What should I know about Dify's license?

Apache-2.0 plus conditions: operating a multi-tenant service (one workspace = one tenant) needs a commercial license; don't strip logo/copyright from the official frontend. Single-tenant internal use is typically fine. The repo LICENSE wins.

How do I self-host Dify?

Official docs want Docker Compose 2.24+ and at least 2 CPU / 4 GiB RAM. Clone, `cd dify/docker`, copy `.env.example` to `.env`, `docker compose up -d`. First visit the install page on local port 80 to create the admin.

Knowledge Q&A, chat apps, models as product features → Dify. Cron, webhooks, wiring existing SaaS/internal systems → n8n. Many stacks run both.

Has the knowledge-Q&A exercise actually run in Dify?

No. Official docs, six bilingual synthetic FAQs, 26 questions and 22 native offline tests were checked. Dify, embeddings, models, native citations and actual human review were not run; later execution needs separate authorization. Expected results are not observations.

DifySelf-hostable · Validate your use caseApache-2.0 + conditions

A workspace to build RAG and agent apps visually

Dify is LangGenius' visual AI app platform: RAG, agent workflows and many model providers in one workspace. Self-host with Docker Compose. License is Apache-2.0 plus conditions (multi-tenant SaaS needs a commercial license).

Editorial update 2026-09-05; GitHub statistics snapshot 2026-10-19.

158.9k

GitHub stars

25.1k

Forks

Self-hostable · Validate your use case

Status

Apache-2.0 + conditions

License

Open GitHub Repo Visit dify.ai Start with Quickstart

From knowledge files to a draft for review

A complete documented exercise, with offline files checked only. These Dify operations were not run; authorize model cost and data egress first.

Check synthetic material before upload

Only in an approved private workspace, upload upload-faq.txt—not test questions or expectations. Inspect all 12 chunks for synthetic notice, Source-ID and complete Q&A.

upload-faq.txt → 12 language-specific chunks (verify in the actual parser)

Check retrieval before connecting answers

Record actual passages and both retrieval-setting layers. Empty results abstain; nonempty results must bind LLM Context. Verify the actual UI/version combination.

Query → Retrieval → empty: fixed abstention / nonempty: LLM → draft

Open citations and review every draft

Map native citations back to passages. Price, delivery, certification and outcomes lack evidence and stay unknown. Scores/labels do not verify facts; no sending or completed approval.

Supported answer / Unknowns / Human review required

Pitfalls

⚠️

Two-layer data boundary

App data can stay on your servers; model calls still usually leave for an external API. Audit that hop — self-host ≠ data never leaves.

⚠️

Don't spider-web the canvas

Split by responsibility and define each branch's inputs, outputs, failures and tests. Node count is not a maintainability measurement; make failures traceable before expanding.

⚠️

Extra terms: multi-tenant and logo

Reselling Dify as multi-tenant AI, or stripping the console logo, hits the extra LICENSE terms. Read them before you productize.

Beside n8n & Ollama

Based on official docs and licenses. Dify is the app layer, n8n the flow layer, Ollama the model layer. They stack.

Tool	License	Surface	Positioning	Best for
Dify	Apache-2.0 + conditions	AI app workspace	RAG + agent workflows + many models	SME PoC, knowledge Q&A
n8n	Fair-code · SUL	Visual workflows	System integration + cron/webhooks	Repetitive business flows
Ollama	MIT (models separate)	Local model runtime	One command to run open models locally	Private / offline inference

Official Directory

Official

Official site (dify.ai)
Product overview, cloud and self-host.
GitHub repository
Source. Apache-2.0 with additional conditions.

Docs & Install

Documentation
Self-host, RAG, workflows and model providers.
Docker Compose deploy
Official self-host install path.
docker/README in repo
Compose, .env and upgrade notes.

GitHub Releases
Read the target-version notes before upgrading.

Latest Updates

2026-09-04 · GitHub · langgenius/dify
Dify keeps shipping: agentic workflows + RAG in one workspace
From LangGenius; cloud, VPC or self-host. LICENSE is Apache-2.0 with extra conditions. 150k+ stars.

What Dify Is: a Visual AI App and RAG Workspace

LaoJin's Pick

149.5k ★ · 2026-07

Dify

For companies that want AI apps fast without writing everything: visual orchestration, RAG, dozens of model providers, access control — all self-hostable. The default base for an SME's AI proof-of-concept.

FAQ

What is Dify?: Dify is a visual AI app builder: RAG, agentic workflows and many model providers in one workspace. Self-host or use official cloud. Made by LangGenius.
What should I know about Dify's license?: Apache-2.0 plus conditions: operating a multi-tenant service (one workspace = one tenant) needs a commercial license; don't strip logo/copyright from the official frontend. Single-tenant internal use is typically fine. The repo LICENSE wins.
How do I self-host Dify?: Official docs want Docker Compose 2.24+ and at least 2 CPU / 4 GiB RAM. Clone, `cd dify/docker`, copy `.env.example` to `.env`, `docker compose up -d`. First visit the install page on local port 80 to create the admin.
Dify vs n8n?: Knowledge Q&A, chat apps, models as product features → Dify. Cron, webhooks, wiring existing SaaS/internal systems → n8n. Many stacks run both.
Has the knowledge-Q&A exercise actually run in Dify?: No. Official docs, six bilingual synthetic FAQs, 26 questions and 22 native offline tests were checked. Dify, embeddings, models, native citations and actual human review were not run; later execution needs separate authorization. Expected results are not observations.

Evidence: docs + offline checks · Dify and models not run

From source to draft: separate support from unknowns

Six bilingual synthetic FAQs, 26 questions and a citation checklist. File integrity is checked; upload, retrieval, models and native citations need separately authorized runtime testing.

Follow the complete walkthrough →

Connecting an existing operation? Start with field names, source permissions and who may approve—not raw customer records.

Discuss input and review requirements View service scope

Cite This Hub

We encourage you to capture this hub into your knowledge base, research, or feed it straight to an AI agent under CC BY 4.0.

Raw Markdown Endpoint (For AI Agents)

Text Citation (For Articles/Reports)

Laojin Global. "Dify Hub". Last updated 2026-09-05. Available at: https://www.laojinchuhai.com/en/dify

How this hub stays current

Version and links live in src/content/dify.ts; star snapshots sync via GitHub API.