Imagine you run a small tutoring business and want a chatbot that answers parents’ questions using only your own course info and price list — without paying a SaaS per month. Or you want a tool that reads your 50 training documents and answers any question about them. Dify lets you build exactly this by dragging boxes on a canvas, no code. It’s one of the most popular open-source AI platforms on GitHub (roughly 150k stars), and it runs free on your own machine.

What you need before you start

A Windows 10/11 PC with at least 2 CPU cores and 4 GB of RAM (8 GB is comfortable), Docker Desktop installed and running, and a model access key or local model. If you already followed our Open WebUI guide, your Docker setup is ready — Dify reuses it.

Step 1 — Install Docker Desktop

Download Docker Desktop from docker.com, install with defaults, start it once and wait until the whale icon is steady. This is the same foundation we used for Open WebUI — if it’s already running, skip ahead.

Step 2 — Download and start Dify

Open PowerShell and run these three lines in order:

git clone https://github.com/langgenius/dify.git cd dify/docker copy .env.example .env docker compose up -d

The first launch downloads several container images, so give it 5–10 minutes on a normal connection. When the command finishes, Dify is running.

Dify official website (English interface)
Dify — build production-grade AI apps with a visual canvas.

Step 3 — Open your Dify and create an account

Open http://localhost in your browser. On first visit Dify asks you to set up an admin account — that account lives on your machine. Then you land on the dashboard, where you can start a blank app or use a template.

Step 4 — Connect a model and build your first app

Open Settings → Model Provider and add a model: paste an OpenAI/Anthropic API key, or point to your local Ollama (API URL http://localhost:11434) if you’d rather stay fully offline. Then create a “Chatbot” app, drag in your knowledge documents under “Knowledge”, and connect the boxes: user question → model → answer. Click Publish and you have a private AI assistant you can share with your family or team.

Common beginner pitfalls

  • “git” is not recognized — install Git for Windows from git-scm.com, or download the Dify zip from GitHub and extract it instead.
  • docker compose fails with “port already in use” — port 80 is taken. Edit .env and change the app port (e.g. APP_PORT=8080), then retry; open http://localhost:8080.
  • No model available — you must add a model in Settings → Model Provider first; the app won’t work without one.
  • First start is slow — image downloads are large; later starts are quick.
  • Knowledge answers are wrong — upload clean documents (PDF/TXT), avoid scanned images, and set the retrieval mode in Knowledge settings.

Bottom line

Dify turns your PC into an AI app factory: drag, connect, publish. Perfect for a family knowledge base, a shop’s customer bot, or just learning how AI apps work. Next in this series: OpenHands, the AI that writes code for you.

Official Websites (Avoid Phishing)

Fake and phishing sites often rank high in search results. Only use the official URLs below.

Open Source & Local AI

Ad · Google AdSense placeholder — paste your ad code here