Create main.py

main.py

@@ -0,0 +1,67 @@
+mport os
+
+os.system("sudo apt-get install xclip")
+
+import nltk
+import pyclip
+import pytesseract
+from nltk.tokenize import sent_tokenize
+from transformers import MarianMTModel, MarianTokenizer
+# Newly added below
+from fastapi import FastAPI, File, UploadFile, Body, Depends, HTTPException
+from fastapi.security.api_key import APIKeyHeader
+from typing import Optional
+from fastapi.encoders import jsonable_encoder
+
+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(
+    api_key: str = Depends(get_api_key),
+    image: UploadFile = File(...),
+    languages: list = Body(["eng"])
+):
+    # if api_key != API_KEY:
+    #     return {"error": "Invalid API key"}, 401
+
+    try:
+        text = image_to_string(await image.read(), lang="+".join(languages))
+    except Exception as e:
+        return {"error": str(e)}, 500
+
+    return jsonable_encoder({"text": text})
+
+
+@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