Update app.py

app.py

@@ -19,7 +19,6 @@ from twilio.rest import Client
 import logging
 import whisper
 import speech_recognition as sr
-from gtts import gTTS
 #model = whisper.load_model("base")
 
 
@@ -87,33 +86,12 @@ def get_questions(prompt, input_text, num_questions=3, max_retries=10):
 
     return new_questions
 
-async def generate_question_audio(question, voice="en-US-AriaNeural"):
+async def generate_question_audio(question, voice="en-IE-EmilyNeural"):
     clean_question = re.sub(r'[^A-Za-z0-9.,?! ]+', '', question)
     tts = edge_tts.Communicate(text=clean_question, voice=voice)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as tmp_file:
-        try:
-            # 🔹 Try Edge-TTS first
-            tts = edge_tts.Communicate(text=clean_question, voice=voice)
-            await tts.save(tmp_file.name)
-            print("✅ Edge-TTS audio generated successfully.")
-            return tmp_file.name
-            
-        except Exception as e:
-            print(f"⚠️ Edge-TTS failed: {e}")
-            print("⏪ Falling back to Google TTS...")
-
-            try:
-                # 🔹 Fallback: Google TTS (works inside Hugging Face)
-                tts = gTTS(text=clean_question, lang="en")
-                tts.save(tmp_file.name)
-                print("✅ gTTS fallback audio generated successfully.")
-                return tmp_file.name
-
-            except Exception as e2:
-                print(f"❌ gTTS also failed: {e2}")
-                return None
-
-
+        await tts.save(tmp_file.name)
+        return tmp_file.name
     
 ########################################///////////////////////////////////////////////////#########################################
 
