zilliztech/deep-searcher

What is it?

Key features

• 01Multi-step search, evaluation and reasoning over private documents.
• 02Works with vector databases including Milvus, Zilliz Cloud and Qdrant.
• 03Supports many LLM providers such as OpenAI, DeepSeek, Claude, Gemini and Ollama.
• 04Loads local PDF, text and Markdown files, plus web pages via crawlers like FireCrawl and Crawl4AI.
• 05Exposes a Python API, a command-line interface and a FastAPI web service.

Use cases

• Building an internal Q&A assistant over company documents that must stay in-house.
• Generating research-style reports from a collection of PDFs and notes.
• Searching a knowledge base with a local model through Ollama for fully offline use.
• Prototyping retrieval-augmented generation pipelines with swappable LLMs and embedding models.

Quick start

1. 1

   Create and activate a virtual environment

   python -m venv .venv && source .venv/bin/activate
2. 2

   Install DeepSearcher with pip

   pip install deepsearcher
3. 3

   Set your OPENAI_API_KEY environment variable, then run the quick start Python demo from the README

Things to know

• !Needs an LLM API key (for example OpenAI) or a locally running model such as Ollama.
• !Requires a vector database such as Milvus or Zilliz Cloud, adding setup effort.
• !Web crawling of documents is still under development and needs a FIRECRAWL_API_KEY.

Topics

FAQ

What is deep-searcher?

Open-source deep research tool that searches and reasons over private data using LLMs and vector databases. DeepSearcher is a Python project for question answering and report generation over private data. It pairs large language models such as OpenAI, Qwen, DeepSeek, Claude and Llama with vector databases such as Milvus and Zilliz Cloud, so that a question can be broken down, searched against your own documents, and answered with a synthesized result.

Is deep-searcher free to use? What license does it have?

deep-searcher is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is deep-searcher still maintained?

Actively maintained. The last commit to deep-searcher was on Sep 22, 2026.

What can I use deep-searcher for?

Building an internal Q&A assistant over company documents that must stay in-house. Generating research-style reports from a collection of PDFs and notes. Searching a knowledge base with a local model through Ollama for fully offline use. Prototyping retrieval-augmented generation pipelines with swappable LLMs and embedding models.

How do I install deep-searcher?

Create and activate a virtual environment: python -m venv .venv && source .venv/bin/activate Install DeepSearcher with pip: pip install deepsearcher Set your OPENAI_API_KEY environment variable, then run the quick start Python demo from the README

Where can I get deep-searcher?

The source code and installation instructions are on GitHub: https://github.com/zilliztech/deep-searcher