fixing agent

benchmarking/llm_providers/medrax_provider.py

@@ -39,7 +39,7 @@ class MedRAXProvider(LLMProvider):
                 "TorchXRayVisionClassifierTool",  # For classifying chest X-ray images using TorchXRayVision
                 "ArcPlusClassifierTool",  # For advanced chest X-ray classification using ArcPlus
                 "ChestXRayReportGeneratorTool",  # For generating medical reports from X-rays
-                "XRayPhraseGroundingTool",  # For locating described features in X-rays
+                # "XRayPhraseGroundingTool",  # For locating described features in X-rays
                 "MedGemmaVQATool", # Google MedGemma VQA tool
                 "MedicalRAGTool",  # For retrieval-augmented generation with medical knowledge
                 "WebBrowserTool",  # For web browsing and search capabilities

medrax/agent/agent.py

@@ -6,7 +6,7 @@ from datetime import datetime
 from typing import List, Dict, Any, TypedDict, Annotated, Optional
 
 from langgraph.graph import StateGraph, END
-from langchain_core.messages import AnyMessage, SystemMessage, ToolMessage
+from langchain_core.messages import AnyMessage, SystemMessage, ToolMessage, HumanMessage
 from langchain_core.language_models import BaseLanguageModel
 from langchain_core.tools import BaseTool
 
@@ -112,8 +112,27 @@ class Agent:
             Dict[str, List[AnyMessage]]: A dictionary containing the model's response.
         """
         messages = state["messages"]
-        if self.system_prompt:
+    
+        # Only add system prompt if it's not already present (i.e., first call in this conversation)
+        # This avoids redundantly sending the system prompt on every model invocation
+        if self.system_prompt and (len(messages) == 0 or not isinstance(messages[0], SystemMessage)):
             messages = [SystemMessage(content=self.system_prompt)] + messages
+        
+        # Check if we just executed tools by checking if the last message is a ToolMessage
+        # This indicates we're in the process node immediately after the execute node
+        has_tool_results = len(messages) > 0 and isinstance(messages[-1], ToolMessage)
+
+        # If we have tool results, add explicit instruction to continue reasoning
+        # This is especially important for models like Gemini that may stop without generating output
+        # Use HumanMessage instead of SystemMessage as it's more compatible with all models
+        # The prompt allows the model to call more tools OR provide final answer, giving it flexibility
+        if has_tool_results:
+            synthesis_prompt = HumanMessage(
+                content="Review the tool results above. If you need more information, you can call additional tools. "
+                "Otherwise, provide your complete final answer synthesizing all the information."
+            )
+            messages = messages + [synthesis_prompt]
+        
         response = self.model.invoke(messages)
         return {"messages": [response]}