Overview: Dify is an LLM application platform that combines visual workflows, prompt development, RAG pipelines, agents, model management, and observability in one workspace.
What beginners can do: Beginners can build a simple input-to-model-to-output flow, publish it as an application or API, and then add document retrieval, tools, branches, and evaluation.
Recommended start: For self-hosting, verify the current minimum system requirements, install Docker and Docker Compose 2.24 or later, use the repository's docker directory, and complete initialization at the local install page.
Requirements and cost: A model provider or supported local endpoint is required. Self-hosting brings multiple containers, persistent data, backups, upgrades, and production networking responsibilities.
Who it suits: Useful for product builders and small teams that want to prototype chat, RAG, or agent workflows visually and later expose an API.
Limits and safety: Complex visual flows can become hard to review and version. The repository uses the Dify Open Source License, based on Apache 2.0 with additional conditions, so review it before hosted redistribution or commercial packaging. Keep provider keys server-side, enable authentication, restrict tools and code execution, back up persistent volumes, and never expose a default installation directly to the internet.
Project and license: Repository content is provided under Dify Open Source License based on Apache-2.0 with additional conditions; check the current LICENSE before reuse. Repository: https://github.com/langgenius/dify