cumulative upgrade - context + safety + response length v3

src/agents/synthesis_agent.py

@@ -267,13 +267,36 @@ class EnhancedSynthesisAgent:
         if not isinstance(user_input, str):
             user_input = str(user_input) if user_input else ''
         
-        # Default agent_outputs to empty list
+        # Default agent_outputs to empty list and normalize format
         if agent_outputs is None:
             agent_outputs = []
         
-            logger.info(f"{self.agent_id} synthesizing {len(agent_outputs)} agent outputs")
-            if context:
-                logger.info(f"{self.agent_id} context has {len(context.get('interactions', []))} interactions")
+        # Normalize agent_outputs: convert dict to list if needed
+        if isinstance(agent_outputs, dict):
+            # Convert dict {task_name: result} to list of dicts
+            normalized_outputs = []
+            for task_name, result in agent_outputs.items():
+                if isinstance(result, dict):
+                    # Add task name to the result dict for context
+                    result_with_task = result.copy()
+                    result_with_task['task_name'] = task_name
+                    normalized_outputs.append(result_with_task)
+                else:
+                    # Wrap non-dict results
+                    normalized_outputs.append({
+                        'task_name': task_name,
+                        'content': str(result),
+                        'result': str(result)
+                    })
+            agent_outputs = normalized_outputs
+        
+        # Ensure it's a list
+        if not isinstance(agent_outputs, list):
+            agent_outputs = [agent_outputs] if agent_outputs else []
+        
+        logger.info(f"{self.agent_id} synthesizing {len(agent_outputs)} agent outputs")
+        if context:
+            logger.info(f"{self.agent_id} context has {len(context.get('interactions', []))} interactions")
             
         # STEP 1: Extract skill probabilities from skills_result
         skill_probabilities = self._extract_skill_probabilities(skills_result)
@@ -459,9 +482,36 @@ RESPONSE GUIDELINES:
         # Build agent outputs section if any
         agent_outputs_section = ""
         if agent_outputs:
-            agent_outputs_section = f"\n\nAgent Analysis Results:\n"
-            for i, output in enumerate(agent_outputs, 1):
-                agent_outputs_section += f"Agent {i}: {output.get('result', output.get('final_response', str(output)))}\n"
+            # Handle both dict and list formats
+            if isinstance(agent_outputs, dict):
+                # Convert dict to list format
+                outputs_list = []
+                for task_name, result in agent_outputs.items():
+                    if isinstance(result, dict):
+                        outputs_list.append(result)
+                    else:
+                        # Wrap string/non-dict results in dict format
+                        outputs_list.append({
+                            'task': task_name,
+                            'content': str(result),
+                            'result': str(result)
+                        })
+                agent_outputs = outputs_list
+            
+            # Ensure it's a list now
+            if isinstance(agent_outputs, list):
+                agent_outputs_section = f"\n\nAgent Analysis Results:\n"
+                for i, output in enumerate(agent_outputs, 1):
+                    # Handle both dict and string outputs
+                    if isinstance(output, dict):
+                        output_text = output.get('content') or output.get('result') or output.get('final_response') or str(output)
+                    else:
+                        # If output is a string or other type
+                        output_text = str(output)
+                    agent_outputs_section += f"Agent {i}: {output_text}\n"
+            else:
+                # Fallback for unexpected types
+                agent_outputs_section = f"\n\nAgent Analysis Results:\n{str(agent_outputs)}\n"
         
         # Construct full prompt
         prompt = f"""{expert_preamble}