local-deep-researchBenchmark45/100 via “rag-based private document indexing and retrieval”
Local Deep Research achieves ~95% on SimpleQA benchmark (tested with GPT-4.1-mini). Supports local and cloud LLMs (Ollama, Google, Anthropic, ...). Searches 10+ sources - arXiv, PubMed, web, and your private documents. Everything Local & Encrypted.
Unique: Implements RAG system with per-user encrypted storage of documents and embeddings, enabling private document search without external vector databases. Document indexing is integrated into research workflow, allowing seamless combination of public source results with private document retrieval in single research execution.
vs others: Simpler deployment than external vector databases (Pinecone, Weaviate) by storing embeddings in encrypted SQLCipher, while maintaining semantic search capability through local or cloud embedding models.