AutoLLM is a Python library designed to rapidly deploy Retrieval Augmented Generation (RAG) based LLM web applications and APIs. It targets developers seeking to simplify the creation of LLM-powered applications by unifying access to over 100 LLMs and 20+ vector databases, offering a streamlined RAG engine and FastAPI integration.

How It Works

AutoLLM abstracts the complexities of LLM orchestration by providing a unified API for various LLM providers (Hugging Face, Ollama, Azure, VertexAI, Bedrock) and vector stores (LanceDB, Qdrant). Its core AutoQueryEngine class allows users to configure RAG pipelines with a single line of code, abstracting away data loading, chunking, embedding, vector storage, and LLM interaction. The library also offers a one-line conversion to a FastAPI application for easy deployment.

Quick Start & Requirements

• Install with: pip install autollm or pip install autollm[readers] for data readers.
• Requires Python >= 3.8.
• Official quickstart notebook: https://colab.research.google.com/github/safevideo/autollm/blob/main/examples/quickstart.ipynb

Highlighted Details

• Supports 100+ LLMs and 20+ vector databases with a unified API.
• Features automated cost calculation for LLM usage.
• Enables 1-line RAG LLM engine and FastAPI app creation.
• Offers easy migration from Llama-Index.

Maintenance & Community

• Active development with a public roadmap.
• Community support via GitHub Discussions (Q&A, Feature Requests).
• Contact: support@safevideo.ai

Licensing & Compatibility

• Licensed under GNU Affero General Public License (AGPL 3.0).
• AGPL 3.0 is a strong copyleft license; commercial use is permitted but requires making source code available under the same license if the software is distributed or accessed over a network. Contact for specific commercial licensing.

Limitations & Caveats

The AGPL 3.0 license may impose significant obligations for commercial use, particularly for network-accessed services, requiring source code disclosure.