OpenAI Proxy Docker

This repository provides a Dockerized proxy for accessing the OpenAI API, allowing for simplified and streamlined interaction with the model.

With the Docker image, you can easily deploy a proxy instance to serve as a gateway between your application and the OpenAI API, reducing the complexity of API interactions and enabling more efficient development.

Use case

1. For users who are restricted from direct access to the OpenAI API, particularly those in countries where OpenAI will be blocking API access starting July 2024
2. For users who need to access private APIs that lack Cross-Origin Resource Sharing (CORS) headers, this solution provides a proxy to bypass CORS restrictions and enable seamless API interactions.

Demo

• API demo https://api.aiql.com

• UI demo ChatUI

For detailed usage of OpenAI API, please check:

• OpenAI API Reference (official docs)

• RESTful OpenAPI (provided by AIQL)

How to use

Just:

sudo docker run -d -p 9017:9017 aiql/openai-proxy-docker:latest

Then, you can use it by YOURIP:9017

For example, the proxied OpenAI Chat Completion API will be: YOURIP:9017/v1/chat/completions

It should be the same as api.openai.com/v1/chat/completions

You can change default port and default target by setting -e in docker, which means that you can use it for any backend followed by OpenAPI format:

Parameter	Default Value
PORT	9017
TARGET	https://api.openai.com

How to dev

Click below to use the GitHub Codespace:

Or fork this repo and create a codespace manually:

1. Wait for env ready in your browser
2. npm install ci
3. npm start

And then, the codespace will provide a forward port (default 9017) for you to check the running.

If everything is OK, check the docker by:

docker build .

Docker Push

If you want to maintaine your own docker image, refer to github Actions

Fork this repo and set DOCKERHUB_USERNAME and DOCKERHUB_TOKEN in your secrets

Normally, the step should be:

1. Fork this repo
2. Settings → Secrets and variables → Actions → New repository secret

Docker Compose

Example 1

You can apply this approach to other APIs, such as Nvidia NIM:

• The proxied Nvidia NIM Completion API will be: YOURIP:9101/v1/chat/completions

  For convenience, a readily available API is provided for those who prefer not to deploy it independently: https://nvidia.aiql.com/v1/chat/completions

services:
  nvidia-proxy:
    image: aiql/openai-proxy-docker:latest
    container_name: nvidia-proxy
    environment:
      PORT: "9101"
      TARGET: "https://integrate.api.nvidia.com"
    restart: always
    network_mode: host

Example 2

You can apply this approach with your own domain over HTTPS:

• YOUREMAILADDR@example.com will be used to get certification notification from ACME server
• The proxied OpenAI Chat Completion API will be: api.example.com/v1/chat/completions

  api.example.com should be replaced by your domain name

services:
  nginx-proxy:
    image: nginxproxy/nginx-proxy
    container_name: nginx-proxy
    ports:
      - "80:80"
      - "443:443/tcp"
      - "443:443/udp"
    environment:
      ENABLE_HTTP3: "true"
    volumes:
      - conf:/etc/nginx/conf.d
      - vhost:/etc/nginx/vhost.d
      - html:/usr/share/nginx/html
      - certs:/etc/nginx/certs:ro
      - /var/run/docker.sock:/tmp/docker.sock:ro
    restart: always
    network_mode: bridge

  acme-companion:
    image: nginxproxy/acme-companion
    container_name: nginx-proxy-acme
    environment:
      - DEFAULT_EMAIL=YOUREMAILADDR@example.com
    volumes_from:
      - nginx-proxy
    volumes:
      - certs:/etc/nginx/certs:rw
      - acme:/etc/acme.sh
      - /var/run/docker.sock:/var/run/docker.sock:ro
    network_mode: bridge

  openai-proxy:
    image: aiql/openai-proxy-docker:latest
    container_name: openai-proxy
    environment:
      LETSENCRYPT_HOST: api.example.com
      VIRTUAL_HOST: api.example.com
      VIRTUAL_PORT: "9017"
    network_mode: host
    depends_on:
      - "nginx-proxy"

volumes:
  conf:
  vhost:
  html:
  certs:
  acme: