Update app.py

app.py

@@ -17,7 +17,8 @@ import traceback
 from streamlit_webrtc import webrtc_streamer, WebRtcMode
 from twilio.rest import Client  
 import logging
-from audiorecorder import audiorecorder
+import numpy as np
+import soundfile as sf
 
 
 # ✅ MUST be the first Streamlit command
@@ -182,24 +183,17 @@ def generate_improvement_suggestions():
     ---
     {cleaned_initial_evaluation}
     ---
-
     Based on all this information, your task is to provide DETAILED and CONSTRUCTIVE suggestions for each question to help the candidate improve. Be supportive and encouraging.
-
     For EACH question, please provide:
     1.  **How to Improve This Answer:** Specific, actionable advice on what the candidate could have added, clarified, or approached differently to make their answer better for their {level_string} level. Focus on 1-2 key improvement points.
     2.  **Hints for an Ideal Answer:** Briefly mention 2-3 key concepts, terms, or elements that a strong answer (appropriate for their {level_string} level) would typically include. DO NOT provide a full model answer, just hints and pointers.
-
     Keep the tone positive and focused on learning.
-
     Structure your response clearly for each question. Example for one question:
-
     ---
     **Regarding Question X: "[Original Question Text Here]"**
-
     *How to Improve This Answer:*
     [Your specific suggestion 1 for improvement...]
     [Your specific suggestion 2 for improvement...]
-
     *Hints for an Ideal Answer (Key Points to Consider):*
     - Hint 1 or Key concept 1
     - Hint 2 or Key concept 2
@@ -289,15 +283,11 @@ def evaluate_answers():
         hr_prompt_template = f"""
         You are an experienced HR interview evaluator assessing a candidate's soft skills based on their answers to interview questions.
         The candidate's performance across ALL answers should inform your scores for the following parameters.
-
         **Parameters to Score (Assign a score from 1 to 5 for each):**
         {hr_prompt_parameter_list}
-
         After providing a score (1-5) for each of the above parameters, also write an **Overall Qualitative Feedback** section. 
         This section should summarize the candidate's general soft skill strengths and areas for improvement, based on their communication, engagement, and professionalism throughout the interview.
-
         **REQUIRED OUTPUT FORMAT (Strictly Adhere):**
-
         **Parameter Scores (1-5):**
         Voice Modulation: [score]
         Confidence: [score]
@@ -308,7 +298,6 @@ def evaluate_answers():
         Basics of Grammar + SVA: [score]
         Persuasiveness: [score]
         Quality of Answers: [score]
-
         **Overall Qualitative Feedback:**
         [Your holistic qualitative feedback here. Be encouraging and constructive.]
         """
@@ -418,11 +407,9 @@ def evaluate_answers():
         {base_assessment_criteria_qualitative_non_hr}
         **YOUR RESPONSE MUST STRICTLY FOLLOW THIS FORMAT. PROVIDE SCORES FOR EACH QUESTION.**
         Output format:
-
         **Per-Question Scores:**
         Question 1 Score: [Score for Q1 out of 5]
         ... (repeat for all {num_answered_questions} questions provided)
-
         **Overall Evaluation Summary:**
         - Concept Understanding: [Overall qualitative feedback here]
         - Communication: [Overall qualitative feedback here]
@@ -502,7 +489,6 @@ def evaluate_answers():
 BEGINNER_PROMPT = """
 You are a friendly mock interview trainer conducting a **Beginner-level** spoken interview in the domain of **{domain}**.
 Ask basic verbal interview questions based on the candidate's input: **{input_text}**.
