Design a RAG (Retrieval-Augmented Generation) pipeline for:

**Use Case:** [chatbot/search/Q&A/etc.]
**Knowledge Base:** [documents/database/API/etc.]
**Scale:** [document count, query volume]
**Quality Needs:** [precision, recall, latency requirements]

Create a complete RAG architecture:

1. **Ingestion Pipeline**
   - Document parsing strategy
   - Chunking approach (size, overlap, semantic)
   - Metadata extraction
   - Embedding model selection

2. **Vector Store Design**
   - Store selection (Pinecone/Weaviate/pgvector/etc.)
   - Index configuration
   - Partitioning strategy
   - Update/refresh patterns

3. **Retrieval Strategy**
   - Query processing (expansion, rewriting)
   - Hybrid search (vector + keyword)
   - Re-ranking approach
   - Context window management

4. **Generation Layer**
   - Prompt template with retrieved context
   - Citation handling
   - Hallucination prevention
   - Response formatting

5. **Quality Assurance**
   - Relevance scoring
   - Answer evaluation metrics
   - Feedback loop design
   - Human review integration

6. **Production Considerations**
   - Caching strategy
   - Error handling
   - Monitoring/logging
   - Cost optimization

Output as an implementable architecture with technology recommendations.