zilliztech/deep-searcher

What it solves

DeepSearcher addresses the challenge of performing accurate search, evaluation, and reasoning over private enterprise data. It prevents the need to rely solely on general LLM knowledge by allowing organizations to utilize their internal documents while maintaining data security.

How it works

The system combines Large Language Models (LLMs) with Vector Databases (such as Milvus or Zilliz Cloud) to create a retrieval-augmented generation pipeline. It allows users to load local files or crawl websites, convert that data into embeddings using a variety of supported embedding models, and store them in a vector database. When a query is made, the system retrieves relevant private data and uses a configured LLM to generate a precise answer or a comprehensive report.

Who it’s for

This tool is designed for enterprises focusing on knowledge management, developers building intelligent Q&A systems, and anyone needing high-accuracy information retrieval from private datasets.

Highlights

• Broad Model Support: Compatible with a wide range of LLMs including OpenAI, DeepSeek, Claude, Gemini, and local models via Ollama.
• Flexible Embeddings: Supports numerous embedding providers such as OpenAI, VoyageAI, and Pymilvus built-in models.
• Enterprise-Ready Storage: Integrated with Milvus and Zilliz Cloud for efficient data partitioning and retrieval.
• Hybrid Data Loading: Supports loading data from local files and integrates web crawling capabilities.