Update process_interview.py

process_interview.py

@@ -26,7 +26,6 @@ from reportlab.lib import colors
 import matplotlib.pyplot as plt
 import matplotlib
 matplotlib.use('Agg')
-from reportlab.platypus import Image
 import io
 from transformers import AutoTokenizer, AutoModel
 import spacy
@@ -37,18 +36,17 @@ from concurrent.futures import ThreadPoolExecutor
 # Setup logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
-logging.getLogger("nemo_logging").setLevel(logging.INFO)
-logging.getLogger("nemo").setLevel(logging.INFO)
+logging.getLogger("nemo_logger").setLevel(logging.WARNING)
 
 # Configuration
-AUDIO_DIR = "./uploads"
+AUDIO_DIR = "./Uploads"
 OUTPUT_DIR = "./processed_audio"
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 
-# API Keys
-PINECONE_KEY = os.getenv("PINECONE_KEY")
-ASSEMBLYAI_KEY = os.getenv("ASSEMBLYAI_KEY")
-GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
+# API Keys (replace with actual keys or environment variables)
+PINECONE_KEY = os.getenv("PINECONE_KEY", "your-pinecone-key")
+ASSEMBLYAI_KEY = os.getenv("ASSEMBLYAI_KEY", "your-assemblyai-key")
+GEMINI_API_KEY = os.getenv("GEMINI_API_KEY", "your-gemini-key")
 
 def download_audio_from_url(url: str) -> str:
     """Downloads an audio file from a URL to a temporary local path."""
@@ -92,11 +90,10 @@ logger.info(f"Using device: {device}")
 
 def load_speaker_model():
     try:
-        import torch
         torch.set_num_threads(5)
         model = EncDecSpeakerLabelModel.from_pretrained(
             "nvidia/speakerverification_en_titanet_large",
-            map_location=torch.device('cpu')
+            map_location=device
         )
         model.eval()
         return model
@@ -190,7 +187,7 @@ def transcribe(audio_path: str) -> Dict:
         logger.error(f"Transcription failed: {str(e)}")
         raise
 
-def process_utterance(utterance, full_audio, wav_file):
+def process_utterance(utterance: Dict, full_audio: AudioSegment, wav_file: str) -> Dict:
     try:
         start = utterance['start']
         end = utterance['end']
@@ -220,7 +217,7 @@ def process_utterance(utterance, full_audio, wav_file):
             'embedding': embedding_list
         }
     except Exception as e:
-        logger.error(f"Utterance processing failed: {str(e)}", exc_info=True)
+        logger.error(f"Utterance processing failed: {str(e)}")
         return {
             **utterance,
             'speaker': 'Unknown',
@@ -267,7 +264,7 @@ def train_role_classifier(utterances: List[Dict]):
                 sum(1 for token in doc if token.pos_ == 'NOUN')
             ])
             features.append(feat)
-            labels.append(0 if i % 2 == 0 else 1)
+            labels.append(0 if i % 2 == 0 else 1)  # Simplified for demo; replace with actual labels
         scaler = StandardScaler()
         X = scaler.fit_transform(features)
         clf = RandomForestClassifier(
@@ -371,46 +368,44 @@ def generate_voice_interpretation(analysis: Dict) -> str:
     if 'error' in analysis:
         return "Voice analysis unavailable due to processing limitations."
     interpretation_lines = [
-        "Vocal Performance Profile:",
-        f"- Speaking Rate: {analysis['speaking_rate']} words/sec - Benchmark: 2.0-3.0 wps for clear delivery",
-        f"- Filler Word Frequency: {analysis['filler_ratio'] * 100:.1f}% - Measures non-content words",
-        f"- Repetition Index: {analysis['repetition_score']:.3f} - Frequency of repeated phrases",
-        f"- Anxiety Indicator: {analysis['interpretation']['anxiety_level']} (Score: {analysis['composite_scores']['anxiety']:.3f}) - Pitch and vocal stability",
-        f"- Confidence Indicator: {analysis['interpretation']['confidence_level']} (Score: {analysis['composite_scores']['confidence']:.3f}) - Vocal strength",
-        f"- Fluency Rating: {analysis['interpretation']['fluency_level']} - Speech flow and coherence",
+        f"- Speaking Rate: {analysis['speaking_rate']} words/sec (Benchmark: 2.0-3.0 wps; affects clarity)",
+        f"- Filler Words: {analysis['filler_ratio'] * 100:.1f}% (High usage reduces credibility)",
+        f"- Anxiety: {analysis['interpretation']['anxiety_level']} (Score: {analysis['composite_scores']['anxiety']:.3f}; stress response)",
+        f"- Confidence: {analysis['interpretation']['confidence_level']} (Score: {analysis['composite_scores']['confidence']:.3f}; vocal strength)",
+        f"- Fluency: {analysis['interpretation']['fluency_level']} (Drives engagement)",
         "",
         "HR Insights:",
-        "- Rapid speech (>3.0 wps) may signal enthusiasm but risks clarity.",
-        "- High filler word use reduces perceived professionalism.",
-        "- Elevated anxiety suggests pressure; training can build resilience.",
-        "- Strong confidence aligns with leadership presence.",
-        "- Fluent speech enhances engagement, critical for team roles."
+        "- Rapid speech (>3.0 wps) may reduce clarity; slower pacing enhances professionalism.",
+        "- High filler word usage undermines perceived confidence.",
+        "- Elevated anxiety suggests pressure; training can improve resilience.",
+        "- Strong confidence supports leadership presence.",
+        "- Fluent speech enhances engagement in team settings."
     ]
     return "\n".join(interpretation_lines)
 
