Predictions

A Python tool that analyzes prediction questions using AI models and web scraping to generate detailed probability estimates.

Features

• Uses Perplexity API for initial research
• Scrapes citation sources using Firecrawl
• Generates detailed analysis using OpenAI's o3-mini model
• Saves research prompts and predictions to files

Setup

1. Clone the repository:

git clone https://github.com/yourusername/predictions.git
cd predictions

2. Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows, use: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

4. Create a .env file in the root directory with your API keys:

PERPLEXITY_API_KEY=your_perplexity_api_key
FIRECRAWL_API_KEY=your_firecrawl_api_key
OPENAI_API_KEY=your_openai_api_key

Usage

Run the script with your prediction questions:

python index.py

The script will:

1. Generate research reports for each question
2. Scrape relevant citations
3. Create detailed analysis prompts
4. Generate probability estimates using AI
5. Save all outputs to files with timestamps

Output Files

For each prediction question, two files are generated:

• {timestamp}_{question}_prompt.md: Contains the research and analysis prompt
• {timestamp}_{question}_o3_prediction.txt: Contains the AI-generated probability estimate

License

MIT License