jaf-py 2.5.10__py3-none-any.whl → 2.5.11__py3-none-any.whl

jaf/core/engine.py

@@ -78,12 +78,12 @@ from .guardrails import (
 def to_event_data(value: Any) -> Any:
     """
     Resilient serializer helper for event payloads.
-    
+
     Converts various types to event-compatible data:
     - dataclasses: uses asdict()
     - Pydantic BaseModel: uses model_dump()
     - other types: returns as-is
-    
+
     This prevents TypeError when serializing nested Pydantic models or non-dataclass types.
     """
     if is_dataclass(value):
@@ -94,12 +94,12 @@ def to_event_data(value: Any) -> Any:
         return value
 
 
-Ctx = TypeVar('Ctx')
-Out = TypeVar('Out')
+Ctx = TypeVar("Ctx")
+Out = TypeVar("Out")
+
 
 async def try_resume_pending_tool_calls(
-    state: RunState[Ctx],
-    config: RunConfig[Ctx]
+    state: RunState[Ctx], config: RunConfig[Ctx]
 ) -> Optional[RunResult[Out]]:
     """
     Try to resume pending tool calls if the last assistant message contained tool_calls
@@ -110,157 +110,192 @@ async def try_resume_pending_tool_calls(
         for i in range(len(messages) - 1, -1, -1):
             msg = messages[i]
             # Handle both string and enum roles
-            role_str = msg.role.value if hasattr(msg.role, 'value') else str(msg.role)
-            if role_str == 'assistant' and msg.tool_calls:
+            role_str = msg.role.value if hasattr(msg.role, "value") else str(msg.role)
+            if role_str == "assistant" and msg.tool_calls:
                 tool_call_ids = {tc.id for tc in msg.tool_calls}
-                
+
                 # Scan forward for tool results tied to these ids
                 executed_ids = set()
                 for j in range(i + 1, len(messages)):
                     m = messages[j]
                     # Handle both string and enum roles
-                    m_role_str = m.role.value if hasattr(m.role, 'value') else str(m.role)
-                    if m_role_str == 'tool' and m.tool_call_id and m.tool_call_id in tool_call_ids:
+                    m_role_str = m.role.value if hasattr(m.role, "value") else str(m.role)
+                    if m_role_str == "tool" and m.tool_call_id and m.tool_call_id in tool_call_ids:
                         executed_ids.add(m.tool_call_id)
-                
+
                 pending_tool_calls = [tc for tc in msg.tool_calls if tc.id not in executed_ids]
-                
+
                 if not pending_tool_calls:
                     continue  # Continue checking other assistant messages
-                
+
                 current_agent = config.agent_registry.get(state.current_agent_name)
                 if not current_agent:
                     return RunResult(
                         final_state=state,
-                        outcome=ErrorOutcome(error=AgentNotFound(agent_name=state.current_agent_name))
+                        outcome=ErrorOutcome(
+                            error=AgentNotFound(agent_name=state.current_agent_name)
+                        ),
                     )
-                
+
                 # Execute pending tool calls
                 tool_results = await _execute_tool_calls(
-                    pending_tool_calls,
-                    current_agent,
-                    state,
-                    config
+                    pending_tool_calls, current_agent, state, config
                 )
-                
+
                 # Check for interruptions
-                interruptions = [r.get('interruption') for r in tool_results if r.get('interruption')]
+                interruptions = [
+                    r.get("interruption") for r in tool_results if r.get("interruption")
+                ]
                 if interruptions:
-                    completed_results = [r for r in tool_results if not r.get('interruption')]
+                    completed_results = [r for r in tool_results if not r.get("interruption")]
                     interrupted_state = replace(
                         state,
-                        messages=list(state.messages) + [r['message'] for r in completed_results],
+                        messages=list(state.messages) + [r["message"] for r in completed_results],
                         turn_count=state.turn_count,
-                        approvals=state.approvals
+                        approvals=state.approvals,
                     )
                     return RunResult(
                         final_state=interrupted_state,
-                        outcome=InterruptedOutcome(interruptions=interruptions)
+                        outcome=InterruptedOutcome(interruptions=interruptions),
                     )
-                
+
                 # Continue with normal execution
                 next_state = replace(
                     state,
-                    messages=list(state.messages) + [r['message'] for r in tool_results],
+                    messages=list(state.messages) + [r["message"] for r in tool_results],
                     turn_count=state.turn_count,
-                    approvals=state.approvals
+                    approvals=state.approvals,
                 )
                 return await _run_internal(next_state, config)
-        
+
     except Exception as e:
         # Best-effort resume; ignore and continue normal flow
         pass
-    
+
     return None
 
