zilliztech/deep-searcher
Open source deep research system that combines LLMs with Milvus vector databases for autonomous reasoning and information retrieval on private enterprise data.
Velocity · 7d
+16
★ / day
Trend
→steady
star history
DeepSearcher is an agentic RAG (Retrieval Augmented Generation) framework designed for deep research on private data. It combines multiple LLMs (OpenAI o3, DeepSeek, Claude, Llama, Qwen, Grok) with vector database retrieval (Milvus, Zilliz Cloud) to perform multi-step reasoning, evaluation, and answer generation. The system supports document loading, flexible embedding models, and is designed for enterprise knowledge management and intelligent Q&A scenarios.