Partial	✅ MIT

🚀 Quick Start

# Install
pip install raglint

# Analyze your RAG pipeline
raglint analyze data.json --smart --provider ollama

# Launch dashboard
raglint dashboard

📊 Features

Core Metrics

• Faithfulness: Verify answers match source documents
• Relevance: Context & answer relevance scoring
• Precision/Recall: Retrieval quality metrics
• Bias Detection: Identify biased responses
• Toxicity: Safety & appropriateness checks
• PII Detection: Security & privacy compliance

Advanced Analytics

• Drift Detection: Monitor metric changes over time
• Cohort Analysis: Compare configurations
• Embedding Visualization: UMAP projections
• Real-time Updates: WebSocket-powered dashboard

Integrations

• ✅ OpenAI, Azure OpenAI, AWS Bedrock
• ✅ Ollama (local LLM)
• ✅ LangChain, LlamaIndex, Haystack
• ✅ Pinecone, Weaviate, Chroma, Qdrant

💡 Example

from raglint.core import RAGPipelineAnalyzer

# Your RAG data
data = [
    {
        "query": "What is machine learning?",
        "retrieved_contexts": ["ML is a subset of AI..."],
        "response": "Machine learning is..."
    }
]

# Analyze
analyzer = RAGPipelineAnalyzer(use_smart_metrics=True)
results = analyzer.analyze(data)

print(f"Faithfulness: {results.faithfulness_scores[0]:.2f}")
print(f"Relevance: {results.semantic_scores[0]:.2f}")

🔌 Plugin Marketplace

Extend RAGLint with custom metrics:

# Browse plugins
raglint plugins list

# Install a plugin
raglint plugins install raglint-pii-advanced

# Create your own
raglint plugins create my-custom-metric

Built-in Plugins:

• Citation Accuracy Checker
• PII Detector
• SQL Injection Detector
• Bias Detector
• Hallucination Confidence

🎯 CI/CD Integration

# .github/workflows/raglint.yml
- name: RAGLint Analysis
  uses: yourusername/raglint-action@v1
  with:
    data-file: 'eval_data.json'
    threshold: 0.7

📈 Dashboard

Beautiful, real-time analytics dashboard:

• 📊 Interactive charts & visualizations
• 🔍 Drill-down into individual runs
• 📉 Drift detection & alerting
• 🎨 Dark mode & modern UI
• 🚀 WebSocket real-time updates

raglint dashboard
# Open http://localhost:8000

🔐 Enterprise-Ready

• Security: RestrictedPython sandbox, code signing
• Scalability: PostgreSQL support, async processing
• Privacy: Self-hosted, local LLM support
• Compliance: PII detection, audit logs

📚 Documentation

• Quick Start Guide
• API Reference
• Cookbooks
  • Pinecone Integration
  • LlamaIndex Advanced
  • Haystack Integration

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

📄 License

MIT License - see LICENSE for details.

⭐ Star Us!

If you find RAGLint useful, please consider starring the repo to help others discover it!

---

Made with ❤️ by the RAGLint team | Website | Discord | Twitter