ChatGPT + noVNC + Docker

You must have docker or podman to be able to run this repo.

There is a docker-compose file that will build a docker image with all the dependencies and run it.

You can use http://localhost:8080/docs to test the API or you can use postman.

(Optional) There is also a postman collection that you can use to test the API.

Create a .env file with the following content:

touch .env
echo "VNC_GEOMETRY=1600x900" > .env
echo "VNC_PASSWD=<your vnc password>" >> .env
echo "USER_PASSWD=<container's user password>" >> .env
echo "EMAIL=<your openai account email>" >> .env
echo "PASSWORD=<your openai account password>" >> .env
echo "CHATGPT_API_SERVER=chatgpt-api-server-link:8080" >> .env

How to run

docker compose up --build

Port 8080 is exposed for fastapi server Port 9000 is exposed for noVNC

http://localhost:9000 allows you to connect to docker container via graphical interface.

For some reason sometimes the noVNC UI does not load properly, if that happens try with other browsers.

How to use

(Optional) Open postman and import the collection from postman_collection.json

1. You can also use http://localhost:8080/docs to test the API

2. Make a request to http://localhost:8080/login, it will open a browser window to login to openai.

3. Wait until it enters the email (it will wait for you to solve the captcha if there is one)

4. Head to the browser window in the noVNC (http://localhost:9000) and solve the captcha.

5. After you solve the captcha make a request to http://localhost:8080/reset-wait to continue the login process.

   • At the end of the login process it will try to create a websocket connection with the ChatGPT-API-server.
   • You can check the logs of the ChatGPT-API-server to see if the connection was successful.
   • You can then use ChatGPT-API-server /api/ask endpoint to make requests to the ChatGPT model.

6. After you login you can make a request to http://localhost:8080/cookies and http://localhost:8080/access-token to make sure you are logged in.

7. Now you can make a request to http://localhost:8080/backend-api/conversation with the same payload you would make to openai's api.

{
    "action": "next",
    "messages": [
        {
            "id": "asdasd1-1231sdad-121",
            "role": "user",
            "content": {
                "content_type": "text",
                "parts": [
                    "Hello, Assistant!",
                ],
            },
        },
    ],
    "parent_message_id": "1sasd3-asc24-asse234",
    "model": "text-davinci-002-render",
}

Response will be as follows:

{
    "status": "",
    "error": [],
    "data": {
        "message": {
            "id": "51d335cd-3d33-4f0f-bfa1-2f2be6256713", 
            "role": "assistant", 
            "user": null, 
            "create_time": null, 
            "update_time": null, 
            "content": {
                "content_type": "text", 
                "parts": [
                    "ChatGPT's response"
                ]
            }, 
            "end_turn": null, 
            "weight": 1.0, 
            "metadata": {}, 
            "recipient": "all"
        }, 
        "conversation_id": "b15d6133-d792-4568-9f8e-71bee1782e28", 
        "error": null
    }
}

Note:

It does not stream the response, it waits until the response is ready and then returns it since we are using tls-client instead of httpx or requests becuse of the tls fingerprinting issue with other http clients.

Demo

docker-chatgpt01.mp4