How to Deploy Dify and Ollama Locally on Windows 11: A Step‑by‑Step Guide

1. Enable Hyper‑V

Open Control Panel → Programs → Turn Windows features on or off and check all Hyper‑V options. This feature is only available in Windows 11 Pro.

2. Install Docker Desktop and configure domestic mirrors

After installing Docker Desktop, edit its daemon configuration ( daemon.json) to use faster Chinese registries:

{
  "registry-mirrors": [
    "https://docker.1ms.run",
    "https://docker.kejilion.pro",
    "https://docker-0.unsee.tech",
    "https://dhub.kubesre.xyz",
    "https://docker.tbedu.top",
    "https://hub.crdz.gq",
    "https://image.cloudlayer.icu"
  ]
}

3. Relocate Docker storage

To avoid filling the C: drive, change the disk image location via Settings → Resources → Advanced → Disk image location in Docker Desktop.

4. Install Ubuntu via WSL

Open PowerShell and run: wsl --install After the download finishes, create a Linux user name and password as prompted.

5. Clone the Dify repository

# Switch to a user directory to avoid permission issues
cd C:\Users\<strong>your-username</strong>\
# Clone from GitHub (or use the Gitee mirror if GitHub is slow)
git clone https://github.com/langgenius/dify.git
# Alternative mirror
# git clone https://gitee.com/langchain-ai/dify.git

cd dify\docker

If the clone fails, an accelerated link such as https://gitcode.com/GitHub_Trending/di/dify/tags can be used.

6. Prepare the environment file

Copy the example file and change the default port (80) because it is already occupied: cp .env.example .env Edit .env and set PORT=8100 (or any unused port).

7. Launch the containers

Remove stale containers if a previous attempt left them behind: docker compose down -d Start the services: docker compose up -d The startup logs show the containers initializing, and the installation page becomes reachable.

8. Verify the service

Open a browser and navigate to http://localhost:8100/install . The UI shows an init_permissic log entry, which indicates successful startup.

9. Configure Ollama in Dify

In the Dify UI, click the avatar → Settings → Model Provider, locate the Ollama card and press “Install”. Fill in the following fields (values must match the output of ollama list):

Model Name : qwen2.5:7b (exactly as shown by ollama list)

Model Type : LLM

Base URL : http://host.docker.internal:11434 (or http:// your‑IP :11434 for cross‑machine access)

API Key : any string, e.g., ollama-local (Ollama itself does not require a key)

10. Final test

After saving the configuration, the Dify interface can invoke the selected Ollama model. The setup has been verified to work on a Windows 11 Pro machine with RTX 4060 Ti, 32 GB RAM, i7‑10700KF and a 1 TB SSD.

References

Dify official site: https://dify-china.com/

Ollama Chinese community: https://ollamacn.github.io/