An autonomous research agent that monitors arXiv daily, summarizes papers using LLMs, and delivers personalized academic digests by email. Built on top of the Zotero-arXiv-Daily framework with custom extensions.
- Fetches new arXiv papers daily based on your Zotero library and research interests
- Generates AI-powered TL;DR summaries for each paper using LLM APIs (OpenAI, Anthropic)
- Ranks papers by relevance to your recent reading history
- Constructs and sends a formatted email digest with paper links, affiliations, and code repos
- Runs fully automated via GitHub Actions at zero cost
| Component | Technology |
|---|---|
| Language | Python 3.11 |
| LLM APIs | OpenAI / Anthropic Claude |
| Deployment | GitHub Actions + Docker |
| SMTP / Gmail | |
| Reference Manager | Zotero API |
main.py # Orchestrator: fetch -> rank -> summarize -> email
paper.py # arXiv paper fetching and filtering logic
llm.py # LLM API wrapper with retry logic
construct_email.py # Email template builder
docker-compose.yml # Containerized deployment
- Fork this repo
- Set the following GitHub Actions secrets:
ZOTERO_API_KEY=your_zotero_api_key
ZOTERO_USER_ID=your_zotero_user_id
OPENAI_API_KEY=your_openai_api_key # or ANTHROPIC_API_KEY
EMAIL_ADDRESS=your_gmail
EMAIL_PASSWORD=your_app_password
TO_EMAIL=recipient@email.com
- The workflow runs daily at 8am UTC automatically.
git clone https://github.com/GuillaumeBld/Research-agent.git
cd Research-agent
cp .env.example .env # fill in your API keys
docker-compose uppip install -r requirements.txt
python main.py- Deployed and running for personal research tracking in AI/Finance
- Processes 20-50 new papers per day across ML, finance, and econometrics
- Average summarization latency: < 2 seconds per paper with Claude Haiku
- Email digest delivered daily with ranked relevance scores
AGPL-3.0 — see LICENSE
