Show HN: LLMWare – Small Specialized Function Calling 1B LLMs for Multi-Step RAG

LLMWare is an open-source RAG framework designed to automate repetitive information processing tasks using small specialized 1B function-calling LLMs (SLIMs) optimized for multi-step workflows. The project, created by a former corporate lawyer and software engineer, provides ~50 models on Hugging Face and aims to make AI-powered document and data analysis practical and accessible.

Key Points

• •LLMWare is an open-source RAG framework designed to automate repetitive tasks in document-heavy industries like finance and insurance
• •The project features ~50 specialized 1B parameter models (SLIMs) optimized for function calling in multi-step RAG workflows
• •Small specialized models reduce computational overhead compared to large LLMs while maintaining accuracy for specific tasks
• •Founded by a former corporate lawyer and software engineer to solve practical AI automation challenges in legal and financial document processing
• •Models are available on Hugging Face and designed for production-ready RAG implementations
• •Multi-step RAG workflow approach enables complex document analysis through sequential specialized model calls
• •Focus on practical automation of high-value repetitive tasks that were previously handled manually by professionals

Workflow Diagram

Concepts