This repository is a curated list of open-source libraries and tools for building and deploying Large Language Models (LLMs) in production environments. It targets engineers, researchers, and developers working with LLMs, providing a structured overview of resources for data preprocessing, training, serving, evaluation, and application development.

How It Works

The list is organized into distinct categories, each covering a critical aspect of the LLM lifecycle. It highlights key projects within these categories, offering brief descriptions and linking to their respective repositories. This approach allows users to quickly identify relevant tools for specific production LLM needs, from data curation to agent benchmarking.

Quick Start & Requirements

This is a curated list, not a runnable project. Users will need to visit individual project links for installation and requirements.

Highlighted Details

• Comprehensive coverage across the LLM production lifecycle: Data Preprocessing, Training/Finetuning, Evaluation, Serving/Inference, Application/RAG, Extraction/Parsing, Testing/Monitoring, Guardrails/Security, Cookbooks, Courses, and Agent Benchmarks.
• Regularly updated with new categories and projects, marked with 📌.
• Includes links to official documentation, examples, and related resources for each listed tool.
• Features prominent projects from leading organizations like Hugging Face, NVIDIA, Microsoft, OpenAI, and Meta.

Maintenance & Community

The list is maintained by jihoo-kim and is inspired by "Awesome Production Machine Learning." Updates are frequent, with new categories added regularly. Links to community resources are not provided for the list itself, but are available via the individual project links.

Licensing & Compatibility

The licensing varies by individual project. Users must consult the license of each tool they choose to adopt. Compatibility for commercial use or closed-source linking depends on the specific licenses of the included projects.

Limitations & Caveats

This is a curated list and does not provide direct functionality. Users must evaluate and integrate individual tools themselves. The rapid evolution of the LLM space means the list may not always reflect the absolute latest or most niche tools.