main.py

import base64
from typing import Annotated, Optional
from fastapi import (Body, FastAPI, File, Form, HTTPException, UploadFile,
                     status)
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import JSONResponse
from pydantic import AnyHttpUrl, UrlConstraints

from config import settings
import uvicorn
from utils.audio_utils import AudioUtils
from utils.caption_utils import ImageCaptioning
from utils.image_utils import UrlTest
from utils.topic_generation import TopicGenerator

app = FastAPI(
    title=settings.PROJECT_NAME,
)

# CORS
if settings:
    app.add_middleware(
        CORSMiddleware,
        allow_origins='*',
        allow_credentials=True,
        allow_methods=["*"],
        allow_headers=["*"],
    )

topic_generator = TopicGenerator()
img_caption = ImageCaptioning()
audio_utils = AudioUtils()
utils = UrlTest()


@app.get("/")
def root():
    return {"message": "Welcome To Rediones API"}


@app.get("/health")
def health():
    return {"message": "OK"}


@app.post("/topicgen")
def generate_topic(
    img: UploadFile = File(
        default=None,
        description="Image file. It mutually excludes ImgUrl",
        # regex=r"^.+\.(jpg|png|jpeg)$"
    ),
    text: Annotated[Optional[str], Form()] = None,
    img_url: Annotated[
        Optional[AnyHttpUrl],
        UrlConstraints(allowed_schemes=["https"]),
        Form(description=(
            "Image url only accepts https scheme. It mutually excludes Img"
        ))
    ] = None,
):
    if img_url and img:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail="Only one of image_url or img can be accepted"
        )
    # if only text is provided
    elif text is not None and img is None and img_url is None:
        generated_topics = topic_generator.generate_topics(text)
        return {"topics": generated_topics}

    # if image/image_url is provided with or without text
    elif img or img_url or text:
        img_file_object = None  # initialize img_file_object
        # decide whether img or img_url is provided
        if img:
            # image file must be ended with .jpg, .png, .jpeg
            if not str(img.filename).endswith(
                (".jpg", ".png", ".jpeg")
            ):
                raise HTTPException(
                    status_code=status.HTTP_400_BAD_REQUEST,
                    detail="Image file must be ended with .jpg, .png, .jpeg"
                )
            img_file_object = img.file
        elif img_url:
            img_file_object = utils.load_image(img_url)

        # decide whether text is provided
        if text is None:
            capt = img_caption.get_caption(img_file_object)
        else:
            capt = str(text) + "." + img_caption.get_caption(img_file_object)

        generated_topics = topic_generator.generate_topics(capt)
        return {"topics": generated_topics}

    else:
        raise HTTPException(
            status_code=status.HTTP_400_BAD_REQUEST,
            detail="enter text or image. "
            "imageurl and img are mutually exclusive"
        )


@app.post("/audioverse")
def generate_audio(
    text: Annotated[str, Body(description="Text to be transcribed.")]
):
    if text is not None:
        audio_bytes = audio_utils.speak(text)
        audio_base64 = base64.b64encode(audio_bytes).decode("utf-8")
        return JSONResponse(content={"audio_base64": audio_base64})

@app.post("/transcribe")
def transcribe_audio(
    audio: UploadFile = File(
        default=None,
        description="Audio file to be transcribed."
    )
):
    if audio is not None:
        audio_transcribe = audio_utils.improved_transcribe(0.8, audio_file=audio.file)
        return JSONResponse(content={"audio_transcription": audio_transcribe})
    

if __name__ == "__main__":
    uvicorn.run(app, host="127.0.0.1", port=8000)