What it does
Dify is a platform for composing LLM applications, retrieval pipelines, and agentic workflows. It combines a visual interface with backend services and exposed APIs.
Why it’s interesting
A full-stack reference shows how model providers, background jobs, document ingestion, and workflow execution fit together. Review its license conditions before redistributing a hosted version.
Real-world use case
Prototype an internal docs assistant and connect its API to a small frontend.
What I can learn
- 01
Graph-based workflow execution
Follow how connected workflow nodes are validated and executed as a graph, including branching and asynchronous steps.
- 02
Document ingestion and retrieval
Study how documents are parsed, divided into chunks, embedded, and retrieved as context for a generated answer.
- 03
Provider and tool adapters
See how provider adapters give model APIs and external tools a common interface within a workflow.
One thing to explore
Workflow execution engine
Follow the api/core/workflow implementation and connect it to the concepts above.
Explore the source (opens in a new tab)10-Minute Challenge
Inspect an example workflow
Before you start: Docker and several containers; full local setup takes longer than ten minutes. Use current docs and avoid pasting sensitive documents into a demo.
- Read the current Docker Compose installation guide.
- Open a sample workflow and identify its nodes, inputs, and outputs.
- Trace one node to its backend handler in the source.