Update process_interview.py

process_interview.py

@@ -39,7 +39,7 @@ matplotlib.use('Agg')
 
 # Concurrency
 from concurrent.futures import ThreadPoolExecutor
-import joblib  # Added import
+import joblib
 
 # ==============================================================================
 # 2. CONFIGURATION AND INITIALIZATION
@@ -67,7 +67,6 @@ if not all([PINECONE_KEY, ASSEMBLYAI_KEY, GEMINI_API_KEY]):
 index, gemini_model, speaker_model, nlp, tokenizer, text_embedding_model = (None,) * 6
 
 def initialize_all_services_and_models():
-    """Initializes all external services and loads all AI models into memory."""
     global index, gemini_model, speaker_model, nlp, tokenizer, text_embedding_model
     logger.info("Initializing all services and loading all models...")
     pc = Pinecone(api_key=PINECONE_KEY)
@@ -144,7 +143,7 @@ def identify_speakers(transcript: Dict, wav_file_path: str) -> List[Dict]:
 
         def process_utterance(utterance):
             start_ms, end_ms = utterance['start'], utterance['end']
-            if end_ms - start_ms < 1000:
+            if end_ms - start_ms < 1000:  # Skip short utterances
                 return {**utterance, 'speaker_id': 'unknown_short_utterance'}
             with temp_audio_file() as temp_path:
                 full_audio[start_ms:end_ms].export(temp_path, format="wav")
@@ -199,7 +198,7 @@ def convert_to_serializable(obj):
         return obj.tolist()
     if isinstance(obj, bytes):
         import base64
-        return base64.b64encode(obj).decode('utf-8')  # تحويل bytes لـ base64 string
+        return base64.b64encode(obj).decode('utf-8')
     if isinstance(obj, dict):
         return {k: convert_to_serializable(v) for k, v in obj.items()}
     if isinstance(obj, list):
@@ -219,7 +218,7 @@ def classify_roles_ultimate(utterances: List[Dict], audio_path: str) -> List[Dic
     interviewer_keywords = r'\b(what|why|how|when|where|who|which|tell me about|can you explain|describe|give me an example)\b'
     for u in utterances:
         sid, text = u.get('speaker_id'), u.get('text', '').lower()
-        if sid not in speaker_data or not text:
+        if sid not in speaker_data or not text or sid.startswith('unknown'):
             continue
         rule_score = 10 if text.endswith('?') else 0
         rule_score += 5 * len(re.findall(interviewer_keywords, text))
@@ -259,9 +258,9 @@ def analyze_interviewee_voice(audio_path: str, utterances: List[Dict]) -> Dict:
     logger.info("Performing detailed voice analysis using your custom function...")
     try:
         y, sr = librosa.load(audio_path, sr=16000)
-        interviewee_utterances = [u for u in utterances if u.get('role') == 'Interviewee']
+        interviewee_utterances = [u for u in utterances if u.get('role') == 'Interviewee' and not u['speaker_id'].startswith('unknown')]
         if not interviewee_utterances:
-            return {'error': 'No interviewee utterances found'}
+            return {'error': 'No valid interviewee utterances found'}
         segments = [y[int(u['start'] * sr / 1000):int(u['end'] * sr / 1000)] for u in interviewee_utterances]
         if not segments:
             return {'error': 'No valid interviewee segments to analyze.'}
@@ -351,13 +350,11 @@ def calculate_acceptance_probability(analysis_data: Dict) -> float:
 # 6. AI-POWERED NARRATIVE AND PDF REPORTING
 # ==============================================================================
 def generate_gemini_report_text(analysis_data: Dict) -> str:
-    """Generates a comprehensive narrative report using the Gemini model, based on your prompt structure."""
     logger.info("Generating AI-powered narrative report with Gemini...")
     voice = analysis_data.get('voice_analysis_metrics', {})
     interviewee_text = "\n".join([f"- {u['text']}" for u in analysis_data['transcript_with_roles'] if u.get('role') == 'Interviewee'])
     acceptance_prob = analysis_data.get('acceptance_probability', 50.0)
 
-    # Format numbers only if they are not 'N/A' or strings
     def format_value(val):
         return f"{val:.2f}" if isinstance(val, (int, float)) else val
 
@@ -397,7 +394,6 @@ def generate_gemini_report_text(analysis_data: Dict) -> str:
         return "Error: Could not generate AI analysis report."
 
 def create_pdf_report(analysis_data: Dict, output_path: str):
-    """Generates a detailed, professional PDF report including all analysis sections, based on your structure."""
     logger.info(f"Generating comprehensive PDF report at {output_path}...")
     doc = SimpleDocTemplate(output_path, pagesize=letter, topMargin=inch, bottomMargin=inch)
     styles = getSampleStyleSheet()
@@ -407,47 +403,55 @@ def create_pdf_report(analysis_data: Dict, output_path: str):
                               textColor=colors.HexColor('#0050BC'), fontName='Helvetica-Bold'))
     styles.add(ParagraphStyle(name='H3', fontSize=12, leading=16, spaceBefore=10, spaceAfter=6,
                               textColor=colors.HexColor('#333333'), fontName='Helvetica-Bold'))
-    # تعديل الـ Body style ليحتوي على indentation
     styles.add(ParagraphStyle(name='Body', fontSize=10, leading=14, spaceAfter=6, alignment=TA_JUSTIFY,
-                              leftIndent=10))  # نقل leftIndent للـ style
-    story = []
+                              leftIndent=10))
 
