离线生存知识库问答系统设计师

You are a systems architect specializing in offline-first AI applications. I need to design an offline survival knowledge base and Q&A system.

Target scenario: [SCENARIO: e.g. emergency preparedness, field research, remote expedition] Hardware constraints: [DEVICE: e.g. Raspberry Pi 5, old laptop, smartphone] Knowledge domains: [DOMAINS: e.g. first aid, navigation, water purification, shelter building]

Design a complete system covering:

1. Architecture

   • Local LLM selection (model size vs capability tradeoff for the hardware)
   • Vector database for knowledge retrieval (SQLite-based options)
   • RAG pipeline optimized for low-resource environments

2. Knowledge Ingestion

   • Sources to scrape/download while online
   • Chunking and embedding strategy for offline search
   • Compression techniques to minimize storage

3. Query System

   • Natural language Q&A with source citations
   • Fallback to keyword search when LLM is too slow
   • Priority routing (medical questions get fastest path)

4. Deployment

   • Docker Compose or single-binary setup
   • Auto-start on boot, power management
   • Simple web UI accessible from any device on local network

5. Update Strategy

   • Sync new knowledge when connectivity is available
   • Delta updates to minimize bandwidth

Provide concrete tool/library recommendations, config files, and a step-by-step setup guide.