The new version is the trend, but if there are problems with the new version, you can switch back to the old version. The old version is still usable and based on the browser, which should be an ultimate solution that can be used for a long time (although it may consume slightly more resources).

English countries does not have the "Great Firewall", so many issues are gone.

More details: linweiyuan/go-chatgpt-api#74

• ChatGPT user login (accessToken will be returned) (currently Google or Microsoft accounts are not supported).

POST /chatgpt/login

{
  "username": "email",
  "password": "password"
}

• get conversation list

GET /chatgpt/conversations?offset=0&limit=20

offset defaults to 0, limit defaults to 20 (max 100).

• get conversation content

GET /chatgpt/conversation/{conversationID}

• create conversation

POST /chatgpt/conversation

{
  "action": "next",
  "messages": [
    {
      "id": "message id",
      "author": {
        "role": "user"
      },
      "content": {
        "content_type": "text",
        "parts": [
          "Hello World"
        ]
      }
    }
  ],
  "parent_message_id": "parent message id",
  "conversation_id": "conversation id",
  "model": "text-davinci-002-render-sha",
  "timezone_offset_min": -480,
  "history_and_training_disabled": false
}

• generate conversation title

POST /chatgpt/conversation/gen_title/{conversationID}

{
  "message_id": "role assistant response message id"
}

• rename conversation

PATCH /chatgpt/conversation/{conversationID}

{
  "title": "new title"
}

• delete conversation

PATCH /chatgpt/conversation/{conversationID}

{
  "is_visible": false
}

• delete all conversations

PATCH /chatgpt/conversations

{
  "is_visible": false
}

• feedback message

POST /chatgpt/conversation/message_feedback

{
  "message_id": "message id",
  "conversation_id": "conversation id",
  "rating": "thumbsUp/thumbsDown"
}

• platform user login (sessionKey will be returned)

POST /platform/login

{
  "username": "email",
  "password": "password"
}

• List models

GET /platform/v1/models

• Retrieve model

GET /platform/v1/models/{model}

• Create completion

POST /platform/v1/completions

{
  "model": "text-davinci-003",
  "prompt": "Say this is a test",
  "max_tokens": 7,
  "temperature": 0,
  "stream": true
}

• Create chat completion

POST /platform/v1/chat/completions

{
  "messages": [
    {
      "role": "user",
      "content": "Hello World"
    }
  ],
  "model": "gpt-3.5-turbo",
  "stream": true
}

• Create edit

POST /platform/v1/edits

{
  "model": "text-davinci-edit-001",
  "input": "What day of the wek is it?",
  "instruction": "Fix the spelling mistakes"
}

• Create image

POST /platform/v1/images/generations

{
  "prompt": "A cute dog",
  "n": 2,
  "size": "1024x1024"
}

• Create embeddings

POST /platform/v1/embeddings

{
  "model": "text-embedding-ada-002",
  "input": "The food was delicious and the waiter..."
}

• List files

GET /platform/v1/files

• get credit grants (only support sessionkey)

GET /platform/dashboard/billing/credit_grants

• get subscription (only support sessionkey)

GET /platform/dashboard/billing/subscription

• get api keys (only support sessionkey)

GET /platform/dashboard/user/api_keys

To set a proxy, you can use the environment variable GO_CHATGPT_API_PROXY, such as GO_CHATGPT_API_PROXY=http://127.0.0.1:20171 or GO_CHATGPT_API_PROXY=socks5://127.0.0.1:20170. If it is commented out or left blank, it will not be enabled.

To use with warp: GO_CHATGPT_API_PROXY=socks5://chatgpt-proxy-server-warp:65535. Since the scenario that requires setting up warp can directly access the ChatGPT website by default, using the same variable will not cause conflicts.

docker-compose.yaml:

services:
  go-chatgpt-api:
    container_name: go-chatgpt-api
    image: linweiyuan/go-chatgpt-api
    ports:
      - 8080:8080
    environment:
      - GIN_MODE=release
      - GO_CHATGPT_API_PROXY=
    restart: unless-stopped

I only develop and test on Arch Linux, which is a rolling release version, meaning that everything on the system is up-to-date. If you encounter a yaml error while using it, you can add version: '3' in front of services:.

If you encounter an Access denied error, but your server is indeed in Supported countries and territories, try this configuration (it is not guaranteed to solve the problem, for example, if your server is in Zone A, but Zone A is not on the list of supported countries, even if you use warp to change to a Cloudflare IP, the result will still be 403):

services:
  go-chatgpt-api:
    container_name: go-chatgpt-api
    image: linweiyuan/go-chatgpt-api
    ports:
      - 8080:8080
    environment:
      - GIN_MODE=release
      - GO_CHATGPT_API_PROXY=socks5://chatgpt-proxy-server-warp:65535
    depends_on:
      - chatgpt-proxy-server-warp
    restart: unless-stopped

  chatgpt-proxy-server-warp:
    container_name: chatgpt-proxy-server-warp
    image: linweiyuan/chatgpt-proxy-server-warp
    environment:
      - LOG_LEVEL=OFF
    restart: unless-stopped

If you want to make sure the image is always latest, try this:

services:
  watchtower:
    container_name: watchtower
    image: containrrr/watchtower
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
    command: --interval 3600
    restart: unless-stopped