-
 Guidelines:
 - Ask simple conceptual questions.
 - Avoid jargon and complex examples.
@@ -511,7 +497,6 @@ Guidelines:
 Ensure the questions are clear, to the point, and suitable for a {difficulty_level}-level interview in {selected_domain}.
 **New Requirement:**
 🚫 **Do NOT repeat any questions from previous generations again and again.** Ensure all generated questions are unique and different from past sessions.
-
 **Guidelines:**
     ✅ Questions should focus on key concepts, best practices, and problem-solving within {selected_domain}.
     ✅ Ensure questions are direct, structured, and relevant to real-world applications.
@@ -522,7 +507,6 @@ Ensure the questions are clear, to the point, and suitable for a {difficulty_lev
 INTERMEDIATE_PROMPT = """
 You are a professional mock interviewer conducting an **Intermediate-level** spoken interview in the domain of **{domain}**.
 Ask moderately challenging verbal interview questions based on the candidate's input: **{input_text}**.
-
 Guidelines:
 - Use a mix of conceptual and real-world scenario questions.
 - Include light critical thinking.
@@ -530,7 +514,6 @@ Guidelines:
 Ensure the questions are clear, to the point, and suitable for a {difficulty_level}-level interview in {selected_domain}.
 **New Requirement:**
 🚫 **Do NOT repeat any questions from previous generations again and again.** Ensure all generated questions are unique and different from past sessions.
-
 **Guidelines:**
     ✅ Questions should focus on key concepts, best practices, and problem-solving within {selected_domain}.
     ✅ Ensure questions are direct, structured, and relevant to real-world applications.
@@ -541,15 +524,13 @@ Ensure the questions are clear, to the point, and suitable for a {difficulty_lev
 ADVANCED_PROMPT = """
 You are a strict mock interviewer conducting an **Advanced-level** spoken interview in the domain of **{domain}**.
 Ask deep, analytical, real-world scenario-based questions from the candidate's input: **{input_text}**.
-
 Guidelines:
 - Expect detailed, logical, well-structured answers.
-- Include challenging “why” and “how” based questions.
+- Include challenging "why" and "how" based questions.
 - No need for code, but assume candidate has high expertise.
 Ensure the questions are clear, to the point, and suitable for a {difficulty_level}-level interview in {selected_domain}.
 **New Requirement:**
 🚫 **Do NOT repeat any questions from previous generations again and again.** Ensure all generated questions are unique and different from past sessions.
-
 **Guidelines:**
     ✅ Questions should focus on key concepts, best practices, and problem-solving within {selected_domain}.
     ✅ Ensure questions are direct, structured, and relevant to real-world applications.
@@ -588,7 +569,6 @@ st.markdown("""
         box-shadow: 0 0 0 0.2rem rgba(0,123,255,.5) !important;
         outline: none !important;
     }
-
     .timer-text {
         font-size: 1.3rem;
         font-weight: 600;
@@ -600,7 +580,6 @@ st.markdown("""
         50% {opacity: 0.4;}
         100% {opacity: 1;}
     }
-
     .summary-card {
         background-color: #f9f9f9;
         padding: 20px;
@@ -795,8 +774,8 @@ def get_ice_servers():
         account_sid = os.environ["TWILIO_ACCOUNT_SID"]
         auth_token = os.environ["TWILIO_AUTH_TOKEN"]
     except KeyError:
-        logger.warning(
-            "Twilio credentials are not set. Fallback to a free STUN server from Google."  # noqa: E501
+        st.warning(
+            "Twilio credentials are not set. Fallback to a free STUN server from Google."
         )
         return [{"urls": ["stun:stun.l.google.com:19302"]}]
 
