Dify

Description

Dify is an open-source platform for building LLM applications and agentic workflows, deployable on your own infrastructure. It combines a visual workflow builder, a prompt IDE, retrieval-augmented generation over your own documents, agent tools, and observability, so teams move from prototype to production without stitching libraries together. It supports dozens of model providers, both local and hosted.

Features

• Visual workflows: Node-based canvas for chains, branching, loops, and multi-agent orchestration.
• RAG engine: Document ingestion, chunking, embeddings, reranking, and citation-backed answers.
• Prompt IDE: Side-by-side model comparison, versioning, and variable testing.
• Agents & tools: Function calling with built-in and custom tools, plus a plugin marketplace.
• Model-agnostic: OpenAI, Anthropic, Azure, and 100+ providers, including local models via Ollama.
• Observability: Full request logging, annotations, and analytics; publish apps as web apps or APIs.

Technology Stack

• Python / Flask backend with Celery workers
• Next.js / React frontend
• PostgreSQL, Redis, and a pluggable vector store (Weaviate, Qdrant, pgvector, ...)
• Docker Compose for self-hosting

Requirements

• Docker and Docker Compose
• Minimum 2 CPU cores and 4 GB RAM
• API keys for at least one model provider (or a local model runtime)