Local Deep Research

Free

A deep research tool for searching academic sources, the web, and private documents with local or cloud LLMs. [#opensource](https://github.com/LearningCircuit/local-deep-research)

FreeFree tier

Inputs: textOutputs: text

Type

Open Source

Company

LearningCircuit

Visit Website GitHub (40 stars)

About Local Deep Research

Local Deep Research is an open-source AI-powered research assistant that performs deep, agentic research using multiple LLMs and search engines with proper citations. It supports local LLMs (llama.cpp, Ollama) and cloud LLMs (Google, etc.), and integrates with over 10 search engines including arXiv, PubMed, and private documents. The tool achieves state-of-the-art results on local hardware: ~95% on SimpleQA (n=500) and 77% on xbench-DeepSearch (n=100) using a single RTX 3090. It is designed for privacy, allowing users to run everything locally and encrypt their data. Deployment is simplified via Docker, and the codebase is fully transparent.

Key Features

Performs deep, agentic research using multiple LLMs and search engines with proper citations

Supports local LLMs (llama.cpp, Ollama) and cloud LLMs (Google, etc.)

Integrates with 10+ search engines including arXiv, PubMed, and private documents

Fully local on a single RTX 3090 with ~95% SimpleQA and 77% xbench-DeepSearch benchmarks

Privacy-focused: everything runs locally and data can be encrypted

Open source with Docker support for easy deployment

Allows building a searchable knowledge base from private documents

Pros & Cons

Pros

Achieves high accuracy on benchmarks (95% SimpleQA, 77% xbench-DeepSearch)
Fully local with privacy (encrypted data and no cloud dependency)
Supports a wide range of LLMs and search engines
Open source and transparent codebase
Easy deployment via Docker Compose

Cons

Requires significant local hardware (e.g., RTX 3090) for optimal local LLM performance
Setup involves multiple Docker containers and network configuration
Docker network limitations on Mac/Windows (--network host not supported)
May not achieve same accuracy with smaller or cloud-based models

Best For

Academic research with arXiv and PubMed integrationWeb research with multiple search enginesPrivate document analysis and knowledge base creationPrivacy-sensitive research requiring local processing

FAQ

What hardware is needed to run Local Deep Research?

It can run fully locally on a single RTX 3090 (tested with Qwen3.6-27B). For lower-end hardware, cloud LLMs are supported as an alternative.

Does it support cloud-based LLMs?

Yes, it supports both local LLMs (llama.cpp, Ollama) and cloud LLMs (e.g., Google).

How do I deploy it?

The recommended method is via Docker. Two options are available: Docker run (Linux only for --network host) or Docker Compose (Linux, Mac, Windows/WSL2).

Is my data private?

Yes, the tool is designed to run locally, and data can be encrypted. You own your data and can see exactly how the tool works.