Knowledge Systems
Local RAG and knowledge management infrastructure — operational.
Overview
721K+ document chunks across 9 domains, queryable via Claude API with CUDA-accelerated embeddings.
Architecture
Paperless-ngx → Document Export → ChromaDB (GPU) → Query API → Claude
Storage
| Component | Location | Size |
|---|---|---|
| ChromaDB | /mnt/storage/knowledge-rag/chroma_db | 16GB+ |
| Symlink | ~/projects/knowledge-rag/chroma_db | → /mnt/storage |
| Available | /mnt/storage | 3.9TB |
Migrated 2026-02-02 to prevent root drive exhaustion
Status
- ✅ Local RAG pipeline (operational)
- ✅ Document ingestion (multi-corpus batch processing)
- ✅ Semantic search across all domains
- ✅ Integration with Dreadbot (Clawdbot)
- ✅ Cross-domain querying (Greek ↔ Science/Medicine)
- ❌ Local LLMs removed (Claude API only)
Corpora
| Domain | Chunks |
|---|---|
| science | 554,335 |
| engineering | 87,746 |
| tech | 43,067 |
| math | 26,311 |
| knowledge_base | 7,127 |
| greek | 1,135 |
| esoteric | 846 |
| quantum_biology | 555 |
| biohacking | 153 |
Total: 721,275 chunks
Documentation
- RAG Setup Notes — Full technical documentation
Related
- PatentBot — Patent discovery system built on this
- Pharmakon Mining — Cross-axis Greek → Science patent discovery
- Dreadbot — AI assistant with RAG access