Dify

Dify is an open-source LLM application development platform that provides a visual workflow editor, built-in RAG pipeline tools, and an observability layer, enabling teams to build production chatbots, knowledge bases, and multi-step agent pipelines without the code complexity of building from LangChain or LlamaIndex primitives directly. The visual canvas wires together models, retrievers, tools, and logic through a drag-and-drop interface that makes the application architecture visible and modifiable by team members without deep LLM framework expertise. The RAG module handles document ingestion, chunking, embedding, and retrieval with configurable parameters accessible through the interface rather than requiring code changes for each tuning iteration. Agent workflows support multi-step execution with tool calling, conditional branching, loop handling, and human-in-the-loop checkpoint nodes.

The annotation system enables human review and correction of model outputs during development, feeding those corrections back into evaluation datasets. Self-hosted Dify runs on Docker and is free for standard single-tenant use under a modified Apache License 2.0; operating it as a multi-tenant service, or removing/modifying the frontend logo and copyright notice, requires a separate commercial license from LangGenius, the producer. Dify Cloud provides a managed version with a generous free tier. With over 90,000 GitHub stars, it is among a widely adopted open-source LLM platforms in production globally, particularly in teams that want the visual workflow development experience with the ability to self-host and customize.