main.py

import os
from fastapi import FastAPI, File, Request, UploadFile, Body, Depends, HTTPException
from fastapi.security.api_key import APIKeyHeader
from typing import Optional, Annotated
from fastapi.encoders import jsonable_encoder
from PIL import Image
from io import BytesIO
import pytesseract
from nltk.tokenize import sent_tokenize
from transformers import MarianMTModel, MarianTokenizer

API_KEY = os.environ.get("API_KEY")

app = FastAPI()
api_key_header = APIKeyHeader(name="api_key", auto_error=False)

def get_api_key(api_key: Optional[str] = Depends(api_key_header)):
    if api_key is None or api_key != API_KEY:
        raise HTTPException(status_code=401, detail="Unauthorized access")
    return api_key

@app.post("/api/ocr", response_model=dict)
async def ocr(
    image: UploadFile = File(...),
    # languages: list = Body(["eng"])
    request: Request,
    # body: dict = Body(...),
    api_key: str = Depends(get_api_key),
):
    # print("[where?] outside try block")
    # print("[image]", image)
    try:
        # print("[where?] inside try block")
        # content = await image.read()
        # # image = Image.open(file.file)
        # print("[content]",content)
        # image = Image.open(BytesIO(content))
        # print("[image]",image)
        # text = pytesseract.image_to_string(image, lang="+".join(languages))
        # print("[text]",text)
        message = "file uploaded"
    except Exception as e:
        return {"error": str(e)}, 500

    # return jsonable_encoder({"text": text})
    return {"messageDetails": message}

@app.post("/api/translate", response_model=dict)
async def translate(
    api_key: str = Depends(get_api_key),
    text: str = Body(...),
    src: str = "en",
    trg: str = "zh",
):
    if api_key != API_KEY:
        return {"error": "Invalid API key"}, 401

    tokenizer, model = get_model(src, trg)

    translated_text = ""
    for sentence in sent_tokenize(text):
        translated_sub = model.generate(**tokenizer(sentence, return_tensors="pt"))[0]
        translated_text += tokenizer.decode(translated_sub, skip_special_tokens=True) + "\n"

    return jsonable_encoder({"translated_text": translated_text})

def get_model(src: str, trg: str):
    model_name = f"Helsinki-NLP/opus-mt-{src}-{trg}"
    tokenizer = MarianTokenizer.from_pretrained(model_name)
    model = MarianMTModel.from_pretrained(model_name)
    return tokenizer, model