#107107th of 153 in Agent frameworks and SDKs

lazyllm

Python framework for building multi-agent and RAG apps with local or hosted models

Stars
3.9k
License
Apache-2.0
Last commit
Nov 2026
Last release
Nov 2026
Language
Python
Stars, last 29 weeks+127
25 Apr 2026: 3,763 stars2 May 2026: 3,772 stars9 May 2026: 3,783 stars16 May 2026: 3,787 stars23 May 2026: 3,791 stars30 May 2026: 3,800 stars6 Jun 2026: 3,804 stars13 Jun 2026: 3,810 stars20 Jun 2026: 3,811 stars27 Jun 2026: 3,817 stars4 Jul 2026: 3,827 stars11 Jul 2026: 3,831 stars18 Jul 2026: 3,833 stars25 Jul 2026: 3,834 stars1 Aug 2026: 3,839 stars8 Aug 2026: 3,846 stars15 Aug 2026: 3,850 stars22 Aug 2026: 3,854 stars29 Aug 2026: 3,863 stars5 Sept 2026: 3,866 stars12 Sept 2026: 3,869 stars19 Sept 2026: 3,871 stars26 Sept 2026: 3,873 stars3 Oct 2026: 3,876 stars10 Oct 2026: 3,881 stars17 Oct 2026: 3,885 stars24 Oct 2026: 3,885 stars31 Oct 2026: 3,889 stars7 Nov 2026: 3,890 stars14 Nov 2026: 3,890 stars

3,763 on 25 Apr 20263,890 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

LazyLLM is a Python library for assembling multi-agent LLM applications from modules (TrainableModule, OnlineChatModule, ServerModule, WebModule) and flows such as pipeline, parallel, switch and loop. It wraps local inference (lightllm, vllm), fine-tuning (collie, peft) and online providers behind one interface, and includes RAG components: Document, Retriever, Reranker. A built-in WebModule serves a chat UI, and launchers target bare metal, Slurm and SenseCore.

Who it is for: Python developers building multi-agent or RAG apps on local or hosted models

Strengths

  • Same code path for local models and online providers such as OpenAI, Kimi, Zhipu, Qwen
  • Flow primitives (pipeline, parallel, diverter, loop, switch) compose RAG and agent graphs briefly
  • Fine-tuning and inference frameworks share one API, with automatic framework selection
  • CLI shortcuts `lazyllm run chatbot` and `lazyllm run rag` start a working app

Weaknesses

  • Local inference limited to lightllm and vllm; fine-tuning limited to collie and peft
  • Online fine-tuning only for GPT, SenseNova and Tongyi Qianwen
  • No Dockerfile or compose file detected; image packaging is described but not shipped
  • Built-in components and examples lean toward Chinese models and providers

What it needs

  • GPU optional
  • Needs lightllm or vllm (local inference), collie or peft (fine-tuning), online provider API key (e.g. OpenAI)
  • Models: internlm2-chat-7b, OpenAI, SenseNova, Kimi, Zhipu GLM

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Agent frameworks and SDKs

See all 153
Also in Agent frameworks and SDKs
RankProjectScore
1ppt-masterAgent skill that turns documents into natively editable PowerPoint decks Live demo ↗ (opens in a new tab)78k starsMIT96 out of 100
2copilotkitReact and Angular SDK for adding agent chat and generative UI to apps38.8k starsMIT92 out of 100
3openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing30.7k starsMIT92 out of 100
#4mastraTypeScript framework for building AI agents, workflows and MCP servers29.6k starscustom license92 out of 100
#5browser-usePython library that lets an LLM agent control a web browser Live demo ↗ (opens in a new tab)121.4k starsMIT90 out of 100