-def generate_anxiety_confidence_chart(composite_scores: Dict, chart_path_or_buffer):
+def generate_anxiety_confidence_chart(composite_scores: Dict, chart_buffer):
     try:
         labels = ['Anxiety', 'Confidence']
         scores = [composite_scores.get('anxiety', 0), composite_scores.get('confidence', 0)]
         fig, ax = plt.subplots(figsize=(5, 3.5))
         bars = ax.bar(labels, scores, color=['#FF5252', '#26A69A'], edgecolor='black', width=0.45)
-        ax.set_ylabel('Score (Normalized)', fontsize=12)
+        ax.set_ylabel('Score', fontsize=12)
         ax.set_title('Vocal Dynamics: Anxiety vs. Confidence', fontsize=14, pad=15)
         ax.set_ylim(0, 1.3)
         for bar in bars:
             height = bar.get_height()
             ax.text(bar.get_x() + bar.get_width()/2, height + 0.05, f"{height:.2f}",
-                    ha='center', color='black', fontweight='bold', fontsize=11)
+                    ha='center', color='black', fontweight='bold', fontsize=10)
         ax.grid(True, axis='y', linestyle='--', alpha=0.7)
         plt.tight_layout()
-        plt.savefig(chart_path_or_buffer, format='png', bbox_inches='tight', dpi=300)
+        plt.savefig(chart_buffer, format='png', bbox_inches='tight', dpi=300)
         plt.close(fig)
     except Exception as e:
         logger.error(f"Error generating chart: {str(e)}")
 
 def calculate_acceptance_probability(analysis_data: Dict) -> float:
     voice = analysis_data.get('voice_analysis', {})
-    if 'error' in voice: return 0.0
+    if 'error' in voice: return 50.0
     w_confidence, w_anxiety, w_fluency, w_speaking_rate, w_filler_repetition, w_content_strengths = 0.35, -0.25, 0.2, 0.15, -0.15, 0.25
     confidence_score = voice.get('composite_scores', {}).get('confidence', 0.0)
     anxiety_score = voice.get('composite_scores', {}).get('anxiety', 0.0)
@@ -428,8 +423,7 @@ def calculate_acceptance_probability(analysis_data: Dict) -> float:
     content_strength_val = 0.85 if analysis_data.get('text_analysis', {}).get('total_duration', 0) > 60 else 0.4
     raw_score = (confidence_score * w_confidence + (1 - anxiety_score) * abs(w_anxiety) + fluency_val * w_fluency + speaking_rate_score * w_speaking_rate + filler_repetition_score * abs(w_filler_repetition) + content_strength_val * w_content_strengths)
     max_possible_score = (w_confidence + abs(w_anxiety) + w_fluency + w_speaking_rate + abs(w_filler_repetition) + w_content_strengths)
-    if max_possible_score == 0: return 50.0
-    normalized_score = raw_score / max_possible_score
+    normalized_score = raw_score / max_possible_score if max_possible_score > 0 else 0.5
     acceptance_probability = max(0.0, min(1.0, normalized_score))
     return float(f"{acceptance_probability * 100:.2f}")
 
