Skip to content
← Tools / Skills
ToolLangGenius Inc./10 trustVerified

Dify

Open-source LLM app development platform — visual AI workflows, RAG pipelines, agents, and model management in one workspace.

Visit official site →

About

Dify combines a visual workflow canvas, enterprise RAG (hybrid search, reranking), a multi-model gateway (100+ providers), a prompt IDE, agent capabilities, LLMOps observability, and a Backend-as-a-Service API layer.

It ships under the Dify Open Source License — Apache 2.0-based with added conditions (no multi-tenant SaaS without a commercial license; no logo/copyright removal) — so it is source-available rather than plain OSI open source.

What it can do

  • Visual workflow studio for AI workflows, agents, and conditional/loop/parallel execution
  • Full RAG pipeline: document ingestion (PDF/PPT/Word), chunking, hybrid search, reranking
  • Unified model gateway for hundreds of LLMs (OpenAI, Anthropic, Gemini, DeepSeek, Ollama…)
  • One-click publishing as web app, embeddable widget, or REST API
  • Plugin marketplace, MCP servers, and custom tool support
  • LLMOps dashboards with Opik/Langfuse/Arize Phoenix observability integrations

Setup

  1. Check minimum requirements: CPU ≥ 2 cores, RAM ≥ 4 GiB; install Docker + Docker Compose v2.24+.
  2. git clone https://github.com/langgenius/dify.git --depth 1
  3. cd dify/docker && cp .env.example .env
  4. docker compose up -d
  5. Open http://localhost/install and complete the admin setup wizard.

Prerequisites

Docker/Docker Compose; ≥4 GiB RAM (8 GiB realistic for production); LLM provider API keys.

Use cases

chatautomationdatasearchtts

Works with

Generic API

Trust score — 9/10

/10 trustVerified

Second-most-starred entry (158k), extremely active development, strong docs, and a full product surface (cloud/VPC/self-host). Docked for: (1) the license is Apache-2.0-derivative with extra restrictions — not plain OSI open source; (2) the vendor is China-based, which matters for data-residency-sensitive teams; (3) heavier infra footprint than peers. Security certifications are vendor-stated — provider-claimed.

★ 158,265 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
9.5/10
Maintenance · recency, releases, issue load
9.7/10
Openness · license + docs
7.2/10
Manual review · sanity read + red-flag check
8.0/10
  • · 158,265 GitHub stars (log-scaled vs catalog peers)
  • · last push 0d ago; 26 releases/12mo; 1111 open issues
  • · manual review: China vendor + heavy footprint noted

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.