Offline Knowledge Q&A System Design Consultant

You are an expert in building offline AI knowledge systems. I want to create a local knowledge base Q&A system that requires no internet connection. Please help me design it. Please provide: 1. **Model Selection** - Compare local LLMs suitable for my hardware (Llama 3, Qwen 2.5, Phi-3, Gemma 2) - Recommend quantization levels based on my memory (Q4_K_M, Q5_K_M, etc.) - Select Embedding models based on language requirements 2. **Vector Database** - Compare offline usage scenarios for ChromaDB, Milvus Lite, and LanceDB - Indexing strategies tailored to my document types - Recommendations for chunk size and overlap 3. **RAG Pipeline** - Document ingestion workflow (PDF, MD, TXT, DOCX) - Retrieval strategies (hybrid search, reranking) - Q&A prompt templates with citations 4. **Deployment Plan** - Docker Compose configuration or bare-metal deployment guide - Resource requirements and performance tuning - Backup and update strategies My configuration: - Hardware: [Your hardware - e.g., 16GB RAM, no GPU] - Documents: [Your documents - e.g., 500 PDF technical manuals] - Language: [e.g., Chinese and English] - Use Case: [e.g., Internal technical support]