Completely update app.py

app.py

@@ -1,109 +1,98 @@
 import gradio as gr
+import openai
 import os
 import uuid
-from faster_whisper import WhisperModel
 from gtts import gTTS
+from faster_whisper import WhisperModel
 import subprocess
-import openai
-from openai import OpenAI
 import shutil
 
-# Use writable cache for Hugging Face
-os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"
+# === Clean /tmp at startup ===
+TMP_DIR = "/tmp"
+for sub in os.listdir(TMP_DIR):
+    sub_path = os.path.join(TMP_DIR, sub)
+    try:
+        if os.path.isfile(sub_path) or os.path.islink(sub_path):
+            os.unlink(sub_path)
+        elif os.path.isdir(sub_path):
+            shutil.rmtree(sub_path)
+    except Exception as e:
+        print(f"Failed to delete {sub_path}. Reason: {e}")
+
+# === Environment Setup ===
 os.environ["HF_HOME"] = "/tmp/hf"
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf"
 os.environ["XDG_CACHE_HOME"] = "/tmp/hf"
 
-chatter_owl_url = "file/images/chatter_owl.png"
-
-# Load Whisper and set OpenAI key
-model = WhisperModel("base", compute_type="int8")  # Fastest CPU option
+# Use writable cache for Hugging Face
+os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"
 openai.api_key = os.getenv("OPENAI_API_KEY")
 
-# Prompt template
-PROMPT_TEMPLATE = """
-You are a communication coach evaluating a user's spoken response.
-
-Evaluate the speech using these 5 dimensions:
-1. Clarity
-2. Structure
-3. Fluency
-4. Content Relevance
-5. Tone & Expression
-
-For each:
-- Score from 0–10
-- Provide brief, specific feedback
-
-Then:
-- Give an overall 3–4 line summary
-- End with a motivational sentence
-
-Lastly:
-- Provide an improved version of the speech that better demonstrates effective communication, based on their original transcript.
-
-Transcript:
-\"\"\"{transcript}\"\"\"
-"""
-
-# Convert uploaded audio to 16kHz mono WAV
+# === Language Codes for GPT + TTS ===
+LANG_CODES = {
+    "English": "en",
+    "Spanish": "es",
+    "Hindi": "hi",
+    "French": "fr",
+    "German": "de",
+    "Arabic": "ar",
+    "Chinese": "zh",
+    "Portuguese": "pt",
+    "Japanese": "ja",
+    "Korean": "ko"
+}
+
+# === Load Whisper Model ===
+model = WhisperModel("base", compute_type="int8")
+
+# === Audio Processing ===
 def convert_to_wav(input_file):
     output_wav = f"/tmp/{uuid.uuid4()}.wav"
     command = ["ffmpeg", "-y", "-i", input_file, "-ar", "16000", "-ac", "1", output_wav]
     subprocess.run(command, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
     return output_wav
 
-# Call GPT to analyze transcript
-client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
-
-def generate_feedback_with_llm(transcript):
-    prompt = PROMPT_TEMPLATE.format(transcript=transcript.strip())
-
-    response = client.chat.completions.create(
-    model="gpt-4",
-    messages=[
-        {"role": "system", "content": "You are a communication coach that gives structured and emotionally aware feedback."},
-        {"role": "user", "content": prompt}
-    ],
-    temperature=0.7
-    )
-
-    return response.choices[0].message.content
-
 def transcribe_audio(audio_path):
     segments, info = model.transcribe(audio_path)
-    transcript = " ".join([segment.text for segment in segments])
-    return transcript
-    
-# Main function
-def tutor_feedback(audio_file):
-    if not audio_file or not os.path.exists(audio_file):
-        return "No audio received.", "Please upload or record again.", None
+    return " ".join([segment.text for segment in segments])
 
-    print(f">>> Audio received: {audio_file}")
-    print(f">>> File size: {os.path.getsize(audio_file)} bytes")
-
-    # Convert to whisper-friendly WAV
-    wav_path = convert_to_wav(audio_file)
-
-    # Transcribe
-    transcript = transcribe_audio(wav_path)
+# === GPT-4 Feedback Generation ===
+def generate_feedback(transcript, language):
+    prompt = f"""
+You are a communication coach. Please respond in [language={language}].
 