@@ -437,39 +431,39 @@ def generate_report(analysis_data: Dict) -> str:
     try:
         voice = analysis_data.get('voice_analysis', {})
         voice_interpretation = generate_voice_interpretation(voice)
-        interviewee_responses = [f"Speaker {u['speaker']} ({u['role']}): {u['text']}" for u in analysis_data['transcript'] if u['role'] == 'Interviewee'][:6]
-        acceptance_prob = analysis_data.get('acceptance_probability', None)
-        acceptance_line = ""
-        if acceptance_prob is not None:
-            acceptance_line = f"\n**Hiring Suitability Score: {acceptance_prob:.2f}%**\n"
-            if acceptance_prob >= 80: acceptance_line += "HR Verdict: Outstanding candidate, highly recommended for immediate advancement."
-            elif acceptance_prob >= 60: acceptance_line += "HR Verdict: Strong candidate, suitable for further evaluation with targeted development."
-            elif acceptance_prob >= 40: acceptance_line += "HR Verdict: Moderate potential, requires additional assessment and skill-building."
-            else: acceptance_line += "HR Verdict: Limited fit, significant improvement needed for role alignment."
+        interviewee_responses = [f"- {u['text']}" for u in analysis_data['transcript'] if u['role'] == 'Interviewee'][:5]
+        acceptance_prob = analysis_data.get('acceptance_probability', 50.0)
+        acceptance_line = f"\n**Suitability Score: {acceptance_prob:.2f}%**\n"
+        if acceptance_prob >= 80:
+            acceptance_line += "HR Verdict: Outstanding candidate, recommended for immediate advancement."
+        elif acceptance_prob >= 60:
+            acceptance_line += "HR Verdict: Strong candidate, suitable for further evaluation."
+        elif acceptance_prob >= 40:
+            acceptance_line += "HR Verdict: Moderate potential, needs additional assessment."
+        else:
+            acceptance_line += "HR Verdict: Limited fit, significant improvement required."
         prompt = f"""
-        You are EvalBot, a senior HR consultant with 20+ years of experience, delivering a polished, concise, and engaging interview analysis report. Use a professional tone, clear headings, and bullet points ('- ') for readability. Avoid redundancy and ensure distinct sections for strengths, growth areas, and recommendations.
+        You are EvalBot, a senior HR consultant delivering a concise, professional interview analysis report. Use clear headings, bullet points ('-'), and avoid redundancy. Focus on candidate suitability, strengths, and actionable recommendations.
         {acceptance_line}
         **1. Executive Summary**
-        - Provide a concise overview of performance, key metrics, and hiring potential.
-        - Interview length: {analysis_data['text_analysis']['total_duration']:.2f} seconds
-        - Speaker turns: {analysis_data['text_analysis']['speaker_turns']}
-        - Participants: {', '.join(analysis_data['speakers'])}
+        - Summarize performance, key metrics, and hiring potential.
+        - Duration: {analysis_data['text_analysis']['total_duration']:.2f} seconds
+        - Speaker Turns: {analysis_data['text_analysis']['speaker_turns']}
+        - Participants: {', '.join(sorted(set(u['speaker'] for u in analysis_data['transcript'])))}
         **2. Communication and Vocal Dynamics**
-        - Evaluate vocal delivery (rate, fluency, confidence) and professional impact.
-        - Offer HR insights on workplace alignment.
+        - Evaluate vocal delivery (rate, fluency, confidence).
+        - Provide HR insights on workplace alignment.
         {voice_interpretation}
-        **3. Competency and Content Evaluation**
-        - Assess competencies: leadership, problem-solving, communication, adaptability.
-        - List strengths and growth areas separately, with specific examples.
+        **3. Competency and Content**
+        - Assess leadership, problem-solving, communication, adaptability.
+        - List strengths and growth areas separately with examples.
         - Sample responses:
         {chr(10).join(interviewee_responses)}
-        **4. Role Fit and Growth Potential**
-        - Analyze cultural fit, role readiness, and long-term potential.
-        - Highlight enthusiasm and scalability.
-        **5. Strategic HR Recommendations**
-        - Provide distinct, prioritized strategies for candidate growth.
-        - Target: Communication, Response Depth, Professional Presence.
-        - List clear next steps for hiring managers (e.g., advance, train, assess).
+        **4. Role Fit and Potential**
+        - Analyze cultural fit, role readiness, and growth potential.
+        **5. Recommendations**
+        - Provide prioritized strategies for growth (communication, technical skills, presence).
+        - Suggest next steps for hiring managers (advance, train, assess).
         """
         response = gemini_model.generate_content(prompt)
         return response.text
@@ -480,14 +474,14 @@ def generate_report(analysis_data: Dict) -> str:
 def create_pdf_report(analysis_data: Dict, output_path: str, gemini_report_text: str):
     try:
         doc = SimpleDocTemplate(output_path, pagesize=letter,
-                                rightMargin=0.7*inch, leftMargin=0.7*inch,
-                                topMargin=0.9*inch, bottomMargin=0.9*inch)
+                                rightMargin=0.75*inch, leftMargin=0.75*inch,
+                                topMargin=1*inch, bottomMargin=1*inch)
         styles = getSampleStyleSheet()
-        h1 = ParagraphStyle(name='Heading1', fontSize=22, leading=26, spaceAfter=20, alignment=1, textColor=colors.HexColor('#003087'), fontName='Helvetica-Bold')
-        h2 = ParagraphStyle(name='Heading2', fontSize=15, leading=18, spaceBefore=14, spaceAfter=8, textColor=colors.HexColor('#0050BC'), fontName='Helvetica-Bold')
-        h3 = ParagraphStyle(name='Heading3', fontSize=11, leading=14, spaceBefore=10, spaceAfter=6, textColor=colors.HexColor('#3F7CFF'), fontName='Helvetica')
-        body_text = ParagraphStyle(name='BodyText', fontSize=10, leading=13, spaceAfter=8, fontName='Helvetica', textColor=colors.HexColor('#333333'))
-        bullet_style = ParagraphStyle(name='Bullet', parent=body_text, leftIndent=20, bulletIndent=10, fontName='Helvetica', bulletFontName='Helvetica', bulletFontSize=10)
+        h1 = ParagraphStyle(name='Heading1', fontSize=20, leading=24, spaceAfter=18, alignment=1, textColor=colors.HexColor('#003087'), fontName='Helvetica-Bold')
+        h2 = ParagraphStyle(name='Heading2', fontSize=14, leading=16, spaceBefore=12, spaceAfter=8, textColor=colors.HexColor('#0050BC'), fontName='Helvetica-Bold')
+        h3 = ParagraphStyle(name='Heading3', fontSize=10, leading=12, spaceBefore=8, spaceAfter=6, textColor=colors.HexColor('#3F7CFF'), fontName='Helvetica')
+        body_text = ParagraphStyle(name='BodyText', fontSize=9, leading=12, spaceAfter=6, fontName='Helvetica', textColor=colors.HexColor('#333333'))
+        bullet_style = ParagraphStyle(name='Bullet', parent=body_text, leftIndent=18, bulletIndent=8, fontName='Helvetica', bulletFontName='Helvetica', bulletFontSize=9)
         
