Dify
A visual platform for building and operating AI workflows, agents, and RAG applications.
By LangGenius · reviewed 2026-08-07
What Dify does
Dify combines model-provider management, workflow composition, prompt and agent tools, knowledge bases, APIs, and operational features. It can be self-hosted or used as a managed cloud service, making it an application platform rather than a model runtime.
Pricing model
Source-available self-hosting; paid cloud plans
Position in the stack
Where Dify fits
Dify is an AI application platform above model providers. It combines workflow and agent composition, prompt management, knowledge bases, tools, APIs, and operational controls so teams can publish applications without building every integration from scratch.
Typical deployment flow
A practical sequence for evaluating Dify before making it part of a permanent stack.
- STEP 1
Choose cloud or self-hosting
Decide whether managed convenience or infrastructure control best fits data, compliance, cost, and operational requirements.
- STEP 2
Configure providers and data
Connect language, embedding, reranking, and media providers, then define knowledge stores and secrets.
- STEP 3
Build and publish
Compose workflows or agents, test failure paths and costs, then expose the application through UI or APIs.
Best for
- Building AI workflows without coding every integration
- Publishing agent and RAG applications
- Using local and cloud models behind one application layer
Not the right layer for
- Direct model inference optimization
- GPU inventory and cluster scheduling
Capabilities
Capabilities refer to the tool's application layer. Hardware and model support can still depend on a connected inference engine.
Deployment and compatibility
- Deployment
- Self-hostedManaged cloud
- Platforms
- Linux · Docker · Kubernetes · Managed cloud
- Hardware backends
- Uses connected model runtime
- Install methods
- Docker Compose · Kubernetes · Managed cloud
Models and integrations
- Model formats
- Uses connected model runtime
- Common integrations
- OllamaOpenAI-compatible APIsCloud model providersVector databases
Strengths
- Visual workflow builder
- Broad provider and knowledge integrations
- Application APIs and operational controls
Limitations
- Self-hosting introduces several dependent services
- License conditions differ from a standard OSI license
Planning checklist
Before you choose Dify
Answer these questions with the exact models, hardware, users, and data you expect to operate.
- Which data and model calls are allowed to leave the private environment?
- Does the workflow require deterministic branching or autonomous agent behavior?
- How will secrets, users, quotas, and application versions be managed?
- Can the team operate the dependent services required by self-hosting?
Dify FAQ
Does Dify replace Ollama or vLLM?
No. Dify builds applications and workflows above model providers. Ollama, vLLM, cloud APIs, and other services can supply the underlying inference.
Can Dify use local models?
Yes, through supported local or OpenAI-compatible providers. Network routing and authentication must allow the Dify deployment to reach those endpoints.
Is self-hosted Dify operationally lightweight?
It is more involved than a desktop application because the platform depends on multiple services. Backups, upgrades, secrets, monitoring, and access control need an operating plan.
Official sources
Use these links to confirm current compatibility and installation requirements.
Related tools
A desktop and self-hosted workspace for document chat, RAG, and AI agents.
A visual low-code platform for composing, evaluating, and deploying AI agents and LLM workflows.
A self-hostable personal AI and knowledge assistant with document search, agents, automations, and research workflows.
Execution evidence
Known working recipes using Dify
Recipes connect hardware, a model artifact, tools, settings, verification, and a reportable result.
No verified recipe is linked to this record yet.
Compatibility estimates remain available in the planner. A recipe appears here only after its exact stack and verification protocol are documented.