-    # Cover Page
-    story.append(Paragraph("Candidate Interview Analysis Report", styles['H1']))
-    story.append(Spacer(1, 0.2 * inch))
-    story.append(Paragraph(f"Candidate ID: {analysis_data.get('user_id', 'N/A')}", styles['Body']))
-    story.append(Paragraph(f"Date of Analysis: {time.strftime('%B %d, %Y')}", styles['Body']))
-    prob = analysis_data.get('acceptance_probability', 0)
-    prob_color = 'green' if prob >= 75 else 'orange' if prob >= 50 else 'red'
-    story.append(Paragraph(f"<b>Overall Suitability Score:</b> <font size=16 color='{prob_color}'>{prob}%</font>", styles['H2']))
-    story.append(PageBreak())
-
-    # Quantitative Analysis Page
-    story.append(Paragraph("Quantitative Vocal Analysis", styles['H2']))
-    if analysis_data.get('chart_image_bytes'):
-        story.append(Image(io.BytesIO(analysis_data['chart_image_bytes']), width=5.5 * inch, height=3.3 * inch))
+    story = []
+    try:
+        # Cover Page
+        story.append(Paragraph("Candidate Interview Analysis Report", styles['H1']))
+        story.append(Spacer(1, 0.2 * inch))
+        story.append(Paragraph(f"Candidate ID: {analysis_data.get('user_id', 'N/A')}", styles['Body']))
+        story.append(Paragraph(f"Date of Analysis: {time.strftime('%B %d, %Y')}", styles['Body']))
+        prob = analysis_data.get('acceptance_probability', 0)
+        prob_color = 'green' if prob >= 75 else 'orange' if prob >= 50 else 'red'
+        story.append(Paragraph(f"<b>Overall Suitability Score:</b> <font size=16 color='{prob_color}'>{prob}%</font>", styles['H2']))
+        story.append(PageBreak())
+
+        # Quantitative Analysis Page
+        story.append(Paragraph("Quantitative Vocal Analysis", styles['H2']))
+        if analysis_data.get('chart_image_bytes'):
+            logger.debug("Adding chart image to PDF")
+            img_buffer = io.BytesIO(analysis_data['chart_image_bytes'])
+            story.append(Image(img_buffer, width=5.5 * inch, height=3.3 * inch))
+        else:
+            story.append(Paragraph("No chart data available.", styles['Body']))
         story.append(Spacer(1, 0.2 * inch))
 
-    voice_text = analysis_data.get('voice_interpretation_text', 'Not available.').replace('\n', '<br/>')
-    story.append(Paragraph(voice_text, styles['Body']))
-    story.append(Spacer(1, 0.2 * inch))
+        voice_text = analysis_data.get('voice_interpretation_text', 'Not available.').replace('\n', '<br/>')
+        story.append(Paragraph(voice_text, styles['Body']))
+        story.append(Spacer(1, 0.2 * inch))
 
-    # AI-Generated Narrative Page
-    story.append(Paragraph("Qualitative AI-Powered Report", styles['H2']))
-    gemini_text = analysis_data.get('gemini_report_text', 'Not available.')
-    for line in gemini_text.split('\n'):
-        line = line.strip()
-        if not line:
-            continue
-        if line.startswith('**') and line.endswith('**'):
-            story.append(Paragraph(line.strip('*'), styles['H3']))
-        elif line.startswith('- ') or line.startswith('* '):
-            story.append(Paragraph(f"• {line[2:]}", styles['Body']))  # ازالة leftIndent من هنا
-        else:
-            story.append(Paragraph(line, styles['Body']))
+        # AI-Generated Narrative Page
+        story.append(Paragraph("Qualitative AI-Powered Report", styles['H2']))
+        gemini_text = analysis_data.get('gemini_report_text', 'Not available.')
+        for line in gemini_text.split('\n'):
+            line = line.strip()
+            if not line:
+                continue
+            if line.startswith('**') and line.endswith('**'):
+                story.append(Paragraph(line.strip('*'), styles['H3']))
+            elif line.startswith('- ') or line.startswith('* '):
+                story.append(Paragraph(f"• {line[2:]}", styles['Body']))
+            else:
+                story.append(Paragraph(line, styles['Body']))
+
+        doc.build(story)
+        logger.info("PDF report generated successfully.")
+    except Exception as e:
+        logger.error(f"Error generating PDF: {e}", exc_info=True)
+        raise
 
-    doc.build(story)
-    logger.info("PDF report generated successfully.")
 # ==============================================================================
 # 7. MAIN PROCESSING PIPELINE
 # ==============================================================================
@@ -471,7 +475,6 @@ def process_interview(audio_path: str, user_id: str = "candidate-123") -> Dict:
         logger.info("Identifying speakers")
         utterances_with_speakers = identify_speakers(transcript, wav_file)
         
-        # Add duration feature here
         logger.info("Extracting duration features")
         utterances_with_duration = extract_duration_feature(utterances_with_speakers)
         
@@ -498,7 +501,7 @@ def process_interview(audio_path: str, user_id: str = "candidate-123") -> Dict:
             },
             'acceptance_probability': calculate_acceptance_probability({'voice_analysis_metrics': voice_analysis}),
             'voice_interpretation_text': voice_interpretation,
-            'chart_image_bytes': chart_buffer.getvalue()
+            'chart_image_bytes': chart_buffer.getvalue() if chart_buffer.tell() > 0 else None
         }
         
         logger.info("Generating report text using Gemini")