         story = []
 
@@ -495,55 +489,54 @@ def create_pdf_report(analysis_data: Dict, output_path: str, gemini_report_text:
             canvas.saveState()
             canvas.setFont('Helvetica', 8)
             canvas.setFillColor(colors.HexColor('#666666'))
-            canvas.drawString(doc.leftMargin, 0.4 * inch, f"Page {doc.page} | EvalBot HR Interview Report | Confidential")
+            canvas.drawString(doc.leftMargin, 0.5*inch, f"Page {doc.page} | EvalBot HR Interview Report | Confidential")
             canvas.setStrokeColor(colors.HexColor('#0050BC'))
-            canvas.setLineWidth(1)
-            canvas.line(doc.leftMargin, doc.height + 0.85*inch, doc.width + doc.leftMargin, doc.height + 0.85*inch)
-            canvas.setFont('Helvetica-Bold', 10)
-            canvas.drawString(doc.leftMargin, doc.height + 0.9*inch, "Candidate Interview Analysis")
-            canvas.drawRightString(doc.width + doc.leftMargin, doc.height + 0.9*inch, time.strftime('%B %d, %Y'))
+            canvas.setLineWidth(0.8)
+            canvas.line(doc.leftMargin, doc.height + 0.9*inch, doc.width + doc.leftMargin, doc.height + 0.9*inch)
+            canvas.setFont('Helvetica-Bold', 9)
+            canvas.drawString(doc.leftMargin, doc.height + 0.95*inch, "Candidate Interview Analysis")
+            canvas.drawRightString(doc.width + doc.leftMargin, doc.height + 0.95*inch, time.strftime('%B %d, %Y'))
             canvas.restoreState()
 
