Main Features

A concise list of Dify capabilities. For concepts and deployment components, see Introduction.

LLM Application Development

Low-Code Builder: Build Assistant, Text Generator, Agent, and Workflow/Chatflow applications through the web UI
App Types: Assistant for chat, Text Generator for text generation, Agent for reasoning and tool use, Workflow for agentic automations
Model Support: Integration with multiple LLM providers (e.g. OpenAI, Azure, Anthropic); set up model provider in the console
Prompt Management: Visual prompt design, variables, and versioning

Knowledge Base: Ingest documents and URLs, chunk and embed for retrieval
Vector Store: The Helm chart currently supports pgvector only (or disable when not using RAG); configure pgvector connection in values
Retrieval Options: Hybrid search, re-ranking, and configurable retrieval strategies
Context Enrichment: Parent-child and extended context for accurate answers
Dataset Management: Versioning, update, and quality control of knowledge datasets

Visual Workflow: Node-based workflow editor with LLM, retrieval, code, and logic nodes; drag-and-drop agentic flows
Agent Capabilities: Tool use, multi-step reasoning, and conversation memory
Error Handling: Configurable error handling and retries for production reliability
Observability: Execution logs and debugging for workflows and agents

Backend Service API: REST endpoints for chat, completion, and workflow invocation; call your app from external systems (backend, scripts, other services).
API Keys: Create and manage API credentials per application in API Access (in the app sidebar); you can create multiple keys for different environments or users. See Developing with APIs.
Web App: Publish your app as a browser-based UI for end users: get a public URL, or embed the chat/completion widget in your site (iframe or script).
SDK & Docs: Client SDKs and API documentation for integration.