Update pipeline.py

pipeline.py

@@ -628,109 +628,78 @@ def draw_question_overlay(frame, text, width, height):
 
 
 ##Main pipeline
-def run_intervision_pipeline(video_path, questions_config, output_dir):
-    """
-    Run the full Intervision analysis pipeline.
-
-    Steps:
-    1. Extract baseline audio
-    2. Run video annotation
-    3. Merge annotated video with original audio
-    4. Generate report
-    """
+# ... (Keep all your imports and utility functions at the top as they were) ...
 
+def run_intervision_pipeline(video_path, questions_config, output_dir):
     os.makedirs(output_dir, exist_ok=True)
-
     print("[PIPELINE] Starting pipeline")
-    print("[PIPELINE] Video path:", video_path)
-
-    # ---------------------------------------------------
-    #Extract baseline audio (first 10 seconds)
-    # ---------------------------------------------------
 
+    # 1. Extract baseline audio
     baseline_wav = os.path.join(output_dir, "baseline.wav")
-
-    print("[PIPELINE] Extracting baseline audio")
-
     subprocess.run([
-        "ffmpeg",
-        "-y",
-        "-i", video_path,
-        "-t", "10",
-        "-vn",
-        "-acodec", "pcm_s16le",
-        "-ar", "16000",
-        baseline_wav
+        "ffmpeg", "-y", "-i", video_path, "-t", "10", 
+        "-vn", "-acodec", "pcm_s16le", "-ar", "16000", baseline_wav
     ], check=True)
-
-    if not os.path.exists(baseline_wav):
-        raise Exception("Baseline audio extraction failed")
-
+    
     y_b, sr_b = librosa.load(baseline_wav, sr=16000)
-
     baseline_features = extract_audio_features(y_b, sr_b)
 
-    # ---------------------------------------------------
-    #Process video frames and annotate
-    # ---------------------------------------------------
-
-    print("[PIPELINE] Running video annotation")
+    # 2. Process and Annotate Video
+    annotated_video_raw = process_full_video(video_path, output_dir, questions_config)
 
-    annotated_video_raw = process_full_video(
-        video_path,
-        output_dir,
-        questions_config
-    )
-
-    if not os.path.exists(annotated_video_raw):
-        raise Exception("Annotated video was not generated")
-
-    # ---------------------------------------------------
-    #Merge annotated video with original audio
-    # ---------------------------------------------------
-
-    final_output = os.path.join(
-        output_dir,
-        "Intervision_Final_Report.mp4"
-    )
-
-    print("[PIPELINE] Merging audio and annotated video")
+    # 3. Analyze each question segment for the Report
+    list_of_answer_report = []
+    full_audio, sr = librosa.load(video_path, sr=16000)
 
+    for q in questions_config:
+        print(f"[PIPELINE] Analyzing Question ID: {q['question_id']}")
+        
+        # Extract specific segment for this question
+        start_sample = int(q['start_time'] * sr)
+        end_sample = int(q['end_time'] * sr)
+        segment_audio = full_audio[start_sample:end_sample]
+        
+        temp_segment_path = os.path.join(output_dir, f"temp_{q['question_id']}.wav")
+        sf.write(temp_segment_path, segment_audio, sr)
+
+        # AI Analysis
+        user_text = get_user_answer(temp_segment_path)
+        similarity = compute_similarity_score(user_text, q['ideal_answer'])
+        relevance = compute_relevance_score(q['question_text'], user_text)
+        audio_scores = analyze_audio_segment(temp_segment_path, baseline=baseline_features)
+
+        list_of_answer_report.append({
+            "questionId": q['question_id'],
+            "userAnswerText": user_text,
+            "score": similarity,
+            "relevance": relevance,
+            "confidence": audio_scores["confidence_audio"],
+            "stress": audio_scores["stress"],
+            "clarity": audio_scores["clarity"],
+            "pauses": audio_scores["pauses"],
+            "toneOfVoice": audio_scores["tone_of_voice"],
+            "status": "answered"
+        })
+        
+        if os.path.exists(temp_segment_path):
+            os.remove(temp_segment_path)
+
+    # 4. Final Merge
+    final_output = os.path.join(output_dir, "Intervision_Final_Report.mp4")
     subprocess.run([
-        'ffmpeg', '-y',
-        '-i', annotated_video_raw,
-        '-i', video_path,
-        '-map', '0:v:0',
-        '-map', '1:a:0',
-        '-c:v', 'libx264',
-        '-preset', 'veryfast',
-        '-crf', '23',
-        '-c:a', 'aac',
-        '-b:a', '160k',
-        '-shortest',
-        final_output
+        'ffmpeg', '-y', '-i', annotated_video_raw, '-i', video_path,
+        '-map', '0:v:0', '-map', '1:a:0', '-c:v', 'libx264', '-preset', 'veryfast',
+        '-crf', '23', '-c:a', 'aac', '-shortest', final_output
     ], check=True)
 
-
-    if not os.path.exists(final_output):
-        raise Exception("Final video merge failed")
-
-    print("[PIPELINE] Final video created:", final_output)
-
-    # ---------------------------------------------------
-    # Generate report JSON
-    # ---------------------------------------------------
-
+    # 5. Save Report with the correct KEY
     report = {
         "status": "completed",
-        "questionsAnalyzed": len(questions_config)
+        "questionsAnalyzed": len(questions_config),
+        "listOfAnswerReport": list_of_answer_report
     }
-
     report_path = os.path.join(output_dir, "report.json")
-
     with open(report_path, "w") as f:
         json.dump(report, f, indent=2)
 
-    print("[PIPELINE] Report saved:", report_path)
-
     return final_output, report_path