         # Title Page
         story.append(Paragraph("Candidate Interview Analysis", h1))
-        story.append(Paragraph(f"Generated: {time.strftime('%B %d, %Y')}", ParagraphStyle(name='Date', alignment=1, fontSize=10, textColor=colors.HexColor('#666666'), fontName='Helvetica')))
-        story.append(Spacer(1, 0.5 * inch))
-        acceptance_prob = analysis_data.get('acceptance_probability')
-        if acceptance_prob is not None:
-            story.append(Paragraph("Hiring Suitability Snapshot", h2))
-            prob_color = colors.HexColor('#2E7D32') if acceptance_prob >= 80 else (colors.HexColor('#F57C00') if acceptance_prob >= 60 else colors.HexColor('#D32F2F'))
-            story.append(Paragraph(f"Suitability Score: <font size=16 color='{prob_color.hexval()}'><b>{acceptance_prob:.2f}%</b></font>",
-                                 ParagraphStyle(name='Prob', fontSize=12, spaceAfter=12, alignment=1, fontName='Helvetica-Bold')))
-            if acceptance_prob >= 80:
-                story.append(Paragraph("<b>HR Verdict:</b> Outstanding candidate, highly recommended for immediate advancement.", body_text))
-            elif acceptance_prob >= 60:
-                story.append(Paragraph("<b>HR Verdict:</b> Strong candidate, suitable for further evaluation with targeted development.", body_text))
-            elif acceptance_prob >= 40:
-                story.append(Paragraph("<b>HR Verdict:</b> Moderate potential, requires additional assessment and skill-building.", body_text))
-            else:
-                story.append(Paragraph("<b>HR Verdict:</b> Limited fit, significant improvement needed for role alignment.", body_text))
-            story.append(Spacer(1, 0.3 * inch))
-            table_data = [
-                ['Metric', 'Value'],
-                ['Interview Duration', f"{analysis_data['text_analysis']['total_duration']:.2f} seconds"],
-                ['Speaker Turns', f"{analysis_data['text_analysis']['speaker_turns']}"],
-                ['Participants', ', '.join(sorted(analysis_data['speakers']))]
-            ]
-            table = Table(table_data, colWidths=[2.2*inch, 3.8*inch])
-            table.setStyle(TableStyle([
-                ('BACKGROUND', (0,0), (-1,0), colors.HexColor('#0050BC')),
-                ('TEXTCOLOR', (0,0), (-1,0), colors.white),
-                ('ALIGN', (0,0), (-1,-1), 'LEFT'),
-                ('VALIGN', (0,0), (-1,-1), 'MIDDLE'),
-                ('FONTNAME', (0,0), (-1,0), 'Helvetica-Bold'),
-                ('FONTSIZE', (0,0), (-1,-1), 9),
-                ('BOTTOMPADDING', (0,0), (-1,0), 10),
-                ('TOPPADDING', (0,0), (-1,0), 10),
-                ('BACKGROUND', (0,1), (-1,-1), colors.HexColor('#F5F6FA')),
-                ('GRID', (0,0), (-1,-1), 0.5, colors.HexColor('#DDE4EB'))
-            ]))
-            story.append(table)
-        story.append(Spacer(1, 0.4 * inch))
+        story.append(Paragraph(f"Generated: {time.strftime('%B %d, %Y')}", ParagraphStyle(name='Date', alignment=1, fontSize=9, textColor=colors.HexColor('#666666'), fontName='Helvetica')))
+        story.append(Spacer(1, 0.4*inch))
+        acceptance_prob = analysis_data.get('acceptance_probability', 50.0)
+        story.append(Paragraph("Hiring Suitability Snapshot", h2))
+        prob_color = colors.HexColor('#2E7D32') if acceptance_prob >= 80 else (colors.HexColor('#F57C00') if acceptance_prob >= 60 else colors.HexColor('#D32F2F'))
+        story.append(Paragraph(f"Suitability Score: <font size=15 color='{prob_color.hexval()}'><b>{acceptance_prob:.2f}%</b></font>",
+                             ParagraphStyle(name='Prob', fontSize=11, spaceAfter=10, alignment=1, fontName='Helvetica-Bold')))
+        if acceptance_prob >= 80:
+            story.append(Paragraph("<b>HR Verdict:</b> Outstanding candidate, recommended for immediate advancement.", body_text))
+        elif acceptance_prob >= 60:
+            story.append(Paragraph("<b>HR Verdict:</b> Strong candidate, suitable for further evaluation.", body_text))
+        elif acceptance_prob >= 40:
+            story.append(Paragraph("<b>HR Verdict:</b> Moderate potential, needs additional assessment.", body_text))
+        else:
+            story.append(Paragraph("<b>HR Verdict:</b> Limited fit, significant improvement required.", body_text))
+        story.append(Spacer(1, 0.3*inch))
+        table_data = [
+            ['Metric', 'Value'],
+            ['Interview Duration', f"{analysis_data['text_analysis']['total_duration']:.2f} seconds"],
+            ['Speaker Turns', f"{analysis_data['text_analysis']['speaker_turns']}"],
+            ['Participants', ', '.join(sorted(set(u['speaker'] for u in analysis_data['transcript'])))],
+        ]
+        table = Table(table_data, colWidths=[2.3*inch, 3.7*inch])
+        table.setStyle(TableStyle([
+            ('BACKGROUND', (0,0), (-1,0), colors.HexColor('#0050BC')),
+            ('TEXTCOLOR', (0,0), (-1,0), colors.white),
+            ('ALIGN', (0,0), (-1,-1), 'LEFT'),
+            ('VALIGN', (0,0), (-1,-1), 'MIDDLE'),
+            ('FONTNAME', (0,0), (-1,0), 'Helvetica-Bold'),
+            ('FONTSIZE', (0,0), (-1,-1), 9),
+            ('BOTTOMPADDING', (0,0), (-1,0), 8),
+            ('TOPPADDING', (0,0), (-1,0), 8),
+            ('BACKGROUND', (0,1), (-1,-1), colors.HexColor('#F5F6FA')),
+            ('GRID', (0,0), (-1,-1), 0.5, colors.HexColor('#DDE4EB')),
+        ]))
+        story.append(table)
+        story.append(Spacer(1, 0.4*inch))
         story.append(Paragraph("Prepared by: EvalBot - AI-Powered HR Analysis", body_text))
         story.append(PageBreak())
 
