PROJECT INTEL
Personal Librarian
A multilingual RAG assistant that answers questions across your whole book and document library.
- ACTIVE SINCE:
- 2025 — present
- STATUS:
- ACTIVE
- FIREPOWER
- 7/10
- ARMOR
- 7/10
- SPEED
- 8/10
- SPECIAL
- 9/10
A personal librarian that has actually read everything on your shelf — and doesn't care what language any of it was written in. An orchestrator/sub-agent architecture splits the work: a query-expansion sub-agent generates five semantically similar queries in the document's source language, the orchestrator judges whether retrieval found enough evidence before answering, and a refiner responds in whatever language you asked in.
Retrieval runs on ChromaDB with hybrid strategies, so a question asked in Persian can surface passages from English sources and come back as one coherent answer.
BATTLE RECORD
- Cross-language retrieval: ask in Persian, find answers in English sources
- Query-expansion sub-agent: five semantic variants in the source language
- Orchestrator checks retrieval sufficiency before answering
- ChromaDB with hybrid retrieval strategies
TECH
- Python
- ChromaDB
- Hybrid retrieval
- Query expansion
- Multi-agent