A production-ready Retrieval-Augmented Generation (RAG) system for intelligent document question-answering with comprehensive evaluation capabilities.
Complete RAG system architecture showing data flow from document upload to answer generation
Detailed sequence diagram showing component interactions and data flow
- Frontend: Streamlit web interface with real-time configuration
- Backend: Dual-mode backend with:
- Streamlit for interactive user experience
- FastAPI for RESTful API access (
/upload,/ask,/evaluate)
- Vector Database: ChromaDB for persistent document embeddings
- LLM: Google Gemini 2.5 Flash for answer generation
- Evaluation: TruLens framework for RAG quality assessment
Main application interface with document upload, Q&A, and evaluation tabs
- Multi-format Document Processing: PDF, TXT with intelligent chunking
- Advanced Retrieval: Semantic search with contextual compression
- AI-Powered Q&A: Context-aware answer generation with source attribution
- Quality Evaluation: Comprehensive metrics for answer relevance and groundedness
- Interactive Interface: Real-time parameter adjustment and conversation history
- Production-Ready: Environment-based configuration and comprehensive logging
- Scalable Architecture: Modular design supporting multiple LLM and embedding providers
- Persistent Storage: ChromaDB with automatic document versioning
- Performance Monitoring: Real-time evaluation metrics and statistics
- Source Attribution: Document-level source tracking and citation
Document processing workflow showing chunking and vector storage
Source attribution and document tracking in Q&A responses
- Processing: Efficient document chunking and vector storage with ChromaDB
- Retrieval: Fast semantic search with configurable similarity thresholds
- Scalability: Modular architecture designed for high-volume document processing
- Evaluation: Comprehensive quality assessment using TruLens framework
- Reliability: Production-ready error handling and session management
| Component | Technology | Purpose |
|---|---|---|
| Frontend | Streamlit | Interactive web interface |
| LLM | Google Gemini 2.5 Flash | Answer generation |
| Vector DB | ChromaDB | Document embeddings storage |
| Processing | LangChain, PyPDF2 | Document processing pipeline |
| Evaluation | TruLens | RAG quality assessment |
| Configuration | Environment Variables | Secure settings management |
-
Install Dependencies:
pip install -r requirements.txt
-
Set API Keys:
export GEMINI_API_KEY="your_gemini_api_key"
-
Run Application:
streamlit run app.py --server.port 5000
-
Access Interface: http://localhost:5000
IntelligentDocumentQA/
├── app.py # Main Streamlit application
├── utils/ # Core utilities
│ ├── config.py # Configuration management
│ ├── document_processor.py # Document loading and chunking
│ ├── vector_store.py # Vector database operations
│ ├── retriever.py # Advanced retrieval logic
│ ├── qa_chain.py # Question answering with LLM
│ └── evaluation.py # RAG evaluation framework
├── .streamlit/config.toml # Streamlit configuration
└── chroma_db/ # Persistent vector database
- Enterprise Knowledge Management: Internal document Q&A systems
- Customer Support: Automated FAQ and documentation assistance
- Research & Analysis: Academic paper and report analysis
- Legal Document Review: Contract and compliance document processing
- Educational Tools: Textbook and course material Q&A systems
- Multi-LLM support (OpenAI, Claude, local models)
- REST API endpoints
- Advanced caching and performance optimization
- Multi-language support
- Enterprise authentication integration
TruLens evaluation dashboard showing comprehensive RAG quality metrics
The system includes comprehensive evaluation using TruLens framework:
- Answer Relevance: Measures how well answers address questions
- Context Relevance: Evaluates retrieval quality and relevance
- Groundedness: Assesses answer factual accuracy against sources
- Real-time Metrics: Live performance monitoring and statistics
Built with ❤️ using modern AI/ML best practices and production-ready architecture.