AI Research Agent
Automated AI Repository Discovery & Analysis
Searches GitHub daily for new AI/ML repositories, analyzes them with local LLMs via Ollama, and commits structured markdown reports — fully autonomous via GitHub Actions.
How It Works
Search GitHub API→
Fetch README & Tech Stack→
Analyze with Ollama LLM→
Clone & Run Tests→
Compare & Rank→
Generate Report→
Git Commit
Runs daily at 08:00 UTC via GitHub Actions
Features
- Core Multi-query GitHub search with deduplication and date filtering
- Core README analysis, language breakdown, and dependency detection
- LLM Ollama-powered summaries (what it does, innovations, impact)
- Advanced Clone repos, detect test frameworks, run tests with timeout
- Advanced Comparative analysis — scores and ranks repos by composite metrics
- v2.0 Matplotlib dashboard charts in every report
- v2.0 Day-over-day delta comparison (self-correcting analysis)
Tech Stack
Python
Ollama (llama3.2)
GitHub REST API
GitPython
Matplotlib
GitHub Actions
PyYAML
Reports
Every day, a new markdown report is generated in the reports/ directory. Each report includes:
- Metadata table (date, repos analyzed, model used)
- Per-repo sections with LLM summary, tech stack, and links
- Comparative ranking with composite scores
- Dashboard charts (PNG)