Auto-Deep-Research - A fully automated personal AI assistant open-sourced by the University of Hong Kong.
Auto-Deep-Research is an open-source, fully automated personal AI assistant developed by Professor Chao Huang's lab at the University of Hong Kong, serving as an open-source alternative to OpenAI Deep Research. Developed based on the AutoAgent framework, it focuses on deep learning...
What is Auto-Deep-Research?
Auto-Deep-Research is an open-source, fully automated personal AI assistant developed by Professor Chao Huang's lab at the University of Hong Kong, serving as an open-source alternative to OpenAI Deep Research. Developed based on the AutoAgent framework, it focuses on deep research functionalities and employs a modular multi-agent architecture, including a Web Agent, a Coding Agent, and a Local File Agent. These agents are responsible for internet information retrieval, programming implementation and debugging, and parsing of various file formats, respectively. It supports multiple large language models (LLMs), such as Anthropic, OpenAI, Mistral, and Hugging Face, and is built solely on Claude-3.5-Sonnet. Auto-Deep-Research supports importing browser cookies for better access to specific websites.
Main functions of Auto-Deep-Research
- In-depth research functionIt focuses on automating complex tasks such as file parsing, web searching, data analysis and visualization, and can generate detailed reports.
- Multilingual model supportIt is compatible with various large language models (LLMs), such as Anthropic, OpenAI, Mistral, Hugging Face, etc.
- High cost performanceBuilt on Claude-3.5-Sonnet, it is cost-effective and is the optimal solution among open-source solutions.
- Community-driven improvementBased on community feedback, features such as one-click start and enhanced LLM compatibility have been added.
- Easy to deployIt supports installation via Conda environment or Docker, and provides detailed startup configuration options.
The technical principles of Auto-Deep-Research
- Multi-Agent ArchitectureIt includes a Web Agent (Internet information search), a Coding Agent (programming implementation and debugging), and a Local File Agent (file parsing and understanding), which work together through the core scheduler (Orchestrator Agent).
- Web Agent: Focusing on accessible and in-depth search of internet information.
- Coding AgentResponsible for programming implementation and debugging, possessing strong logical analysis skills.
- Local File AgentDedicated to parsing and understanding the content of multi-format files.
Auto-Deep-Research project address
- Github repository:https://github.com/HKUDS/Auto-Deep-Research
Application scenarios of Auto-Deep-Research
- Scientific Research and Data AnalysisResearchers can use Auto-Deep-Research to quickly process and analyze data, automatically generating high-quality analysis reports.
- Financial and Market AnalysisFinancial analysts can use tools to track industry dynamics, assess market trends, generate investment research reports, and support data-driven decision-making.
- Education and LearningStudents and educators can use Auto-Deep-Research to conduct literature reviews, organize learning materials, and generate learning reports.
- Corporate Strategy and Business DecisionsBusinesses can use tools to conduct industry analysis, competitor research, and business strategy evaluation, and optimize product planning and market expansion strategies.