GPT Researcher is an autonomous AI agent designed to conduct in-depth research on any topic, both online and from local documents, producing comprehensive, cited reports. It targets individuals and organizations needing efficient, factual, and unbiased information gathering, aiming to overcome limitations of traditional research methods and standard LLMs.

How It Works

The system employs a "planner" and "execution" agent architecture. The planner breaks down a research query into specific questions, while execution agents concurrently gather information from multiple web sources and local files for each question. Summaries are then aggregated, filtered, and sourced to create a final, detailed report, with a "Deep Research" feature enabling recursive exploration of subtopics.

Quick Start & Requirements

• Install: pip install gpt-researcher or via Docker (docker-compose up --build).
• Prerequisites: Python 3.11+, OpenAI API Key, Tavily API Key.
• Setup: Clone repo, set API keys (env vars or .env), install dependencies (pip install -r requirements.txt), run with uvicorn main:app --reload.
• Docs: https://docs.gptr.dev
• Colab: https://colab.research.google.com/github/assafelovic/gpt-researcher/blob/master/docs/docs/examples/pip-run.ipynb

Highlighted Details

• Generates reports exceeding 2,000 words by aggregating over 20 sources.
• Supports JavaScript-enabled web scraping and smart image scraping.
• Features a "Deep Research" mode for recursive, tree-like exploration of topics.
• Offers both lightweight (FastAPI) and feature-rich (NextJS) frontend options.

Maintenance & Community

• Active development with a public roadmap.
• Community support via Discord: https://discord.gg/QgZXvJAccX
• Author: Assaf Elovic.

Licensing & Compatibility

• Licensed under Apache 2.0.
• Permissive license suitable for commercial use and integration with closed-source projects.

Limitations & Caveats

The project is experimental and provided "as-is." While aiming to reduce bias and misinformation through broad source aggregation, it does not claim to eliminate bias entirely. Costs for research are estimated at ~$0.4 per run using specific models and settings.