openai api to ollama api translator

server.py

import httpx
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel

app = FastAPI()

# Define a Pydantic model for request input
class GenerateRequest(BaseModel):
    model: str
    prompt: str
    options: dict = None
    system: str = None
    template: str = None
    context: list = None

# Function to make a request to the given endpoint and return the response
async def make_api_request(endpoint, request_data):
    # Define the base URL for the API
    # change as needed
    base_url = "http://localhost:11434/api"

    async with httpx.AsyncClient() as client:
        response = await client.post(f"{base_url}/{endpoint}", json=request_data)

        # Check if the request was successful (HTTP status code 200)
        if response.status_code != 200:
            raise HTTPException(status_code=response.status_code, detail="API request failed")

        return response.json()

# Implement routes to match the ChatGPT API structure
@app.post("/v1/completions")
async def generate_completion(request_data: GenerateRequest):
    # Make a request to the "Generate a completion" endpoint
    response_data = await make_api_request("generate", request_data.dict())

    return response_data

@app.post("/v1/messages")
async def generate_message_completion(request_data: GenerateRequest):
    # Modify the request to match the message-based conversation format
    conversation = [{"role": "user", "content": request_data.prompt}]
    if request_data.context:
        conversation.extend(request_data.context)

    # Make a request to the "Generate a completion" endpoint with the conversation format
    modified_request_data = GenerateRequest(
        model=request_data.model,
        messages=conversation,
        options=request_data.options,
        system=request_data.system,
        template=request_data.template
    )

    response_data = await make_api_request("generate", modified_request_data.dict())

    return response_data


if __name__ == "__main__":
    import uvicorn
    uvicorn.run(app, host="localhost", port=8000)