@@ -870,39 +849,101 @@ if st.session_state["generated_questions"]:
             if remaining > 0:
                 st.markdown(f"<h4 class='timer-text'>🎙️ {remaining} seconds to answer...</h4>", unsafe_allow_html=True)
                 
-                audio_bytes = audiorecorder("Click to start recording", "Click to stop")
-                if audio_bytes:
-                    wav_path = f"response_{idx}.wav"
-                    with open(wav_path, "wb") as f:
-                        f.write(audio_bytes)
-                    st.audio(wav_path, format="audio/wav")
-                        
-
-                    st.session_state.update({
-                        "record_phase": "listening",
-                        "response_file": wav_path,
-                         })
-                    st.rerun()
-                            
+                # Create two columns for recording options
+                col1, col2 = st.columns(2)
+                
+                with col1:
+                    st.markdown("**🎤 Voice Recording (WebRTC):**")
+                    # Try WebRTC recording
+                    try:
+                        webrtc_ctx = webrtc_streamer(
+                            key = f"webrtc_{idx}",
+                            mode=WebRtcMode.SENDONLY,
+                            audio_receiver_size=1024,
+                            media_stream_constraints={"audio": True, "video": False},
+                            rtc_configuration={"iceServers": get_ice_servers()},
+                        )
+
+                        if webrtc_ctx.state.playing:
+                            if st.button("⏹️ Stop Voice Recording", key=f"stop_voice_{idx}"):
+                                wav_path = f"response_{idx}.wav"
+                                try:
+                                    frames = webrtc_ctx.audio_receiver.get_frames(timeout=1)
+                                except Exception as e:
+                                    st.error(f"⚠️ Audio capture error: {e}")
+                                    frames = []
+                                    
+                                if frames:
+                                    try:
+                                        pcm = np.concatenate([f.to_ndarray() for f in frames], axis=0)
+                                        sample_rate = frames[0].sample_rate
+                                        sf.write(wav_path, pcm, sample_rate)
+                                        st.audio(wav_path)
+                                        st.session_state["answers"].append({
+                                            "question": question,
+                                            "response_file": wav_path
+                                        })
+                                    except Exception as e:
+                                        st.error(f"⚠��� Error saving recording: {e}")
+                                        st.session_state["answers"].append({
+                                            "question": question,
+                                            "response": "[Error saving recording]"
+                                        })
+                                else:
+                                    st.warning("⚠️ No audio captured.")
+                                    st.session_state["answers"].append({
+                                        "question": question,
+                                        "response": "[No response]"
+                                    })
+
+                                st.session_state.update({
+                                    "record_phase": "idle",
+                                    "question_played": False,
+                                    "current_question_index": idx + 1
+                                })
+
+                                if st.session_state["current_question_index"] == len(st.session_state["generated_questions"]):
+                                    evaluate_answers()
+                                    st.session_state["show_summary"] = True
+
+                                st.experimental_rerun()
+                    except Exception as webrtc_error:
+                        st.warning("⚠️ Voice recording not available")
+                        st.info("This is common on Hugging Face Spaces")
+                
+                with col2:
+                    st.markdown("**✍️ Text Input (Alternative):**")
+                    # Always show text input as alternative
+                    text_response = st.text_area(
+                        f"Type your answer:",
+                        key=f"text_response_{idx}",
+                        height=150,
+                        placeholder="Enter your response here..."
+                    )
+                    
+                    if st.button("✅ Submit Text Answer", key=f"submit_text_{idx}"):
+                        if text_response.strip():
+                            st.session_state["answers"].append({
+                                "question": question,
+                                "response": text_response.strip()
+                            })
+                        else:
+                            st.session_state["answers"].append({
+                                "question": question,
+                                "response": "[No text response provided]"
+                            })
                         
-                else:
-                    st.warning("⚠️ No audio captured.")
-                    st.session_state["answers"].append({
-                        "question": question,
-                        "response": "[No response]"
-                    })
-
-                    st.session_state.update({
-                        "record_phase": "idle",
-                        "question_played": False,
-                        "current_question_index": idx + 1
+                        st.session_state.update({
+                            "record_phase": "idle",
+                            "question_played": False,
+                            "current_question_index": idx + 1
                         })
 
-                    if st.session_state["current_question_index"] == len(st.session_state["generated_questions"]):
-                        evaluate_answers()
-                        st.session_state["show_summary"] = True
+                        if st.session_state["current_question_index"] == len(st.session_state["generated_questions"]):
+                            evaluate_answers()
+                            st.session_state["show_summary"] = True
 
-                    st.rerun()
+                        st.rerun()
                                 
             else:
                 st.markdown("<div style='padding:10px; background:#fff3e0; border-left:5px solid orange;'>⚠️ <strong>No response detected.</strong> Moving to next question...</div>", unsafe_allow_html=True)
@@ -920,23 +961,17 @@ if st.session_state["generated_questions"]:
                 st.rerun()
 
         elif st.session_state["record_phase"] == "listening":
-            st.success("🎧 Review your response below:")
-            st.audio(st.session_state["response_file"], format="audio/wav")
-            if st.button("⏹️ Confirm & Next"):
-                st.session_state["answers"].append({
-                "question": question,
-                "response_file": st.session_state["response_file"],
-                "response_text": st.session_state.get("recorded_text", "")
-                })
-
+            st.success("🎧 Listening... You may stop when ready.")
+            if st.button("⏹️ Stop Recording"):
+                st.session_state["answers"].append({"question": question, "response": st.session_state["recorded_text"]})
                 st.session_state.update({
                     "record_phase": "idle",
+                    "recorded_text": "",
                     "recording_started": False,
                     "question_played": False,
                     "question_start_time": 0.0,
                     "current_question_index": idx + 1
                 })
-                
                 if st.session_state["current_question_index"] == len(st.session_state["generated_questions"]):
                     evaluate_answers()
                     st.session_state["show_summary"] = True
@@ -1083,4 +1118,4 @@ if st.session_state.get("show_summary", False):
             keys_to_fully_clear = list(st.session_state.keys())
             for key_to_del_full in keys_to_fully_clear:
                 del st.session_state[key_to_del_full]
-           
+