This is my own proxy for OpenAI to replace GPT models with Google Gemini ones.
Tested on:
- SillyTavern
- Python OpenAI package
- Cursor
git clone https://github.com/obezbolen67/openai-to-gemini-proxy.git cd openai-to-gemini-proxy npm inode server.js~$ node server.js Proxy server running on port 3333 My proxy provides video, image and audio input. For now you can send your media through direct links (works well with discord attachments or imgur images)
from openai import OpenAI base_url = "http://localhost:3333/v1" API_KEY = "your_api_key" model = OpenAI(api_key=API_KEY, base_url=base_url) response = model.chat.completions.create( model="gemini-1.5-flash", messages=[ { "role": "user", "content": [ { "type": "text", "text": "Describe the image in every detail." }, { "type": "image_url", "image_url": { "url": "https://i.natgeofe.com/n/548467d8-c5f1-4551-9f58-6817a8d2c45e/NationalGeographic_2572187_square.jpg", }, }, ], } ] ) print(response.choices[0].message.content) # The image is a close-up shot of a cat's face agai...from openai import OpenAI base_url = "http://localhost:3333/v1" API_KEY = "your_api_key" model = OpenAI(api_key=API_KEY, base_url=base_url) response = model.chat.completions.create( model="gemini-1.5-flash", messages=[ { "role": "user", "content": [ { "type": "text", "text": "Describe the video in every detail." }, { "type": "video_url", "video_url": { "url": "https://www.dropbox.com/scl/fi/oss8nx5p4ck4u3bcfz24d/2024-06-18-19-33-36.mp4?rlkey=pl751s7kcqgeksdjs4hx6n5um&st=cp5uzd7h&dl=1", }, }, ], } ] ) print(response.choices[0].message.content) # The video shows a screen recording of a computer running Python code to detect objects in the Minecraft game. The code is in the left half of the screen, and the Minecraft game is in the right half of the screen.from openai import OpenAI base_url = "http://localhost:3333/v1" API_KEY = "your_api_key" model = OpenAI(api_key=API_KEY, base_url=base_url) response = model.chat.completions.create( model="gemini-1.5-flash", messages=[ { "role": "user", "content": [ { "type": "text", "text": "Make subtitles for this audio." }, { "type": "audio_url", "audio_url": { "url": "https://www.eslfast.com/robot/audio/smalltalk/smalltalk0101.mp3", }, }, ], } ] ) print(response.choices[0].message.content) # 00:00 Hi, how are you doing? \n 00:02 I'm fine, how about yourself? \n 00:04 I'm pretty good. Thanks for asking.If Gemini blocked in your region or you want to have a remote server, you can deploy repo on Render, although it can delay requests up to 2 minutes when inactive. But be warned that Render storage is limited so avoid uploading large videos to API.
/v1/chat/completions(with streaming support!)/v1/models
And, if you want to support me, you can buy me a coffe or become my patron!
