Update app.py

app.py

@@ -7,13 +7,14 @@ from oauth2client.service_account import ServiceAccountCredentials
 from datetime import datetime
 from gtts import gTTS
 import tempfile
+import requests  
 
 # --- CONFIGURATION ---
 MODEL_K2H_REPO = "ankitklakra/kurukh-to-hindi"
 MODEL_H2K_REPO = "ankitklakra/hindi-to-kurukh"
 SHEET_NAME = "Kurukh_Feedback_Log"
 
-# --- LOAD TEXT MODELS ---
+# --- LOAD MODELS ---
 print("Loading Translation Models...")
 tokenizer = AutoTokenizer.from_pretrained("google/mt5-small")
 model_k2h = AutoModelForSeq2SeqLM.from_pretrained(MODEL_K2H_REPO)
@@ -22,12 +23,22 @@ model_h2k = AutoModelForSeq2SeqLM.from_pretrained(MODEL_H2K_REPO)
 pipe_k2h = pipeline("text2text-generation", model=model_k2h, tokenizer=tokenizer)
 pipe_h2k = pipeline("text2text-generation", model=model_h2k, tokenizer=tokenizer)
 
-# --- LOAD AUDIO MODEL (WHISPER) ---
 print("Loading Voice Model...")
-
 asr_pipeline = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
 
-# --- GOOGLE SHEETS FUNCTION  ---
+# --- HELPER FUNCTIONS ---
+
+def transliterate_to_hindi(text):
+    
+    try:
+        url = "https://inputtools.google.com/request?text={}&itc=hi-t-i0-und&num=1"
+        response = requests.get(url.format(text))
+        result = response.json()
+        # The API returns a nested list; we grab the first suggestion
+        return result[1][0][1][0]
+    except:
+        return text  # If it fails (no internet), return original text
+
 def save_to_sheet(original, translation, correction, direction):
     try:
         json_creds = os.getenv("GOOGLE_CREDENTIALS")
@@ -44,41 +55,32 @@ def save_to_sheet(original, translation, correction, direction):
     except Exception as e:
         return f"❌ Error: {str(e)}"
 
-# --- CORE FUNCTIONS ---
-
 def speech_to_text(audio_path):
-    """Converts Microphone Audio -> Hindi/English Text"""
-    if audio_path is None:
-        return ""
-    print("Transcribing audio...")
-    text = asr_pipeline(audio_path)["text"]
-    return text
+    if audio_path is None: return ""
+    return asr_pipeline(audio_path)["text"]
 
 def text_to_speech(text, language="hi"):
-    """Converts Text -> Audio File"""
-    if not text:
-        return None
+    if not text: return None
     try:
-        # Save audio to a temporary file
         tts = gTTS(text=text, lang=language)
         temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
         tts.save(temp_file.name)
         return temp_file.name
-    except:
-        return None
+    except: return None
 
-def process_translation(text, audio_input, direction):
-    # 1. Handle Input Source (Audio vs Text)
-    if audio_input is not None:
-        # If user spoke, transcribe it first
-        original_text = speech_to_text(audio_input)
-    else:
-        original_text = text
+# --- MAIN TRANSLATION LOGIC ---
+def process_translation(text, audio_input, direction, is_hinglish):
+    
+    # 1. Get Text from Voice or Typing
+    original_text = speech_to_text(audio_input) if audio_input else text
+    if not original_text: return "", "", None
 
-    if not original_text:
-        return "", "", None
+    # 2. Handle Hinglish (NEW FEATURE)
+    # Only applies if translating TO Kurukh (User typing Hindi in English letters)
+    if direction == "Hindi -> Kurukh" and is_hinglish:
+        original_text = transliterate_to_hindi(original_text)
 
-    # 2. Translate
+    # 3. Translate
     target_pipeline = pipe_k2h if direction == "Kurukh -> Hindi" else pipe_h2k
     try:
         results = target_pipeline(original_text, max_length=128)
@@ -86,43 +88,41 @@ def process_translation(text, audio_input, direction):
     except Exception as e:
         return str(e), "", None
 
-    # 3. Handle Audio Output (Read aloud if Hindi)
+    # 4. Audio Output (For Hindi)
     audio_output = None
-    
     if direction == "Kurukh -> Hindi":
         audio_output = text_to_speech(translated_text, "hi")
     
-    # Return: (Input Text shown to user), (Translation), (Audio File)
+    # Return: (Updated Input Box), (Translation), (Audio)
     return original_text, translated_text, audio_output
 
 # --- THE UI ---
 with gr.Blocks() as demo:
     gr.Markdown("# 🇮🇳 AI Kurukh (Kurux) Translator")
-    gr.Markdown("### Voice-Enabled Tribal Language Preserver")
+    gr.Markdown("### Voice & Hinglish Supported")
     
     with gr.Tabs():
-        with gr.TabItem("🗣️ Voice & Text Translator"):
+        with gr.TabItem("🗣️ Translator"):
             
             with gr.Row():
                 direction = gr.Radio(["Kurukh -> Hindi", "Hindi -> Kurukh"], label="Mode", value="Kurukh -> Hindi")
+                # NEW CHECKBOX
+                is_hinglish = gr.Checkbox(label="I am typing Hindi in English (e.g. 'Tumhara')", value=False)
             
-            # INPUT SECTION
             with gr.Row():
                 with gr.Column():
-                    input_text = gr.Textbox(label="Type Here", placeholder="Or use the microphone below...", lines=3)
-                    input_audio = gr.Audio(sources=["microphone"], type="filepath", label="Or Speak (Hindi Only)")
+                    input_text = gr.Textbox(label="Input", placeholder="Type here...", lines=3)
+                    input_audio = gr.Audio(sources=["microphone"], type="filepath", label="Voice Input (Hindi)")
                     translate_btn = gr.Button("Translate 🚀", variant="primary")
                 
-                # OUTPUT SECTION
                 with gr.Column():
                     output_text = gr.Textbox(label="Translation", lines=3, interactive=False)
-                    output_audio = gr.Audio(label="Listen to Translation (Hindi)", interactive=False)
+                    output_audio = gr.Audio(label="Listen (Hindi Only)", interactive=False)
             
-            # LOGIC
             translate_btn.click(
                 fn=process_translation, 
-                inputs=[input_text, input_audio, direction], 
-                outputs=[input_text, output_text, output_audio] # Updates input box with transcribed text too!
+                inputs=[input_text, input_audio, direction, is_hinglish], 
+                outputs=[input_text, output_text, output_audio]
             )
 
         with gr.TabItem("📝 Improve the AI"):