AI Safety Events and Training

This project helps find, process, and evaluate information about AI safety events and training programs using the Exa API and OpenRouter API.

Features

• Search for AI safety events (conferences, workshops, seminars)
• Process and export results to JSON and Markdown
• Customizable queries for AI safety, governance, ethics, etc.
• Automatic date range for up-to-date results
• AI-powered scoring of results for relevance to AI safety
• Error handling and retry mechanism

Requirements

• Rye
• Exa API Key (set EXA_API_KEY in .env file)
• OpenRouter API Key (set OPENROUTER_API_KEY in .env file)

Setup and Usage

1. Clone the repository
2. Create a .env file in the project root and add your API keys:

   EXA_API_KEY=your_exa_api_key_here
   OPENROUTER_API_KEY=your_openrouter_api_key_here
   

3. Run rye sync to set up the project
4. Run rye run python -m aisafetyeventsandtraining.main
5. Check the ./results folder for output files (JSON and Markdown)

Output

The script generates two types of output files in the results folder:

1. JSON file: Contains raw data of search results and AI safety scores
2. Markdown file: Presents a formatted list of results, sorted by AI safety relevance score

Customization

You can modify the following variables in main.py to customize the script's behavior:

• NUM_RESULTS: Number of results to fetch per query
• DAYS: Number of days in the past to search for events
• OPENAI_MODEL: The OpenAI model to use for scoring (via OpenRouter)
• MAX_SUMMARY_LENGTH: Maximum length of result summaries

License

MIT License

License

MIT License