import gradio as gr import json def chat_completions(request: gr.Request): # Parse the incoming JSON request data = request.json # Create a placeholder response response = { "id": "chatcmpl-123", "object": "chat.completion", "created": 1677652288, "choices": [{ "index": 0, "message": { "role": "assistant", "content": f"Placeholder response. Received: {data['messages'][-1]['content']}" }, "finish_reason": "stop" }], "usage": { "prompt_tokens": 9, "completion_tokens": 12, "total_tokens": 21 } } return json.dumps(response) demo = gr.Interface( fn=chat_completions, inputs=None, outputs=None, title="Chat Completions API", description="Send a POST request to /v1/chat/completions" ) if __name__ == "__main__": demo.launch() demo.queue() demo.launch() api = gr.mount_gradio_app(demo, "/v1/chat/completions", api_name="chat_completions")