# llmware

A framework for building local, private, enterprise retrieval-augmented generation applications, pairing a catalog of 300+ small, optimized language models with document parsing and knowledge-base pipelines.

llmware is a framework for building knowledge-based large language model applications that can run locally, privately and on self-hosted infrastructure. It is optimized for AI PCs and laptops as well as edge and self-hosted deployment across Windows, Mac and Linux, with support for GGUF, OpenVINO, ONNX Runtime, ONNX Runtime for Qualcomm NPUs, WindowsLocalFoundry and PyTorch, choosing the inference technology suited to the target platform.

It has two main parts. A model catalog gives one consistent interface to more than 300 prepackaged, quantized models, including over 50 llmware fine-tuned SLIM, Bling, Dragon and Industry-BERT models built for specific enterprise tasks, alongside cloud models from OpenAI, Anthropic and Google. A retrieval-augmented generation pipeline handles the rest of the lifecycle: parsing and ingesting documents, chunking and embedding text, and building knowledge bases that scale from a single library to a cluster, which can then be queried with text, semantic, hybrid, metadata or custom filters and combined with an LLM prompt.

Website: https://llmware-ai.github.io/llmware/