@@ -413,41 +391,91 @@ def evaluate_answers():
         base_assessment_criteria_qualitative_non_hr = """
         For the OVERALL qualitative summary, assess responses based on:
         - Conceptual Understanding (effort and relevance more than perfect accuracy for the level)
-        - Communication Clarity (can the core idea be understood?)
+        - Communication Clarity (is the idea understandable and logically stated?)
         - Depth of Explanation (relative to expected level)
-        - Use of Examples (if any, and if appropriate for the level)
-        - Logical Flow (is there a basic structure or train of thought?)
+        - Use of Examples (if any, and appropriate for the level)
+        - Logical Flow (basic structure or reasoning flow)
+
+        Focus on both understanding and reasoning. Responses should demonstrate thinking, not memorization.
         """
         per_question_scoring_guidelines_non_hr = f"""
-        For EACH question and its answer, provide a score from 0 to 5 points.
+        For EACH question and its answer, assign a score from 0 to 5 points.
         The candidate is at a {level_string} level.
-        Consider the following when assigning the per-question score:
-        - Effort: Did the candidate attempt a meaningful answer, even if partially incorrect?
-        - Relevance: Is the response at least partially related to the question topic?
-        - Clarity of thought for the candidate's level.
-        - Basic logical structure.
-        - Use of examples, if any were given and appropriate.
+        Use the numeric scale and notes below for calibration.
+
+        **Scoring Scale (per question):**
+        - **5 (Excellent / 90–100%)** → Comprehensive, accurate, and well-structured. Includes reasoning or an example. Rare and well-deserved.
+        - **4 (Good / 75–89%)** → Mostly correct, relevant, and clear. Minor conceptual gaps but good structure.
+        - **3 (Fair / 60–74%)** → Partially correct or lacks depth, but shows understanding and effort.
+        - **2 (Basic / 45–59%)** → One-line or short answer with minimal reasoning; incomplete or overly generic.
+        - **1 (Poor / 30–44%)** → Attempted but largely irrelevant or unclear.
+        - **0 (No Effort / <30%)** → Incorrect, off-topic, or explicitly “I don’t know”.
+
+        **Important Rules:**
+        - *One-word or one-line answers* (e.g., just definitions or keywords) must NOT score more than **2 out of 5**, regardless of correctness, because they lack reasoning and depth.
+        - Encourage clarity, structure, and explanation over memorized phrases.
+        """
+        scoring_tightness_guidelines = """
+        **Scoring Calibration (Strictness Guidance):**
+        - Maintain a slightly tight scoring approach.
+        - Incomplete or short one-line answers score **below 60% (1–2 out of 5)**.
+        - Scores of **5/5 (100%)** should be **rare** — reserved for comprehensive, insightful, and well-reasoned answers.
+        - Most competent answers should fall between **3 and 4**.
+        - When unsure, choose the **lower score** to maintain scoring consistency.
         """
         if level_string == "beginner":
             level_specific_instructions_non_hr = """
-            You are an **extremely understanding, encouraging, and supportive** interview evaluator for a **BEGINNER/FRESHER**. Your primary goal is to **build confidence**.
-            **Scoring Guidelines for Beginners (0-5 points per question):**
-            - **5 points:** Accurate, clear, and well-structured answer. Shows clear effort and basic understanding.
-            - **4 points:** Mostly correct with minor gaps or unclear phrasing.Good attempt, relevant, shows some understanding or key terms (e.g., one/two relevant words).
-            - **3 points:** Partially correct with evident effort, but lacks clarity or completeness.
-            - **1-2 points:** Minimal effort, mostly irrelevant, but an attempt beyond silence.
-            - **0 points:** Candidate explicitly says "I don’t know", "I'm not sure", or provides placeholder/non-answers. No relevant effort or understanding shown.Incorrect or unrelated.
-            Provide VERY positive feedback.
+            You are a **supportive, understanding evaluator** for a **BEGINNER/FRESHER**.
+            Focus on clarity, effort, and attempt — not perfection.
+            Encourage learning through feedback, but ensure fair scoring.
+
+            **Scoring Guidelines (0–5):**
+            - **5** → Accurate, clear, well-structured, and shows strong effort and reasoning. Rare.
+            - **4** → Mostly correct, relevant, and shows basic reasoning or understanding.
+            - **3** → Partial correctness with effort; may lack completeness or flow.
+            - **1–2** → One-line or definition-only answers; minimal reasoning. (Below 60%)
+            - **0** → No effort or irrelevant response.
+            Avoid giving high scores to short, memorized, or definition-only responses.
+            Provide motivating feedback that highlights areas of improvement.
             """
         elif level_string == "intermediate":
-            level_specific_instructions_non_hr = """Supportive evaluator for **INTERMEDIATE**. Scoring (0-5): 5=Correct/Clear; 3-4=Mostly correct; 1-2=Partial/Gaps; 0=Incorrect."""
+            level_specific_instructions_non_hr = """
+            You are a **balanced and fair evaluator** for an **INTERMEDIATE** candidate.
+            Expect conceptual clarity, structured reasoning, and relevant examples.
+            Be encouraging yet objective in scoring.
+
+            **Scoring Guidelines (0–5):**
+            - **5** → Clear, accurate, structured response with reasoning and relevance. Rare.
+            - **4** → Mostly correct with some logical structure and explanation.
+            - **3** → Some understanding; missing clarity or key detail.
+            - **1–2** → Short, definition-like, or minimal response. (Below 60%)
+            - **0** → Irrelevant or incorrect.
+            Never assign high scores to one-line or superficial answers.
+            """
         else: # Advanced
-            level_specific_instructions_non_hr = """Discerning evaluator for **ADVANCED**. Scoring (0-5): 5=Accurate/Comprehensive; 3-4=Correct lacks nuance; 1-2=Inaccurate; 0=Fundamentally incorrect."""
+            level_specific_instructions_non_hr = """
+            You are a **discerning but fair evaluator** for an **ADVANCED** professional.
+            Expect precision, applied understanding, and structured reasoning.
+            Maintain fairness without excessive strictness.
+
+            **Scoring Guidelines (0–5):**
+            - **5** → Exceptionally comprehensive, insightful, and accurate. (Rare)
+            - **4** → Correct and well-reasoned; may lack minor nuance or application.
+            - **3** → Adequate but missing depth, structure, or examples.
+            - **1–2** → Generic, incomplete, or one-line responses without reasoning. (Below 60%)
+            - **0** → Fundamentally incorrect or irrelevant.
+            Be concise and consistent in judgment; reward depth, not brevity.
+            """
         
         evaluation_prompt_template_non_hr = f"""
         {level_specific_instructions_non_hr} 
         {per_question_scoring_guidelines_non_hr}
         {base_assessment_criteria_qualitative_non_hr}
+        {scoring_tightness_guidelines}
+
+         When evaluating, be supportive yet fair. Encourage clarity and effort but avoid over-rewarding shallow or memorized answers.
+        Maintain a balanced tone — neither too strict nor too lenient.
+
         **YOUR RESPONSE MUST STRICTLY FOLLOW THIS FORMAT. PROVIDE SCORES FOR EACH QUESTION.**
         Output format:
 