@@ -558,11 +551,11 @@ def create_pdf_report(analysis_data: Dict, output_path: str, gemini_report_text:
                 ['Metric', 'Value', 'HR Insight'],
                 ['Speaking Rate', f"{voice_analysis.get('speaking_rate', 0):.2f} words/sec", 'Benchmark: 2.0-3.0 wps; impacts clarity'],
                 ['Filler Words', f"{voice_analysis.get('filler_ratio', 0) * 100:.1f}%", 'High usage reduces credibility'],
-                ['Anxiety', voice_analysis.get('interpretation', {}).get('anxiety_level', 'N/A'), f"Score: {voice_analysis.get('composite_scores', {}).get('anxiety', 0):.3f}; stress response"],
-                ['Confidence', voice_analysis.get('interpretation', {}).get('confidence_level', 'N/A'), f"Score: {voice_analysis.get('composite_scores', {}).get('confidence', 0):.3f}; vocal strength"],
-                ['Fluency', voice_analysis.get('interpretation', {}).get('fluency_level', 'N/A'), 'Drives engagement']
+                ['Anxiety', voice_analysis.get('interpretation', {}).get('anxiety_level', 'N/A'), f"Score: {voice_analysis.get('composite_scores', {}).get('anxiety', 0):.3f}"],
+                ['Confidence', voice_analysis.get('interpretation', {}).get('confidence_level', 'N/A'), f"Score: {voice_analysis.get('composite_scores', {}).get('confidence', 0):.3f}"],
+                ['Fluency', voice_analysis.get('interpretation', {}).get('fluency_level', 'N/A'), 'Drives engagement'],
             ]
-            table = Table(table_data, colWidths=[1.7*inch, 1.2*inch, 3.1*inch])
+            table = Table(table_data, colWidths=[1.6*inch, 1.2*inch, 3.2*inch])
             table.setStyle(TableStyle([
                 ('BACKGROUND', (0,0), (-1,0), colors.HexColor('#0050BC')),
                 ('TEXTCOLOR', (0,0), (-1,0), colors.white),
@@ -570,124 +563,138 @@ def create_pdf_report(analysis_data: Dict, output_path: str, gemini_report_text:
                 ('VALIGN', (0,0), (-1,-1), 'MIDDLE'),
                 ('FONTNAME', (0,0), (-1,0), 'Helvetica-Bold'),
                 ('FONTSIZE', (0,0), (-1,-1), 9),
-                ('BOTTOMPADDING', (0,0), (-1,0), 10),
-                ('TOPPADDING', (0,0), (-1,0), 10),
+                ('BOTTOMPADDING', (0,0), (-1,0), 8),
+                ('TOPPADDING', (0,0), (-1,0), 8),
                 ('BACKGROUND', (0,1), (-1,-1), colors.HexColor('#F5F6FA')),
-                ('GRID', (0,0), (-1,-1), 0.5, colors.HexColor('#DDE4EB'))
+                ('GRID', (0,0), (-1,-1), 0.5, colors.HexColor('#DDE4EB')),
             ]))
             story.append(table)
-            story.append(Spacer(1, 0.2 * inch))
+            story.append(Spacer(1, 0.2*inch))
             chart_buffer = io.BytesIO()
             generate_anxiety_confidence_chart(voice_analysis.get('composite_scores', {}), chart_buffer)
             chart_buffer.seek(0)
-            img = Image(chart_buffer, width=4.8*inch, height=3.2*inch)
+            img = Image(chart_buffer, width=4.5*inch, height=3*inch)
             img.hAlign = 'CENTER'
             story.append(img)
         else:
             story.append(Paragraph("Vocal analysis unavailable.", body_text))
-        story.append(Spacer(1, 0.3 * inch))
+        story.append(Spacer(1, 0.2*inch))
 
