- Currently, mainly targets computer science research, especially machine learning research.
- More functionalities will be added soon.
- Example usage and API documentation are available.
Auto-Research is a framework designed to simplify and accelerate academic research tasks. It offers a modular and extensible architecture to help researchers, developers, and academics efficiently search, organize, summarize, and analyze academic papers.
See home page for API documentation and detailed examples
See this Google Colab notebook for an installation-free quick demo
While there are many existing tools for research automation. But here are the key distinctions and advantages of AutoResearch:
-
No additional API keys besides LLM API keys are required (No API keys, such as Semantic Scholar keys, are needed for literature search and downloading papers)
-
Support multiple search keywords for one inquiry.
-
Rank the papers based on their impacts, and consider the most important papers first.
-
Fast literature search process. It only takes about 3 seconds to automatically download a paper.
-
Python code-base, which enables convenient deployment, such as Google Colab notebook, as well as efficient integration with other ML-based tools, such as other LLM agents
-
API documentation is available
OS: Linux-based
Python: >= 3.10
git clone https://github.com/JLX0/auto_research
cd auto_research
pip install .
First, fill in your API keys for LLMs with your actual keys in key.json. See Setting up API keys for LLMs for more information. You only need to fill in at least one type of key.
Then, check examples.
For example, python examples/topic_to_survey converts your research topic or
question of interests to a survey over relevant papers. An explanation about the motivation behind
this functionality and an example run of the script is available here.
If you want to use DeepSeek models, change the argument target_key in get_api_key and the argument model in various instance initiations accordingly.