app.py

from fastapi import FastAPI, Depends, Query, File, UploadFile
from fastapi.responses import StreamingResponse
from pydantic import BaseModel
from typing import Annotated
from mistralai import Mistral
from google import genai
from google.genai import types
from auth import verify_token
from enum import Enum
import os
import httpx
import base64
import json

app = FastAPI()

mistral = os.environ.get('MISTRAL_KEY', '')
mistral_client = Mistral(api_key=mistral)

gemini = os.environ.get('GEMINI_KEY', '')
gemini_client = genai.Client(api_key=gemini)

open_router_key = os.environ.get('OPEN_ROUTER_KEY', '')

@app.get("/")
def hello():
    return {"Hello": "World!"}

class LLMRequest(BaseModel):
    model: str
    prompt: str

@app.post("/mistral")
async def mistral(request: LLMRequest, token: Annotated[str, Depends(verify_token)]):
    async def generate():
        response = await mistral_client.chat.stream_async(
            model=request.model,
            messages=[
                {
                    "role": "user",
                    "content": request.prompt,
                }
            ],
        )
        async for chunk in response:
            if chunk.data.choices[0].delta.content is not None:
                yield chunk.data.choices[0].delta.content

    return StreamingResponse(generate(), media_type="text/plain")

@app.post("/gemini")
async def gemini(request: LLMRequest, token: Annotated[str, Depends(verify_token)]):
    async def generate():
        response = gemini_client.models.generate_content_stream(
            model=request.model,
            contents=[request.prompt])

        for chunk in response:
            if chunk.text:
                yield chunk.text

    return StreamingResponse(generate(), media_type="text/plain")

class GeminiMultimodalRequest(BaseModel):
    model: str
    prompt: str
    image: str # url or base64

@app.post("/gemini/multimodal")
async def gemini_multimodal(request: GeminiMultimodalRequest, token: Annotated[str, Depends(verify_token)]):
    if request.image.startswith('http'):
        async with httpx.AsyncClient() as client:
            image = await client.get(request.image)
            #image = types.Part.from_bytes(data=image.content, mime_type="image/jpeg")
    else:
        image = types.Part.from_bytes(data=base64.b64decode(request.image), mime_type="image/jpeg")
    
    response = gemini_client.models.generate_content(
        model=request.model,
        contents=[request.prompt, image]
    )

    return {"response": response.text}

class ModelName(str, Enum):
    deepseek_r1 = "deepseek/deepseek-r1:free"
    gemini_2_flash_lite = "google/gemini-2.0-flash-lite-preview-02-05:free"
    gemini_2_pro = "google/gemini-2.0-pro-exp-02-05:free"
    llama_3_3 = "meta-llama/llama-3.3-70b-instruct:free"
    mistral_small_3 ="mistralai/mistral-small-24b-instruct-2501:free"
    
@app.post("/open-router/text")
async def open_router_text(
    token: Annotated[str, Depends(verify_token)],
    model: ModelName = Query(..., description="Select a model"), 
    prompt: str = Query(..., description="Enter your prompt")
):  
    async with httpx.AsyncClient() as client:
        response = await client.post(
            url="https://openrouter.ai/api/v1/chat/completions",
            headers={
                "Authorization": f"Bearer {str(open_router_key)}",
                "Content-Type": "application/json",
                "HTTP-Referer": "<YOUR_SITE_URL>",  # Optional
                "X-Title": "<YOUR_SITE_NAME>",  # Optional
            },
            json={
                "model": model,
                "messages": [
                    {
                        "role": "user",
                        "content": prompt,
                    }
                ],
            }
        )

        response.raise_for_status()  # Raise HTTPError for bad responses (4xx or 5xx)
        return response.json()

class MultiModelName(str, Enum):
    qwen_vl_plus = "qwen/qwen-vl-plus:free"
    qwen_vl_72b = "qwen/qwen2.5-vl-72b-instruct:free"
    gemini_2_flash_lite = "google/gemini-2.0-flash-lite-preview-02-05:free"
    gemini_2_pro = "google/gemini-2.0-pro-exp-02-05:free"
    llama_3_2_vision = "meta-llama/llama-3.2-11b-vision-instruct:free"

@app.post("/open-router/multimodal-url")
async def open_router_multimodal(
    token: Annotated[str, Depends(verify_token)],
    model: MultiModelName = Query(..., description="Select a model"), 
    prompt: str = Query(..., description="Enter your prompt (ex: What is in this image?)"),
    image_url: str = Query(..., description="Enter the image URL"),
): 
    async with httpx.AsyncClient() as client:
        response = await client.post(
            url="https://openrouter.ai/api/v1/chat/completions",
            headers={
                "Authorization": f"Bearer {str(open_router_key)}",
                "Content-Type": "application/json",
                "HTTP-Referer": "<YOUR_SITE_URL>",  # Optional
                "X-Title": "<YOUR_SITE_NAME>",  # Optional
            },
            json={
                "model": model,
                "messages": [
                {
                    "role": "user",
                    "content": [
                    {
                        "type": "text",
                        "text": prompt,
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                          "url": image_url,
                        }
                    }
                    ]
                }
                ],
            }
        )

        response.raise_for_status()  # Raise HTTPError for bad responses (4xx or 5xx)
        return response.json()

@app.post("/open-router/multimodal-b64")
async def open_router_multimodal_upload(
    token: Annotated[str, Depends(verify_token)], 
    image: UploadFile = File(...),
    model: MultiModelName = Query(..., description="Select a model"),
    prompt: str = Query(..., description="Enter your prompt (ex: What is in this image?)")
):
    image_bytes = await image.read()
    encoded_string = base64.b64encode(image_bytes).decode('utf-8')
    img = f"data:{image.content_type};base64,{encoded_string}"

    async with httpx.AsyncClient() as client:
        response = await client.post(
            url="https://openrouter.ai/api/v1/chat/completions",
            headers={
                "Authorization": f"Bearer {str(open_router_key)}",
                "Content-Type": "application/json",
                "HTTP-Referer": "<YOUR_SITE_URL>",  # Optional
                "X-Title": "<YOUR_SITE_NAME>",  # Optional
            },
            json={
                "model": model,
                "messages": [
                {
                    "role": "user",
                    "content": [
                    {
                        "type": "text",
                        "text": prompt,
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                          "url": img,
                        }
                    }
                    ]
                }
                ],
            }
        )
    
        response.raise_for_status()  # Raise HTTPError for bad responses (4xx or 5xx)
        return response.json()