Deploy to a Workflow via Dify

Dify is an open-source LLM application development platform with AI workflows, RAG pipelines, agents and model management. You can connect your WeClone clone to Dify and use it in workflows.

1. Deploy Dify: Install and configure it following the Dify documentation, or use the cloud service.

   • Dify Cloud
   • Deploy the Dify Community Edition with Docker Compose

2. Start the WeClone server and generate a token: Follow Server and Bot Token. If Dify runs in Docker on Linux, add --host 172.17.0.1:

   weclone-cli server --inference --port 8005 --host 172.17.0.1

3. Configure the model provider: In Dify, go to Settings → Model Providers → OpenAI → Add Model and fill in:

   • Model name: gpt-3.5-turbo or another GPT model name
   • API key: the generated token
   • API base:
     • Dify in Docker on Linux: http://172.17.0.1:8005
     • Dify on Docker Desktop (macOS, Windows): http://host.docker.internal:8005
     • Dify Cloud needs the local server exposed to the internet: http://<public IP>:8005

   WARNING

   Exposing the local server to the internet is not recommended. The token decrypts all data. See Data Security.

   

4. Add a WeClone model node to a workflow: In the Dify workflow editor, add an LLM node, bind it to the model you just added (such as gpt-3.5-turbo), and set its SYSTEM prompt to the default_system used for fine-tuning. With the default configuration, enter:

   请你扮演一名人类，不要说自己是人工智能

   

---

Your clone should now be connected to Dify. Chat with it to see how it does.