         # Parse Gemini Report
         sections = {
             "Executive Summary": [],
-            "Communication and Vocal Dynamics": [],
-            "Competency and Content Evaluation": {"Strengths": [], "Growth Areas": []},
-            "Role Fit and Growth Potential": [],
-            "Strategic HR Recommendations": {"Development Priorities": [], "Next Steps": []}
+            "Communication": [],
+            "Competency": {"Strengths": [], "Growth Areas": []},
+            "Recommendations": {"Development": [], "Next Steps": []},
+            "Role Fit": [],
         }
-        report_parts = re.split(r'(\s*\*\*\s*\d\.\s*.*?\s*\*\*)', gemini_report_text)
         current_section = None
-        for part in report_parts:
-            if not part.strip(): continue
-            is_heading = False
-            for title in sections.keys():
-                if title.lower() in part.lower():
-                    current_section = title
-                    is_heading = True
-                    break
-            if not is_heading and current_section:
-                if current_section == "Competency and Content Evaluation":
-                    if 'strength' in part.lower() or any(k in part.lower() for k in ['leadership', 'problem-solving', 'communication', 'adaptability']):
-                        sections[current_section]["Strengths"].append(part.strip())
-                    elif 'improve' in part.lower() or 'grow' in part.lower() or 'challenge' in part.lower():
-                        sections[current_section]["Growth Areas"].append(part.strip())
-                elif current_section == "Strategic HR Recommendations":
-                    if any(k in part.lower() for k in ['communication', 'depth', 'presence', 'improve']):
-                        sections[current_section]["Development Priorities"].append(part.strip())
-                    elif any(k in part.lower() for k in ['advance', 'train', 'assess', 'next step']):
-                        sections[current_section]["Next Steps"].append(part.strip())
+        current_subsection = None
+        lines = gemini_report_text.split('\n')
+        for line in lines:
+            line = line.strip()
+            if not line: continue
+            if re.match(r'\s*\*\*\s*\d*\.?\s*.*?)\s*\*\*', line):
+                section_match = re.search(r'\s*\*\*\s*\d*\.?\s*(.*?)\s*\*\*', line)
+                section_title = section_match.group(1).strip()
+                if section_title.startswith('Executive Summary'):
+                    current_section = 'Executive Summary'
+                    current_subsection = None
+                elif 'Communication' in section_title:
+                    current_section = 'Communication'
+                    current_subsection = None
+                elif 'Competency' in section_title:
+                    current_section = 'Competency'
+                    current_subsection = None
+                elif 'Role Fit' in section_title:
+                    current_section = 'Role Fit'
+                    current_subsection = None
+                elif 'Recommendations' in section_title:
+                    current_section = 'Recommendations'
+                    current_subsection = None
+            elif line.startswith(('-', '*', '•')) and current_section:
+                clean_line = line.lstrip('-*• ').strip()
+                if not clean_line: continue
+                if current_section == 'Competency':
+                    if any(k in clean_line.lower() for k in ['leadership', 'problem-solving', 'communication', 'adaptability', 'strength']):
+                        current_subsection = 'Strengths'
+                    elif any(k in clean_line.lower() for k in ['improve', 'grow', 'depth', 'challenge']):
+                        current_subsection = 'Growth Areas'
+                    if current_subsection:
+                        sections[current_section][current_subsection].append(clean_line)
+                elif current_section == 'Recommendations':
+                    if any(k in clean_line.lower() for k in ['communication', 'technical', 'depth', 'presence']):
+                        current_subsection = 'Development'
+                    elif any(k in clean_line.lower() for k in ['advance', 'train', 'assess', 'next', 'mentor']):
+                        current_subsection = 'Next Steps'
+                    if current_subsection:
+                        sections[current_section][current_subsection].append(clean_line)
                 else:
-                    sections[current_section].append(part.strip())
+                    sections[current_section].append(clean_line)
 
         # Executive Summary
         story.append(Paragraph("2. Executive Summary", h2))
         if sections['Executive Summary']:
             for line in sections['Executive Summary']:
-                if line.startswith(('-', '•', '*')):
-                    story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
-                else:
-                    story.append(Paragraph(line, body_text))
+                story.append(Paragraph(line, bullet_style))
         else:
-            story.append(Paragraph("Summary unavailable.", body_text))
-        story.append(Spacer(1, 0.3 * inch))
+            story.append(Paragraph("No summary provided.", body_text))
+        story.append(Spacer(1, 0.2*inch))
 
         # Competency and Content
-        story.append(Paragraph("3. Competency & Content", h2))
+        story.append(Paragraph("3. Competency & Evaluation", h2))
         story.append(Paragraph("Strengths", h3))
-        if sections['Competency and Content Evaluation']['Strengths']:
-            for line in sections['Competency and Content Evaluation']['Strengths']:
-                story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
+        if sections['Competency']['Strengths']:
+            for line in sections['Competency']['Strengths']:
+                story.append(Paragraph(line, bullet_style))
         else:
             story.append(Paragraph("No strengths identified.", body_text))
-        story.append(Spacer(1, 0.2 * inch))
+        story.append(Spacer(1, 0.1*inch))
         story.append(Paragraph("Growth Areas", h3))
-        if sections['Competency and Content Evaluation']['Growth Areas']:
-            for line in sections['Competency and Content Evaluation']['Growth Areas']:
-                story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
+        if sections['Competency']['Growth Areas']:
+            for line in sections['Competency']['Growth Areas']:
+                story.append(Paragraph(line, bullet_style))
         else:
-            story.append(Paragraph("No growth areas identified.", body_text))
-        story.append(Spacer(1, 0.3 * inch))
+            story.append(Paragraph("No growth areas identified; maintain current strengths.", body_text))
+        story.append(Spacer(1, 0.2*inch))
 