@@ -461,7 +489,10 @@ def evaluate_answers():
         - Depth of Explanation: [Overall qualitative feedback here]
         - Examples: [Overall qualitative feedback here]
         - Logical Flow: [Overall qualitative feedback here]
+        - Final Remarks: [Brief encouraging but fair closing note]
         [Any additional overall encouraging remarks can optionally follow here]
+
+        Provide the final tone as **professional, balanced, and confidence-building**.
         """
         candidate_responses_formatted_non_hr = "\n\n".join(
             [f"Question {i+1}: {entry['question']}\nAnswer {i+1}: {str(entry.get('response', '[No response provided]'))}" for i, entry in enumerate(st.session_state["answers"])]
@@ -966,7 +997,7 @@ if st.session_state.get("generated_questions"):
 
         # Phase 2: Waiting to Start Recording
         elif st.session_state["record_phase"] == "waiting_to_start":
-            remaining = 20 - int(elapsed)
+            remaining = 15 - int(elapsed)
             if remaining > 0:
                 st.markdown(f"<h4 class='timer-text'>⏳ {remaining} seconds to click 'Start Recording'...</h4>", unsafe_allow_html=True)
                 if st.button("🎙️ Start Recording"):
@@ -1078,23 +1109,6 @@ if st.session_state.get("show_summary", False):
     current_percentage_score = st.session_state.get('percentage_score', 0.0)
     current_overall_score = st.session_state.get('overall_score', 0.0)
 
-    # --- Retrieve stored configuration info ---
-    selected_domain = st.session_state.get("selected_domain", "N/A")
-    input_type = st.session_state.get("section_choice", st.session_state.get("soft_skill_mode", "N/A"))
-    difficulty_level = st.session_state.get("difficulty_level_select", "N/A")
-    total_questions_selected = st.session_state.get("num_qs", num_qs_in_session)
-    selected_company = st.session_state.get("selected_company")
-    selected_job_role = st.session_state.get("selected_job_role")
-
-    # --- Display configuration summary ---
-    st.markdown("### ⚙️ Test Configuration Summary")
-    st.markdown(f"""
-    - **Domain Selected:** {selected_domain}  
-    - **Input Type / Mode:** {section_choice}  
-    - **Difficulty Level / Job Role:** {difficulty_level if selected_domain != "Finance" else job_roles}  
-    - **Total Questions Selected:** {total_questions_selected}
-    """)
-
     if st.session_state["selected_domain"] == "Soft Skills":
         hr_table_data = []
         for param, config in HR_PARAMETERS_CONFIG.items():
@@ -1181,19 +1195,10 @@ if st.session_state.get("show_summary", False):
 
     # Helper function to prepare summary text for download
     def prepare_summary_for_download():
-        #download_text = f"# GrillMaster Mock Interview Summary\n\n"
-        #download_text += f"**Selected Domain:** {st.session_state.get('selected_domain', 'N/A')}\n"
-        #dl_difficulty = st.session_state.get('difficulty_level_select', 'N/A')
-        #download_text += f"**Difficulty Level:** {dl_difficulty}\n"
         download_text = f"# GrillMaster Mock Interview Summary\n\n"
-        download_text += f"**Selected Domain:** {selected_domain}\n"
-        download_text += f"**Input Type** {section_choice}\n"
-        download_text += f"**Difficulty Level / Job Role:** {difficulty_level if selected_domain != 'Finance' else job_roles}\n"
-        download_text += f"**Total Questions Selected:** {total_questions_selected}\n"
-        download_text += f"**Company Selected:** {selected_company}\n"
-        download_text += f"**Job Role:** {selected_job_role}\n"
-        #download_text += f"**Calculated Overall Score:** {current_overall_score:.1f} / {max_score_possible_for_session:.1f} ({current_percentage_score:.1f}%)\n\n"
-        download_text += "## Questions & Candidate's Answers:\n"
+        download_text += f"**Selected Domain:** {st.session_state.get('selected_domain', 'N/A')}\n"
+        dl_difficulty = st.session_state.get('difficulty_level_select', 'N/A')
+        download_text += f"**Difficulty Level:** {dl_difficulty}\n"
         
         num_q_for_max_score = len(st.session_state.get("generated_questions", st.session_state.get("answers",[])))
         max_s_for_dl = num_q_for_max_score * 5.0