Fixed Multi Agent RAG issue

Author: debabratamishra

app/services/rag_service.py

@@ -1003,6 +1003,7 @@ async def query(self, query_text, system_prompt="You are a helpful assistant.",
         async for chunk in stream_ollama(llm_messages, model=model_name):
             yield chunk
 
+
 class CrewAIRAGOrchestrator:
     """Coordinates refined querying and answer composition using an Ollama-backed model."""
     def __init__(self, rag_service: RAGService, model_name="gemma3n:e2b"):
@@ -1033,6 +1034,19 @@ def __init__(self, rag_service: RAGService, model_name="gemma3n:e2b"):
             llm=self.ollama_llm
         )
 
+    def _get_ollama_url(self) -> str:
+        """Get the appropriate Ollama URL based on execution environment."""
+        try:
+            from app.services.host_service_manager import host_service_manager
+            url = host_service_manager.environment_config.ollama_url
+            logger.debug(f"Using Ollama URL from host service manager: {url}")
+            return url
+        except ImportError:
+            logger.warning("Host service manager not available, using fallback config")
+            url = os.getenv("OLLAMA_API_BASE", "http://localhost:11434")
+            logger.debug(f"Using fallback Ollama URL: {url}")
+            return url
+
     async def _get_context_length(self, model_name: str) -> int:
         """Query Ollama for the model's context length; fall back to a sensible default."""
         model = model_name.replace("ollama/", "") if "ollama/" in model_name else model_name
@@ -1073,7 +1087,7 @@ async def _generate_summary(self, text: str, system_prompt: str) -> str:
 
         return response
 
-    def simple_chunk_text(self, text: str, chunk_size: int):
+    def chunk_text(self, text: str, chunk_size: int):
         """Simple character-based chunking without overlap (used for specific cases)."""
         if not text or not text.strip():
             return []

main.py

@@ -611,17 +611,40 @@ async def event_generator():
 
             return StreamingResponse(event_generator(), media_type="text/plain")
         else:
-            # Ollama RAG
+            # Ollama RAG - check if multi-agent is requested
             async def event_generator():
-                async for chunk in rag_service.query(
-                    request.query, request.system_prompt, request.messages,
-                    request.n_results, request.use_hybrid_search, request.model
-                ):
-                    yield chunk + "\n"
+                if request.use_multi_agent:
+                    # Use CrewAI multi-agent orchestration
+                    logger.info("Using CrewAI multi-agent orchestration")
+                    from app.services.rag_service import CrewAIRAGOrchestrator
+                    
+                    orchestrator = CrewAIRAGOrchestrator(
+                        rag_service=rag_service,
+                        model_name=request.model or "gemma3n:e2b"
+                    )
+                    
+                    async for chunk in orchestrator.query(
+                        user_query=request.query,
+                        system_prompt=request.system_prompt,
+                        messages=request.messages,
+                        n_results=request.n_results,
+                        use_hybrid_search=request.use_hybrid_search,
+                        model=request.model
+                    ):
+                        yield chunk + "\n"
+                else:
+                    # Use regular RAG
+                    logger.info("Using standard RAG without multi-agent orchestration")
+                    async for chunk in rag_service.query(
+                        request.query, request.system_prompt, request.messages,
+                        request.n_results, request.use_hybrid_search, request.model
+                    ):
+                        yield chunk + "\n"
 
             return StreamingResponse(event_generator(), media_type="text/plain")
 
     except Exception as e:
+        logger.error(f"RAG query error: {e}")
         raise HTTPException(status_code=500, detail=str(e))