llm-min 0.1.3

Generates LLM context by scraping and summarizing documentation for Python libraries listed in a requirements.txt file.

LLM-Min: Generate Compact Docs for LLMs

Problem: Large Language Models (LLMs) work best with focused, concise context. Feeding them entire documentation websites is inefficient and often counterproductive.

Solution: llm-min-generator automatically crawls Python library documentation and uses Google Gemini to generate compact, structured summaries (llm-min.txt) optimized for LLM consumption. It also saves the full crawled text (llm-full.txt) for reference.

Stop wasting tokens! Give your LLMs the focused context they need.

Key Features

• Automated Crawling: Finds and scrapes official Python package docs.
• LLM-Powered Summarization: Creates concise, structured summaries using the PCS (Progressive Compaction Strategy) via Google Gemini.
• Flexible Input: Process packages from requirements.txt, folders, or direct input.
• Easy Integration: Use via CLI or the Python LLMMinClient.
• Organized Output: Saves results neatly per package (output_dir/package_name/).

Quick Start

1. Installation:

Using pip (Recommended for users):

pip install llm-min

For Development/Contribution (Using uv):

# Clone (if you haven't already)
# git clone <repository_url>
# cd llm-min-generator

# Install dependencies (using uv)
python -m venv .venv
source .venv/bin/activate # or .venv\Scripts\activate on Windows
uv pip install -r requirements.txt
uv pip install -e .

# Install browser binaries for crawling
playwright install

# Optional: Install pre-commit hooks for development
# uv pip install pre-commit
# pre-commit install

2. Configure API Key:

• Recommended: Copy .env.example to .env and add your GEMINI_API_KEY. The application will automatically load it.
• Alternatively: You can provide the key directly using the --gemini-api-key CLI flag or pass it as the api_key parameter when initializing LLMMinClient in Python.

3. Generate Docs (CLI Example):

Process packages from a requirements file and save to my_llm_docs:

llm-min-generator -f path/to/your/requirements.txt -o my_llm_docs

• Use -pkg "requests\ntyper" for direct package input.
• Use -d /path/to/project to find requirements.txt in a folder.
• See llm-min-generator --help for more options (crawl depth, chunk size, etc.).

4. Generate Docs (Python Client Example):

from llm_min.client import LLMMinClient
import os

# Assumes GEMINI_API_KEY is in .env or environment
try:
    client = LLMMinClient()

    # Example: Compact existing text content
    long_text = "Your very long documentation text here..."
    subject = "My Custom Library"
    compacted_text = client.compact(content=long_text, subject=subject)

    print(f"--- Compacted {subject} ---")
    print(compacted_text)

    # You can also use client.process_package("package_name")
    # or client.process_requirements("path/to/requirements.txt")
    # See client documentation for details.

except (ValueError, FileNotFoundError) as e:
    print(f"Error initializing client (API Key or PCS Guide missing?): {e}")
except Exception as e:
    print(f"An error occurred: {e}")

Output

For each package, you'll get:

output_dir/
└── package_name/
    ├── llm-full.txt  # Raw crawled content
    └── llm-min.txt   # Compacted PCS content for LLMs

What is PCS (Packed Code Syntax)?

PCS is a highly condensed, machine-centric format designed for representing code structure and essential metadata with maximum information density. It uses single-character codes, minimal delimiters, and positional context to create a compact, single-string representation optimized for LLM context windows.

Think of it as a "minified" version of code documentation, focusing purely on the structural elements and relationships an LLM needs to understand an API or library, discarding natural language explanations. The full specification can be found in docs/pcs-guide.md.

Contributing

Contributions are welcome! See CONTRIBUTING.md (if available) or focus on improving discovery, compaction, LLM support, or tests.

License

MIT License.