LazyLLM

Why we included this project

LazyLLM is a low-code framework for building multi-agent LLM applications without doing the integration work yourself. Instead of juggling several tools, you assemble agents from built-in data flows and modules, then keep improving the same project as it grows: prototype an agent, look at which bad cases keep showing up, and fine-tune models at the stages that need it. It also covers the practical pieces, with support for vLLM and LMDeploy inference, RAG, MCP-based tool calling, and optional fine-tuning through LLaMA-Factory, all exposed through a Python API and CLI. For a small team, that makes it a reasonable home base when you want one framework to take a project from a quick demo to a deployed service that handles concurrent users and tolerates faults.