Run vision AI models on a free Google Colab GPU (or any Linux machine) and use them from your local PC through a public API.
This project lets you run a vision-language model in the cloud and access it locally.
It uses:
- vLLM to serve the model
- Cloudflare Tunnel to create a public URL
- OpenAI-compatible API for easy integration
You can send images and text from your local machine and get AI responses from Colab.
The goal is to make powerful AI models easy to use without needing a strong local computer.
Run the model in the cloud and access it locally through a simple API.
- A Python script runs a vision model using vLLM on Colab or Linux
- A Cloudflare Tunnel exposes the server as a public URL
- Your local PC sends requests (image + prompt)
- The model processes and returns a response
- Works on Colab, local Linux, or cloud GPU
- Uses free Colab GPU
- OpenAI-compatible API
- Supports image and text input
- Public API via Cloudflare Tunnel
- Automatic hardware detection (CPU / GPU / TPU)
- Simple setup
This project is not limited to Colab.
You can run it on:
- Google Colab
- Local Linux (WSL2 supported)
- Cloud GPU servers
It automatically detects your hardware and configures itself.
Follow the instructions in GUIDE.md to:
- Run the server
- Get the public API URL
- Send requests from your local PC
google colab ai, run llm on colab, free gpu ai, vllm server, openai compatible api, cloudflare tunnel api, expose colab gpu as api, multimodal ai, vision language model
Turn free Google Colab GPU into your own OpenAI-style AI server using a single script file. It combines vLLM, Cloudflare tunnel, and vision models into a simple workflow that runs on Colab, local Linux, or cloud GPU with no setup. The script installs dependencies, detects hardware (CPU/GPU/TPU), runs the model, exposes a public API, and lets you send image and text requests from your local PC—no GPU needed, no complex setup, just a clean and practical AI backend.