+Evaluate the user's speech on:
+1. Clarity
+2. Structure
+3. Fluency
+4. Content Relevance
+5. Tone & Expression
 
-    # Feedback via GPT
-    feedback_text = generate_feedback_with_llm(transcript)
+Each category:
+- Score out of 10
+- Short explanation
 
-    # TTS
-    tts = gTTS(feedback_text)
-    mp3_path = f"/tmp/{uuid.uuid4()}.mp3"
-    tts.save(mp3_path)
+End with:
+- Overall feedback summary
+- One motivational line
 
-    return transcript, feedback_text, mp3_path
+Transcript:
+{transcript}
+"""
+    response = openai.ChatCompletion.create(
+        model="gpt-4",
+        messages=[
+            {"role": "system", "content": f"You are a supportive communication coach responding in {language}."},
+            {"role": "user", "content": prompt}
+        ],
+        temperature=0.7
+    )
+    return response["choices"][0]["message"]["content"]
 
-# Generate example version of the same speech
-def generate_example_response(transcript):
+# === GPT-4 Suggested Example ===
+def generate_example_response(transcript, language):
     prompt = f"""
-You are a communication tutor. Rewrite this speech transcript to be a more polished and confident version, while keeping the meaning and tone similar.
+You are a communication tutor. Rewrite this speech transcript to be a more polished and confident version, while keeping the meaning and tone similar. Respond in {language}.
 
 Transcript:
 {transcript}
@@ -118,30 +107,40 @@ Transcript:
     )
     return response["choices"][0]["message"]["content"]
 
-# Hook up logic
-def show_example_feedback(transcript):
-    # Extract only example portion from GPT or return a pre-generated version
-    return generate_example_response(transcript)
+# === Main Feedback Function ===
+def tutor_feedback(audio_file, language):
+    if not audio_file or not os.path.exists(audio_file):
+        return "", "No audio received.", None, ""
 
-# Gradio interface
-# Path to your CSS file
-css_path = "light_mode_chatter_owl.css"  # Make sure this file is in the same directory as app.py
+    wav_path = convert_to_wav(audio_file)
+    transcript = transcribe_audio(wav_path)
+    feedback_text = generate_feedback(transcript, language)
 
-# Path to your CSS file
-css_path = "light_mode_chatter_owl.css"  # Make sure this file is in the same directory as app.py
+    lang_code = LANG_CODES.get(language, "en")
+    tts = gTTS(feedback_text, lang=lang_code)
+    mp3_path = f"/tmp/{uuid.uuid4()}.mp3"
+    tts.save(mp3_path)
 
-# App UI
+    return transcript, feedback_text, mp3_path, transcript
+
+# === Gradio Interface ===
 with gr.Blocks(css="light_mode_chatter_owl.css") as app:
     gr.Markdown(
         """
         <div id="header" style="text-align: center;">
             <img src="file/images/chatter_owl.png" width="120">
             <h2>🦉 Meet <strong>Chatter the Owl</strong></h2>
-            <p>Speak into the mic, and I’ll give you structured feedback to help you grow as a communicator!</p>
+            <p>Choose your language, speak into the mic, and I’ll give you structured feedback to help you grow as a communicator!</p>
         </div>
         """
     )
 
+    language_dropdown = gr.Dropdown(
+        label="🌍 Select Your Language",
+        choices=list(LANG_CODES.keys()),
+        value="English"
+    )
+
     with gr.Row():
         audio_input = gr.Audio(type="filepath", label="🎙 Speak or Upload Audio")
 
@@ -156,16 +155,26 @@ with gr.Blocks(css="light_mode_chatter_owl.css") as app:
 
     example_box = gr.Textbox(label="🗣 Suggested Improvement", visible=False)
 
-    # Interactions
-    audio_input.change(fn=tutor_feedback, inputs=audio_input,
-                       outputs=[transcript_box, feedback_box, audio_output, hidden_transcript])
-
-    try_again.click(fn=lambda: ("", "", None, "", ""), inputs=None,
-                    outputs=[transcript_box, feedback_box, audio_output, hidden_transcript, example_box])
+    # Connect functions
+    audio_input.change(
+        fn=tutor_feedback,
+        inputs=[audio_input, language_dropdown],
+        outputs=[transcript_box, feedback_box, audio_output, hidden_transcript]
+    )
 
-    show_example.click(fn=generate_example_response, inputs=hidden_transcript, outputs=example_box)
+    try_again.click(
+        fn=lambda: ("", "", None, "", ""),
+        inputs=None,
+        outputs=[transcript_box, feedback_box, audio_output, hidden_transcript, example_box]
+    )
 
+    show_example.click(
+        fn=generate_example_response,
+        inputs=[hidden_transcript, language_dropdown],
+        outputs=example_box
+    )
 
+# === Launch App ===
 if __name__ == "__main__":
     print("✅ App is launching...")
     app.launch(server_name="0.0.0.0", server_port=7860, debug=True)