Paper Agent - Zotero Arxiv Daily

English | 中文

Your personal AI research assistant that serves you a daily dose of arXiv papers, tailored to your interests from Zotero or the pdfs from your local folder.

---

✨ Features

• Dual Preference Sources: Build your interest profile based on your Zotero library OR a local directory of PDF files.
• Personalized Paper Recommendations: Leverages your chosen preference source to understand your research interests.
• Daily arXiv Digest: Fetches the latest papers from your favorite arXiv categories.
• AI-Powered Summaries: Uses a Large Language Model (LLM) to generate concise summaries of each paper.
• Multi-language Support: Get summaries in your preferred language (English or Chinese supported out-of-the-box).
• Highly Configurable: Easily tune the agent's behavior through a simple config.yaml file.

🚀 Getting Started

📋 Prerequisites

• Python 3.11+
• A Zotero account (if using the Zotero preference source) or from your local directory of PDF files.
• An OpenAI-compatible API key for paper summarization

⚙️ Installation

1. Clone the repository:

   git clone https://github.com/Marverlises/Paper-Agent-Zotero.git
   cd Paper-Agent-Zotero
   

2. Play with:

   uv run main.py
   

🛠️ Configuration

1. If it doesn't exist, create a src/config.yaml file. You can use the following as a template:

2. Open src/config.yaml and configure the following:

   • preference_source: Set this to "zotero" or "local" to define your source of interest papers.

   • zotero (if preference_source: "zotero"):

     • id: Your Zotero user ID.
     • key: Your Zotero API key.
     • recency_months: How recent the papers from your Zotero library should be to be considered for interest calculation.

   • local (if preference_source: "local"):

     • path: The file path to your local directory of preferred papers in this folder.

   • arxiv:

     • query: The arXiv categories you want to follow (e.g., cs.AI+cs.CV).

   • llm:

     • openai_api_key: Your OpenAI-compatible API key.
     • openai_api_base: The base URL for your LLM API.
     • model_name: The model you want to use for summarization (e.g., gpt-4o).

   For more details, see the comments in the config.yaml file.

🏃‍♀️ Usage

To run the agent and generate your daily paper digest, simply run:

uv run main.py

The daily digest will be saved as a Markdown file in the year-month-day-hour directory.

👓Example Output

• Example output folder tree

[YYYY-MM-DD-HH]/ 
│   └── (Example output)
├── Daily_Report.md
│   └── (A quick look of the Top-N most relevant papers)
├── assets_dir/
│   └── (Each paper's detailed report with images)
├── [arXiv_ID]v[version].[Paper_Title].pdf
├── [arXiv_ID]v[version].[Paper_Title].pdf
└── ...

• The detailed info in assets_dir

In each folder, the result is like this:

In each paper's assets_dir:

In each paper's detailed info markdown file (Part of the result):

🤝 Contributing

Contributions are welcome! Please feel free to submit a pull request or open an issue.

📄 License

Please see the LICENSE file for details.