README_DOCKER.md

Deploying KAI-API with Z.ai Online (Docker)

To run Z.ai (which requires a full browser) online, you must use a Docker container. The best free option is Hugging Face Spaces.

Option A: Hugging Face Spaces (Recommended - Free)

1. Create a Space:

   • Go to Hugging Face Spaces.
   • Click Create new Space.
   • Name: kai-api-gateway.
   • SDK: Select Docker.
   • Space Hardware: CPU basic (free) (2 vCPU, 16GB RAM) is sufficient.
   • Visibility: Public or Private.

2. Upload Files:

   • You can drag and drop your project files, or use Git.
   • Essential files: Dockerfile, requirements.txt, main.py, config.py, engine.py, providers/, static/.

   Git Command:

   git remote add space https://huggingface.co/spaces/YOUR_USERNAME/kai-api-gateway
   git push space main
   

3. Wait for Build:

   • Hugging Face will build the Docker image (takes ~2-3 mins).
   • Once "Running", your API is live!
   • URL: https://YOUR_USERNAME-kai-api-gateway.hf.space

4. Access Your App:

   • Dashboard (UI): https://YOUR_USERNAME-kai-api-gateway.hf.space/
   • API Docs (Swagger): https://YOUR_USERNAME-kai-api-gateway.hf.space/docs
   • Admin Panel: https://YOUR_USERNAME-kai-api-gateway.hf.space/qazmlp

5. Test Z.ai:

   • Go to your Dashboard URL.
   • /models should listed glm-5.
   • Chat with provider: zai — it will work!

Option B: Render / Railway / Koyeb

1. Connect your GitHub repo.
2. Select Docker as the deployment type.
3. Set the internal port to 7860 (or update CMD in Dockerfile to 8080).
4. Deploy.

Note on Vercel: Vercel Serverless Function does NOT support the browser. You can keep your Vercel deployment for lightweight tasks, but use this Docker instance for heavy AI tasks (Z.ai).