A Python script that generates llms.txt and llms-full.txt files for any website using Firecrawl and OpenAI APIs.

llms.txt is a standardized format for making website content more accessible to Large Language Models (LLMs). It provides:

• llms.txt: A concise index of all pages with titles and descriptions
• llms-full.txt: Complete content of all pages for comprehensive access

• 🗺️ Website Mapping: Automatically discovers all URLs on a website using Firecrawl's map endpoint
• 📄 Content Scraping: Extracts markdown content from each page
• 🤖 AI Summaries: Uses OpenAI's GPT-4o-mini to generate concise titles and descriptions
• ⚡ Parallel Processing: Processes multiple URLs concurrently for faster generation
• 🎯 Configurable Limits: Set maximum number of URLs to process
• 📁 Flexible Output: Choose to generate both files or just llms.txt

• Python 3.7+
• Firecrawl API key (Get one here)
• OpenAI API key (Get one here)

1. Clone the repository:

git clone <repository-url>
cd <repository-directory>

2. Install dependencies:

pip install -r requirements.txt

3. Set up API keys (choose one method):

   Option A: Using .env file (recommended)

   cp env.example .env
   # Edit .env and add your API keys

   Option B: Using environment variables

   export FIRECRAWL_API_KEY="your-firecrawl-api-key"
   export OPENAI_API_KEY="your-openai-api-key"

   Option C: Using command line arguments (See usage examples below)

Generate llms.txt and llms-full.txt for a website:

python generate-llmstxt.py https://example.com

# Limit to 50 URLs
python generate-llmstxt.py https://example.com --max-urls 50

# Save to specific directory
python generate-llmstxt.py https://example.com --output-dir ./output

# Only generate llms.txt (skip full text)
python generate-llmstxt.py https://example.com --no-full-text

# Enable verbose logging
python generate-llmstxt.py https://example.com --verbose

# Specify API keys via command line
python generate-llmstxt.py https://example.com \
  --firecrawl-api-key "fc-..." \
  --openai-api-key "sk-..."

• url (required): The website URL to process
• --max-urls: Maximum number of URLs to process (default: 20)
• --output-dir: Directory to save output files (default: current directory)
• --firecrawl-api-key: Firecrawl API key (defaults to .env file or FIRECRAWL_API_KEY env var)
• --openai-api-key: OpenAI API key (defaults to .env file or OPENAI_API_KEY env var)
• --no-full-text: Only generate llms.txt, skip llms-full.txt
• --verbose: Enable verbose logging for debugging

# https://example.com llms.txt

- [Page Title](https://example.com/page1): Brief description of the page content here
- [Another Page](https://example.com/page2): Another concise description of page content

# https://example.com llms-full.txt

<|firecrawl-page-1-lllmstxt|>
## Page Title
Full markdown content of the page...

<|firecrawl-page-2-lllmstxt|>
## Another Page
Full markdown content of another page...

1. Website Mapping: Uses Firecrawl's /map endpoint to discover all URLs on the website
2. Batch Processing: Processes URLs in batches of 10 for efficiency
3. Content Extraction: Scrapes each URL to extract markdown content
4. AI Summarization: For each page, GPT-4o-mini generates:
   • A 3-4 word title
   • A 9-10 word description
5. File Generation: Creates formatted llms.txt and llms-full.txt files

• Failed URL scrapes are logged and skipped
• If no URLs are found, the script exits with an error
• API errors are logged with details for debugging
• Rate limiting is handled with delays between batches

• Processing time depends on the number of URLs and response times
• Default batch size is 10 URLs processed concurrently
• Small delays between batches prevent rate limiting
• For large websites, consider using --max-urls to limit processing

python generate-llmstxt.py https://small-blog.com --max-urls 20

python generate-llmstxt.py https://docs.example.com --max-urls 100 --verbose

python generate-llmstxt.py https://example.com --no-full-text --max-urls 50

The script checks for API keys in this order:

1. Command line arguments (--firecrawl-api-key, --openai-api-key)
2. .env file in the current directory
3. Environment variables (FIRECRAWL_API_KEY, OPENAI_API_KEY)

Ensure you've either:

• Created a .env file with your API keys (copy from env.example)
• Set environment variables: FIRECRAWL_API_KEY and OPENAI_API_KEY
• Or pass them via command line arguments

If you encounter rate limits:

• Reduce concurrent workers in the code
• Add longer delays between batches
• Process fewer URLs at once

For very large websites:

• Use --max-urls to limit the number of pages
• Process in smaller batches
• Use --no-full-text to skip full content generation

MIT License - see LICENSE file for details