Dify
Dify is an open-source platform for building generative AI applications and agents, combining a visual workflow builder, prompt orchestration, retrieval-augmented generation pipelines, and model management behind one backend and console. It supports self-hosting the full application stack via Docker Compose.
What gets deployed with it
- PostgreSQL HA Cluster — required Dify requires PostgreSQL for application metadata, user data, workflow execution history, and knowledge base indexing records.
- Redis Cluster — required Dify uses Redis for Celery task brokering, session caching, and rate limiting. Without Redis the worker queue and real-time features are unavailable.
Baseline topologies
| Topology | Minimum nodes | Roles |
|---|---|---|
| single-node | 1 | dify |
| app-plus-workers | 2 | dify-api, dify-worker |
What drives the cost
- Peak active app users
- Peak users simultaneously interacting with Dify apps and chatbots. Drives API server thread pool and web worker sizing.
- Daily workflow executions
- Approximate agentic workflow or pipeline executions per day. Drives Celery worker capacity and sandbox concurrency planning.
Where it runs
bare_metal, dedicated, cloud_vm
Size Dify and get a cost estimate