Update app/services/rag_service.py

app/services/rag_service.py

@@ -1,6 +1,7 @@
 from app.config import settings
 from sqlalchemy.orm import Session
 from app.services.search_service import SearchService
+from app.models.database_models import PathologyReport
 
 try:
     from groq import Groq
@@ -25,17 +26,22 @@ class RAGService:
 
     @staticmethod
     def format_search_results(search_results: list) -> str:
-        """Format search results as context"""
+        """Format search results as detailed medical context"""
         if not search_results:
             return "No previous pathology records retrieved for this query.\n\n"
         context = "RELEVANT MEDICAL DOCUMENTS FROM EMR DATABASE:\n\n"
         for i, result in enumerate(search_results, 1):
             context += f"Document #{i} (ID: {result.get('document_id', 'Unknown')})\n"
             context += f"  Patient Name: {result.get('patient_name', 'Unknown')}\n"
+            context += f"  Patient ID / MRN: {result.get('patient_id', 'Unknown')}\n"
             context += f"  Test Type: {result.get('test_type', 'Unknown')}\n"
-            context += f"  Diagnosis: {result.get('diagnosis', 'Unknown')}\n"
+            context += f"  Diagnosis / Assessment: {result.get('diagnosis', 'Unknown')}\n"
             context += f"  Report Summary: {result.get('summary', 'Unknown')}\n"
-            context += f"  Text Content Preview: {result.get('text_preview', '')}\n\n"
+            if result.get('findings'):
+                context += f"  Biomarker Findings / Lab Results: {result.get('findings')}\n"
+            elif result.get('text_preview'):
+                context += f"  Text Content Preview: {result.get('text_preview', '')}\n"
+            context += "\n"
         return context
 
     def answer_question(
@@ -51,15 +57,75 @@ class RAGService:
                     "source_documents": [],
                 }
 
-            # 1. Retrieve semantically relevant documents if available in DB
+            # 1. Retrieve relevant documents: Direct Patient/MRN lookup + Semantic vector search
             results_list = []
+            seen_doc_ids = set()
+
+            # A. Direct query matching for patient name, MRN, document ID, and test type
+            import re
+            tokens = [t for t in re.findall(r'[a-zA-Z0-9_-]+', question) if len(t) >= 3 and t.lower() not in ['summarize', 'diagnostic', 'history', 'abnormal', 'flags', 'findings', 'what', 'with', 'from', 'explain', 'show', 'tell', 'about', 'report', 'reports', 'test', 'results']]
+            
+            for token in tokens:
+                matched_reports = (
+                    db.query(PathologyReport)
+                    .filter(
+                        (PathologyReport.patient_id.ilike(f"%{token}%")) | 
+                        (PathologyReport.patient_name.ilike(f"%{token}%")) | 
+                        (PathologyReport.document_id.ilike(f"%{token}%")) |
+                        (PathologyReport.test_type.ilike(f"%{token}%"))
+                    )
+                    .limit(5)
+                    .all()
+                )
+                for r in matched_reports:
+                    if r.document_id not in seen_doc_ids:
+                        seen_doc_ids.add(r.document_id)
+                        results_list.append({
+                            "document_id": r.document_id,
+                            "patient_name": r.patient_name or "Unknown",
+                            "patient_id": r.patient_id or "Unknown",
+                            "test_type": r.test_type or "Pathology Report",
+                            "diagnosis": r.diagnosis or "Normal",
+                            "summary": r.summary or "Diagnostic report recorded in EMR",
+                            "findings": r.findings or "",
+                            "text_preview": r.summary or "",
+                        })
+
+            # B. Vector / Semantic search fallback
             try:
                 search_result = SearchService.semantic_search(db, question, top_k=3)
                 if search_result.get("status") == "success":
-                    results_list = search_result.get("results", [])
+                    for res in search_result.get("results", []):
+                        doc_id = res.get("document_id")
+                        if doc_id and doc_id not in seen_doc_ids:
+                            seen_doc_ids.add(doc_id)
+                            results_list.append(res)
             except Exception as search_err:
                 print(f"[WARN] Vector search lookup skipped: {search_err}")
 
+            # C. If still 0 documents found, check if there are existing reports in DB as contextual fallback
+            if len(results_list) == 0:
+                all_recent = db.query(PathologyReport).order_by(PathologyReport.created_at.desc()).limit(3).all()
+                for r in all_recent:
+                    # If query mentions any part of patient name or diagnosis
+                    pname = (r.patient_name or "").lower()
+                    diag = (r.diagnosis or "").lower()
+                    ttype = (r.test_type or "").lower()
+                    q_lower = question.lower()
+                    if any(part in q_lower for part in pname.split() if len(part) > 2) or any(part in q_lower for part in ttype.split() if len(part) > 2):
+                        if r.document_id not in seen_doc_ids:
+                            seen_doc_ids.add(r.document_id)
+                            results_list.append({
+                                "document_id": r.document_id,
+                                "patient_name": r.patient_name or "Unknown",
+                                "patient_id": r.patient_id or "Unknown",
+                                "test_type": r.test_type or "Pathology Report",
+                                "diagnosis": r.diagnosis or "Normal",
+                                "summary": r.summary or "Diagnostic report recorded in EMR",
+                                "findings": r.findings or "",
+                                "text_preview": r.summary or "",
+                            })
+
             # 2. Format context from search results
             context = self.format_search_results(results_list)