Update app.py

app.py

@@ -4,8 +4,7 @@ import os
 import random
 import whisper
 
-# ========== Gemini API Setup ==========
-
+# ========== Gemini API Setup (fallback with 2 keys) ==========
 API_KEYS = [os.getenv("API_KEY_1"), os.getenv("API_KEY_2")]
 model = None
 
@@ -14,7 +13,7 @@ def initialize_model():
     for key in API_KEYS:
         try:
             genai.configure(api_key=key)
-            model = genai.GenerativeModel("gemini-1.5-flash")
+            model = genai.GenerativeModel("gemini-2.0-flash")
             test = model.generate_content("Hello!")
             if "Hello" in test.text:
                 print(f"[INFO] Using Gemini API key ending with: ...{key[-4:]}")
@@ -25,19 +24,10 @@ def initialize_model():
 
 initialize_model()
 
-# ========== Whisper ASR ==========
+# ========== Load Whisper ==========
 whisper_model = whisper.load_model("base")
 
-# ========== Session State ==========
-session = {
-    "questions": [],
-    "current_index": 0,
-    "scores": [],
-    "feedbacks": [],
-    "transcripts": []
-}
-
-# ========== Interview Questions ==========
+# ========== Questions ==========
 all_questions = {
     "general": [
     "Tell me about yourself.",
@@ -338,7 +328,27 @@ all_questions = {
     ]
 }
 
-# ========== Evaluation ==========
+# ========== Session State ==========
+session = {
+    "questions": [],
+    "current_index": 0,
+    "scores": [],
+    "feedbacks": [],
+    "transcripts": []
+}
+
+# ========== Handle Audio ==========
+def handle_audio(audio):
+    if not audio:
+        return "No audio provided"
+    result = whisper_model.transcribe(audio)
+    return result["text"]
+
+# Auto-transcription for transcript box
+def transcribe_and_show(audio):
+    return handle_audio(audio)
+
+# ========== Evaluate Answer ==========
 def get_score(question, answer):
     prompt = f"""
 You are an interview expert. A user was asked:
@@ -364,21 +374,15 @@ Format:
             feedback = line.split("**Feedback:**")[1].strip()
     return score, feedback
 
+# ========== Summary ==========
 def summarize_feedback(feedbacks):
     prompt = "Summarize the following feedback into 4-5 improvement tips:\n" + "\n".join(feedbacks)
     return model.generate_content(prompt).text.strip()
 
-# ========== Audio Processing ==========
-def handle_audio(audio):
-    if not audio:
-        return "No audio provided"
-    result = whisper_model.transcribe(audio)
-    return result["text"]
-
-# ========== Interview Logic ==========
+# ========== Submit Answer ==========
 def run_session(audio):
     if session["current_index"] >= len(session["questions"]):
-        return "✅ Interview complete!", gr.update(visible=False)
+        return "✅ Interview complete!", "", gr.update(visible=False)
 
     question = session["questions"][session["current_index"]]
     transcript = handle_audio(audio)
@@ -391,12 +395,21 @@ def run_session(audio):
 
     if session["current_index"] < len(session["questions"]):
         next_q = session["questions"][session["current_index"]]
-        return f"🗣️ Answer: {transcript}\n\n🔢 Score: {score}/10\n💬 Feedback: {feedback}\n\n➡️ Next: {next_q}", gr.update(visible=True)
+        return (
+            f"🗣️ Answer: {transcript}\n\n🔢 Score: {score}/10\n💬 Feedback: {feedback}\n\n➡️ Next: {next_q}",
+            "",
+            gr.update(visible=True),
+        )
     else:
         avg_score = sum(session["scores"]) / len(session["scores"])
         summary = summarize_feedback(session["feedbacks"])
-        return f"✅ Interview Complete!\n\n📊 Avg Score: {avg_score:.2f}/10\n\n📝 Summary:\n{summary}", gr.update(visible=False)
+        return (
+            f"✅ Interview Complete!\n\n📊 Avg Score: {avg_score:.2f}/10\n\n📝 Summary:\n{summary}",
+            "",
+            gr.update(visible=False),
+        )
 
+# ========== Start New Session ==========
 def start_session(field):
     general_qs = random.sample(all_questions["general"], 10)
     domain_qs = random.sample(all_questions[field], 10)
@@ -410,18 +423,23 @@ def start_session(field):
 
 # ========== Gradio UI ==========
 with gr.Blocks() as demo:
-    gr.Markdown("## 🎤 AI Mock Interview — Voice Only")
-
+    gr.Markdown("## 🎤 AI Mock Interview (Voice-Based)")
     with gr.Row():
-        field = gr.Dropdown(label="Select Domain", choices=list(all_questions.keys()), value="ml")
+        field = gr.Dropdown(
+            label="Select your domain",
+            choices=list(all_questions.keys()),
+            value="ml"
+        )
         start_btn = gr.Button("Start Interview")
 
     question_display = gr.Textbox(label="Current Question", interactive=False)
     audio_input = gr.Audio(type="filepath", label="🎙️ Record Your Answer")
+    transcript_display = gr.Textbox(label="📝 Transcript", interactive=False)
     submit_btn = gr.Button("Submit Answer", visible=False)
-    output = gr.Textbox(label="Feedback", lines=8)
+    output = gr.Textbox(label="Feedback", lines=10)
 
     start_btn.click(start_session, inputs=field, outputs=[question_display, submit_btn])
-    submit_btn.click(run_session, inputs=audio_input, outputs=[output, submit_btn])
+    audio_input.change(transcribe_and_show, inputs=audio_input, outputs=transcript_display)
+    submit_btn.click(run_session, inputs=audio_input, outputs=[output, transcript_display, submit_btn])
 
 demo.launch()