ai-service/app/api/chat.py

@@ -32,12 +32,12 @@ async def get_orchestrator_service_with_memory(
     Ensures each request has a fresh MemoryService with database session.
     """
     from app.services.llm.factory import get_llm_config_manager
-    from app.services.retrieval.vector_retriever import get_vector_retriever
+    from app.services.retrieval.optimized_retriever import get_optimized_retriever
     
     memory_service = MemoryService(session)
     llm_config_manager = get_llm_config_manager()
     llm_client = llm_config_manager.get_client()
-    retriever = await get_vector_retriever()
+    retriever = await get_optimized_retriever()
     
     return OrchestratorService(
         llm_client=llm_client,

ai-service/app/services/orchestrator.py

@@ -402,6 +402,12 @@ class OrchestratorService:
             messages.extend(ctx.merged_context.messages)
 
         messages.append({"role": "user", "content": ctx.current_message})
+        
+        logger.info(
+            f"[AC-AISVC-02] Built {len(messages)} messages for LLM: "
+            f"system_len={len(system_content)}, history_count={len(ctx.merged_context.messages) if ctx.merged_context else 0}"
+        )
+        logger.debug(f"[AC-AISVC-02] System prompt preview: {system_content[:500]}...")
 
         return messages