         # Role Fit
         story.append(Paragraph("4. Role Fit & Potential", h2))
-        if sections['Role Fit and Growth Potential']:
-            for line in sections['Role Fit and Growth Potential']:
-                if line.startswith(('-', '•', '*')):
-                    story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
-                else:
-                    story.append(Paragraph(line, body_text))
+        if sections['Role Fit']:
+            for line in sections['Role Fit']:
+                story.append(Paragraph(line, bullet_style))
         else:
-            story.append(Paragraph("Fit and potential analysis unavailable.", body_text))
-        story.append(Spacer(1, 0.3 * inch))
+            story.append(Paragraph("No fit analysis provided.", body_text))
+        story.append(Spacer(1, 0.2*inch))
 
-        # Strategic Recommendations
-        story.append(Paragraph("5. Strategic Recommendations", h2))
+        # Recommendations
+        story.append(Paragraph("5. Recommendations", h2))
         story.append(Paragraph("Development Priorities", h3))
-        if sections['Strategic HR Recommendations']['Development Priorities']:
-            for line in sections['Strategic HR Recommendations']['Development Priorities']:
-                story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
+        if sections['Recommendations']['Development']:
+            for line in sections['Recommendations']['Development']:
+                story.append(Paragraph(line, bullet_style))
         else:
             story.append(Paragraph("No development priorities specified.", body_text))
-        story.append(Spacer(1, 0.2 * inch))
-        story.append(Paragraph("Next Steps for Managers", h3))
-        if sections['Strategic HR Recommendations']['Next Steps']:
-            for line in sections['Strategic HR Recommendations']['Next Steps']:
-                story.append(Paragraph(line.lstrip('-•* ').strip(), bullet_style))
+        story.append(Spacer(1, 0.1*inch))
+        story.append(Paragraph("Next Steps", h3))
+        if sections['Recommendations']['Next Steps']:
+            for line in sections['Recommendations']['Next Steps']:
+                story.append(Paragraph(line, bullet_style))
         else:
             story.append(Paragraph("No next steps provided.", body_text))
-        story.append(Spacer(1, 0.3 * inch))
-        story.append(Paragraph("This report provides a data-driven evaluation to guide hiring and development decisions.", body_text))
+        story.append(Spacer(1, 0.2*inch))
+        story.append(Paragraph("This report provides actionable insights to support hiring and candidate development.", body_text))
 
         doc.build(story, onFirstPage=header_footer, onLaterPages=header_footer)
         return True
     except Exception as e:
-        logger.error(f"PDF creation failed: {str(e)}", exc_info=True)
+        logger.error(f"PDF generation failed: {str(e)}")
         return False
 
 def convert_to_serializable(obj):
     if isinstance(obj, np.generic): return obj.item()
     if isinstance(obj, dict): return {k: convert_to_serializable(v) for k, v in obj.items()}
-    if isinstance(obj, list): return [convert_to_serializable(i) for i in obj]
+    if isinstance(obj, list): return [convert_to_serializable(item) for item in obj]
     if isinstance(obj, np.ndarray): return obj.tolist()
     return obj
 
@@ -730,18 +737,19 @@ def process_interview(audio_path_or_url: str):
         base_name = str(uuid.uuid4())
         pdf_path = os.path.join(OUTPUT_DIR, f"{base_name}_report.pdf")
         json_path = os.path.join(OUTPUT_DIR, f"{base_name}_analysis.json")
-        create_pdf_report(analysis_data, pdf_path, gemini_report_text=gemini_report_text)
+        if create_pdf_report(analysis_data, pdf_path, gemini_report_text):
+            logger.info(f"PDF report generated at: {pdf_path}")
         with open(json_path, 'w') as f:
             serializable_data = convert_to_serializable(analysis_data)
             json.dump(serializable_data, f, indent=2)
         logger.info(f"Processing completed for {audio_path_or_url}")
         return {'pdf_path': pdf_path, 'json_path': json_path}
     except Exception as e:
-        logger.error(f"Processing failed for {audio_path_or_url}: {str(e)}", exc_info=True)
+        logger.error(f"Processing failed for {audio_path_or_url}: {str(e)}")
         raise
     finally:
         if wav_file and os.path.exists(wav_file):
             os.remove(wav_file)
         if is_downloaded and local_audio_path and os.path.exists(local_audio_path):
             os.remove(local_audio_path)
-            logger.info(f"Cleaned up temporary downloaded file: {local_audio_path}")
+            logger.info(f"Cleaned up temporary audio file: {local_audio_path}")