LLM-RSS

Introduction

LLM-RSS is a tool that reads title and abstract from RSS sources like Nature, Arxiv, etc., filters them based on specific research areas or keywords provided by users, and then generates a new RSS feed in xml. Modify the prompt as needed. The generated xml file can then be hosted via nginx and accessed via Zotero, for example. Unlike keyword matching, LLM-RSS uses large language models (LLMs) to understand the content, providing a more contextually relevant feed.

Installation

To install and run LLM-RSS, follow these steps:

1. Clone the repository:

   git clone https://github.com/xsgeng/llm-rss.git
   cd llm-rss

2. Install the required dependencies:

   pip install -r requirements.txt

3. Configure the project:

   • Copy the config.d/comfig.toml.example to config.d/config.toml.
   • Modify the config.toml file to include your RSS feed URLs, research areas, and other settings.
   • The program sequentially processes each toml file.

Supported RSS Providers

LLM-RSS currently supports the following RSS providers:

• Nature
• Arxiv
• APS (American Physical Society)
• BioRxiv
• Cell
• AIP (American Institute of Physics)
• IOP (Institute of Physics)

Use with Cron

LLM-RSS is intended to be used along with a cron job to periodically update the filtered RSS feed. Here is an example of how to set up a cron job:

1. Open the crontab editor:

   crontab -e

2. Add a new cron job:

   0 0 * * * /usr/bin/python3 /path/to/llm-rss/main.py

   This cron job will run the script every day at 0:00.

Hosting via Nginx

A docker-compose.yaml is provided to host the generated RSS feed via Nginx:

docker-compose up -d

The RSS feed will be available at http://localhost:8080/rss/rss.xml depending on your setup.

Feed to Zotero

Add http://localhost:8080/rss/rss.xml to your Zotero RSS feed. Then you can read and access your papers in Zotero.