// the find
1Panel-dev/MaxKB
🔥 MaxKB is an open-source platform for building enterprise-grade agents. 强大易用的开源企业级智能体平台。
MaxKB is a self-hosted RAG + agentic workflow platform built on Django/Python, pgvector, and LangChain. It gives teams a UI-driven way to build knowledge bases and wire up LLM-powered workflows without writing much code. Target audience is enterprises that want to run their own knowledge base chat system on-prem, especially those already using Chinese LLMs like Qwen or DeepSeek.
- Single Docker command deployment with pgvector bundled — genuinely easy to get running compared to assembling your own LangChain stack from scratch.
- Unusually wide set of workflow nodes out of the box: loops, condition branching, MCP tool calling, speech-to-text/text-to-speech, image generation, reranker — the node library is more complete than most open-source competitors.
- Model-agnostic by design with adapters for both local (Ollama, vLLM) and hosted providers, so you're not locked into one inference backend.
- GPL v3 license means the hosted/SaaS use case is covered and forks can't quietly go closed-source without consequences.
- GPLv3 is a real problem for commercial adoption — if you embed this in a product you need to open-source your entire stack or buy a commercial license. The README doesn't mention this risk at all.
- The codebase is heavily Chinese-first: default workflow JSON files exist in zh/zh_Hant/en variants, docs lean toward Chinese users, and community support outside Chinese forums is thin.
- No test suite visible in the directory tree — for a platform handling enterprise document ingestion and agent execution, the absence of automated tests is a liability when you need to upgrade or customize.
- Built on LangChain, which has a history of breaking API changes between minor versions; any significant LangChain upgrade could require substantial rework of the pipeline internals.