Update app.py

app.py

@@ -1,66 +1,40 @@
-from fastapi import FastAPI
-from pydantic import BaseModel
-from fastapi.middleware.cors import CORSMiddleware
-from transformers import MarianMTModel, MarianTokenizer, pipeline
+from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer
 import gradio as gr
 
-# Load translation models (xx -> en)
-models = {
-    "ar": pipeline("translation", model="Helsinki-NLP/opus-mt-ar-en"),
-    "ur": pipeline("translation", model="Helsinki-NLP/opus-mt-ur-en"),
-    "hi": pipeline("translation", model="Helsinki-NLP/opus-mt-hi-en"),
-}
-
-# Dummy fastapi backend
-app = FastAPI()
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-class TranslationRequest(BaseModel):
-    text: str
-
-def detect_language(text: str) -> str:
-    """Very basic language detector based on script/characters."""
-    if any("\u0600" <= c <= "\u06FF" for c in text):  # Arabic/Urdu
-        # Urdu detection heuristic
-        if "ے" in text or "ں" in text:  
-            return "ur"
-        return "ar"
-    elif any("\u0900" <= c <= "\u097F" for c in text):  # Devanagari -> Hindi
-        return "hi"
-    return "en"
-
-@app.post("/translate")
-def translate(req: TranslationRequest):
-    lang = detect_language(req.text)
-    if lang == "en":
-        return {"input": req.text, "translation": req.text, "lang": "en"}
+# Load M2M100 multilingual model (supports 100+ languages)
+model_name = "facebook/m2m100_418M"
+tokenizer = M2M100Tokenizer.from_pretrained(model_name)
+model = M2M100ForConditionalGeneration.from_pretrained(model_name)
+
+def translate_to_english(text):
+    # Detect language (roughly by script, can be improved later)
+    if any("\u0600" <= c <= "\u06FF" for c in text):   # Arabic/Urdu script
+        lang = "ar"  # works for Urdu too, since M2M100 uses ISO codes
+    elif any("\u0900" <= c <= "\u097F" for c in text): # Hindi (Devanagari)
+        lang = "hi"
     else:
-        translated = models[lang](req.text)[0]["translation_text"]
-        return {"input": req.text, "translation": translated, "lang": lang}
+        lang = "en"
 
-# ---- Gradio Interface ----
-def chat_interface(user_input):
-    lang = detect_language(user_input)
+    # If already English, return as is
     if lang == "en":
-        return f"🗣 You: {user_input}\n\n🌐 Detected: English\n\n✅ Translation: {user_input}"
-    else:
-        translated = models[lang](user_input)[0]["translation_text"]
-        return f"🗣 You: {user_input}\n\n🌐 Detected: {lang.upper()}\n\n✅ Translation: {translated}"
+        return f"🗣 You: {text}\n🌐 Detected: English\n✅ Translation: {text}"
+
+    # Set tokenizer to source lang
+    tokenizer.src_lang = lang
+    encoded = tokenizer(text, return_tensors="pt")
+    generated_tokens = model.generate(**encoded, forced_bos_token_id=tokenizer.get_lang_id("en"))
+    translation = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
+
+    return f"🗣 You: {text}\n🌐 Detected: {lang.upper()}\n✅ Translation: {translation}"
 
+# ---- Gradio UI ----
 with gr.Blocks(css=".gradio-container {font-family: 'Poppins', sans-serif;}") as demo:
-    gr.Markdown("## 🌍 Multilingual → English Chatbot")
-    gr.Markdown("Type anything in **Arabic, Urdu, Hindi, or English**, and I’ll always answer in English!")
+    gr.Markdown("## 🌍 Multilingual → English Chatbot (Arabic, Urdu, Hindi, English)")
 
     with gr.Row():
-        inp = gr.Textbox(placeholder="Type here...", label="Your Message")
+        inp = gr.Textbox(placeholder="Type something here...", label="Your Message")
     out = gr.Textbox(label="Chat Response", interactive=False)
 
-    inp.submit(chat_interface, inp, out)
+    inp.submit(translate_to_english, inp, out)
 
 demo.launch(server_name="0.0.0.0", server_port=7860)