Fix data loading: use correct JSONL file and display group information

- Load from assets/structured_results_o3_with_inputs.jsonl instead of Qwen JSONL
- Display patient group, age, gender, race, and note length in UI
- Update expected patient count from 20 to 24
- Add patient info header to raw data display

chatgpt.py

@@ -903,38 +903,36 @@ with gr.Blocks() as app:
         return text.strip()
     
     # Load from new JSONL format first (single admission per record)
-    # IMPORTANT: Only load from Qwen JSONL, skip all other data sources
-    jsonl_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'Qwen__Qwen2.5-7B-Instruct_io.jsonl')
+    # IMPORTANT: Load from assets/structured_results_o3_with_inputs.jsonl
+    jsonl_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'assets', 'structured_results_o3_with_inputs.jsonl')
     jsonl_loaded = False
     eval_data_by_patient = {}  # Initialize empty to ensure clean start
     
-    # FORCE: Only load from Qwen JSONL, completely skip all other data sources
+    # FORCE: Only load from structured_results JSONL, completely skip all other data sources
     if os.path.exists(jsonl_path):
-        print(f"[INFO] ========== LOADING FROM QWEN JSONL ONLY ==========")
+        print(f"[INFO] ========== LOADING FROM STRUCTURED RESULTS JSONL ONLY ==========")
         print(f"Loading data from JSONL: {jsonl_path}")
         jsonl_data = load_jsonl(jsonl_path)
         print(f"Found {len(jsonl_data)} entries in JSONL file")
         entry_count = 0
         for entry in jsonl_data:
             entry_count += 1
-            full_id = entry.get('patient_id', '') or ''
-            # Extract patient_id from format like "11318742_admission_29646478_input"
-            # Only keep the patient_id part (e.g., "11318742")
-            if '_admission_' in full_id:
-                patient_id = full_id.split('_admission_')[0]
-                # Extract admission_id from the remaining part (e.g., "29646478_input" -> "29646478")
-                admission_part = full_id.split('_admission_', 1)[1]
-                admission_id = admission_part.split('_')[0] if '_' in admission_part else admission_part
-            else:
-                # Fallback: if format doesn't match, try to extract just the numeric patient ID
-                patient_id = full_id.split('_')[0] if '_' in full_id else full_id
-                admission_id = '1'  # Default admission ID
+            # Get patient and admission IDs directly from the entry
+            patient_id = str(entry.get('subject_id', ''))
+            admission_id = str(entry.get('hadm_id', ''))
+            
+            # Get group information
+            group = entry.get('group', 'Unknown')
+            age = entry.get('age', 'Unknown')
+            gender = entry.get('gender', 'Unknown')
+            race = entry.get('race', 'Unknown')
+            note_length = entry.get('note_length_non_ws', 0)
             
             # Get input and output fields
-            input_text = entry.get('input', '')
-            raw_summary = entry.get('output', '')
+            input_text = entry.get('raw_note', '')
+            raw_summary = entry.get('content', '')
             # Clean the output to extract only markdown summary
-            summary = clean_summary_output(raw_summary)
+            summary = clean_summary_output(raw_summary) if raw_summary else ''
             
             # Extract admission time from input text
             admittime = None
@@ -949,6 +947,11 @@ with gr.Blocks() as app:
                 'patient_id': patient_id,
                 'admission_id': admission_id,
                 'admission_index': 1,
+                'group': group,  # Add group information
+                'age': age,
+                'gender': gender,
+                'race': race,
+                'note_length': note_length,
                 'input_text': input_text,  # This will be shown in "Patient raw data from EHR"
                 'summary': summary,  # This will be shown in "AI Generated Summary" (cleaned)
                 'admittime': admittime,
@@ -1114,12 +1117,12 @@ with gr.Blocks() as app:
     print(f"[DEBUG] JSONL loaded: {jsonl_loaded}")
     print(f"[DEBUG] Total patients in UI list: {len(patient_ids_list)}")
     print(f"[DEBUG] Patient IDs: {patient_ids_list}")
-    if len(patient_ids_list) != 20:
-        print(f"[ERROR] Expected 20 patients but found {len(patient_ids_list)}!")
-        print(f"[ERROR] This indicates data is being loaded from multiple sources!")
-        print(f"[ERROR] Check the logs above to see which sources were loaded.")
+    if len(patient_ids_list) != 24:
+        print(f"[WARNING] Expected 24 patients but found {len(patient_ids_list)}!")
+        print(f"[WARNING] This may indicate data is being loaded from multiple sources!")
+        print(f"[WARNING] Check the logs above to see which sources were loaded.")
     else:
-        print(f"[DEBUG] ✓ Correct number of patients (20) loaded from Qwen JSONL")
+        print(f"[DEBUG] ✓ Correct number of patients (24) loaded from structured results JSONL")
     print(f"[DEBUG] ============================================")
 
     with gr.Tabs():
@@ -1541,12 +1544,32 @@ with gr.Blocks() as app:
 
                     summary_text = clean_summary_output(rec.get('summary', '') or 'No summary available')
                     input_text = rec.get('input_text', '') or 'No input data available'
+                    
+                    # Get patient metadata
+                    group = rec.get('group', 'Unknown')
+                    age = rec.get('age', 'Unknown')
+                    gender = rec.get('gender', 'Unknown')
+                    race = rec.get('race', 'Unknown')
+                    note_length = rec.get('note_length', 0)
+                    admission_id = rec.get('admission_id', 'Unknown')
+                    
+                    # Add patient info header
+                    patient_info_header = (
+                        f"=== PATIENT INFORMATION ===\n"
+                        f"Patient ID: {pid} | Admission ID: {admission_id}\n"
+                        f"Group: {group} | Note Length: {note_length:,} chars\n"
+                        f"Demographics: Age {age}, {gender}, {race}\n"
+                        f"{'='*80}\n\n"
+                    )
 
                     # Replace ICD codes and Lab IDs with descriptions
                     input_text_with_descriptions = replace_icd_codes_and_lab_ids(input_text)
+                    
+                    # Prepend patient info to input text
+                    full_input_text = patient_info_header + input_text_with_descriptions
 
                     sanitized_input = (
-                        input_text_with_descriptions.replace('&', '&')
+                        full_input_text.replace('&', '&')
                         .replace('<', '&lt;')
                         .replace('>', '&gt;')
                     )