AI4Sci Daily Paper Agent

An automated, AI-driven daily paper tracking and intelligent filtering system tailored for the AI for Science (Bio + Chem) domain.

This project utilizes a stateful workflow built with LangGraph to automatically fetch, score, and summarize the latest research from top sources, ensuring you never miss a critical breakthrough in geometric deep learning, AIDD, single-cell omics, and protein structure.

🌟 Key Features

• Multi-source Fetching: Automatically parses the latest daily papers from arXiv, bioRxiv, and select Nature journals (e.g., Nature Biotechnology, Nature Machine Intelligence).
• Intelligent LLM Filtering: Powered by DeepSeek-R1, the system evaluates and scores papers on a 1-10 scale based on relevance to predefined core subjects (e.g., GNNs, Flow Matching, Perturb-seq, Molecular Glue). Only high-quality papers (score >= 7) pass the filter.
• Automated Summarization: For papers that pass the threshold, the agent generates multidimensional summaries covering the core contributions, technical routes, and innovations.
• Markdown Reports: Automatically compiles an easy-to-read daily Markdown report into the reports/ directory.

⚙️ Architecture

The pipeline is implemented as a LangGraph StateGraph to guarantee structured, predictable, and fully automated processing:

1. Fetch Node: Retrieves feed data.
2. Filter Node: Runs batch inferences to score abstract relevance.
3. Summarize Node: Provides in-depth analysis on valid papers.
4. Report Node: Generates the markdown output and saves it locally.

🚀 Getting Started

Prerequisites

We recommend using pixi for dependency management:

pixi install

Alternatively, use pip:

pip install -r requirements.txt

Configuration

1. Copy the environment configuration template:

cp .env.example .env

2. Fill in the required API keys (e.g., your SiliconCloud/DeepSeek API key) and email configurations in your newly created .env file.

Running the Agent

Start the agent pipeline by simply running:

python main.py

(Ensure that your main.py entrypoint executes the LangGraph workflow.)

🛠️ Tech Stack

• Langchain & LangGraph: Workflow management and orchestration
• DeepSeek-R1: Core reasoning and selection engine
• Python-dotenv: Environment and secrets management
• Feedparser: RSS and feed processing

🤝 Contributing

Contributions, issues, and feature requests are welcome!

📄 License

This project is licensed under the MIT License.