-async def run(
-    initial_state: RunState[Ctx],
-    config: RunConfig[Ctx]
-) -> RunResult[Out]:
+
+async def run(initial_state: RunState[Ctx], config: RunConfig[Ctx]) -> RunResult[Out]:
     """
     Main execution function for running agents.
     """
     try:
         # Set the current RunConfig in context for agent tools
         from .agent_tool import set_current_run_config
+
         set_current_run_config(config)
-        
+
         state_with_memory = await _load_conversation_history(initial_state, config)
-        
+
         # Emit RunStartEvent AFTER loading conversation history so we have complete context
         if config.on_event:
-            config.on_event(RunStartEvent(data=to_event_data(RunStartEventData(
-                run_id=initial_state.run_id,
-                trace_id=initial_state.trace_id,
-                session_id=config.conversation_id,
-                context=state_with_memory.context,
-                messages=state_with_memory.messages,  # Now includes full conversation history
-                agent_name=state_with_memory.current_agent_name
-            ))))
-        
+            config.on_event(
+                RunStartEvent(
+                    data=to_event_data(
+                        RunStartEventData(
+                            run_id=initial_state.run_id,
+                            trace_id=initial_state.trace_id,
+                            session_id=config.conversation_id,
+                            context=state_with_memory.context,
+                            messages=state_with_memory.messages,  # Now includes full conversation history
+                            agent_name=state_with_memory.current_agent_name,
+                        )
+                    )
+                )
+            )
+
         # Load approvals from storage if configured
         if config.approval_storage:
-            print(f'[JAF:ENGINE] Loading approvals for runId {state_with_memory.run_id}')
+            print(f"[JAF:ENGINE] Loading approvals for runId {state_with_memory.run_id}")
             from .state import load_approvals_into_state
+
             state_with_memory = await load_approvals_into_state(state_with_memory, config)
-        
+
         result = await _run_internal(state_with_memory, config)
 
         # Store conversation history only if this is a final completion of the entire conversation
         # For HITL scenarios, storage happens on interruption to allow resumption
         # We only store on completion if explicitly indicated this is the end of the conversation
-        if (config.memory and config.memory.auto_store and config.conversation_id and 
-            result.outcome.status == 'completed' and getattr(config.memory, 'store_on_completion', True)):
-            print(f'[JAF:ENGINE] Storing final completed conversation for {config.conversation_id}')
+        if (
+            config.memory
+            and config.memory.auto_store
+            and config.conversation_id
+            and result.outcome.status == "completed"
+            and getattr(config.memory, "store_on_completion", True)
+        ):
+            print(f"[JAF:ENGINE] Storing final completed conversation for {config.conversation_id}")
             await _store_conversation_history(result.final_state, config)
-        elif result.outcome.status == 'interrupted':
-            print('[JAF:ENGINE] Conversation interrupted - storage already handled during interruption')
+        elif result.outcome.status == "interrupted":
+            print(
+                "[JAF:ENGINE] Conversation interrupted - storage already handled during interruption"
+            )
         else:
-            print(f'[JAF:ENGINE] Skipping memory store - status: {result.outcome.status}, store_on_completion: {getattr(config.memory, "store_on_completion", True) if config.memory else "N/A"}')
+            print(
+                f"[JAF:ENGINE] Skipping memory store - status: {result.outcome.status}, store_on_completion: {getattr(config.memory, 'store_on_completion', True) if config.memory else 'N/A'}"
+            )
 
         if config.on_event:
-            config.on_event(RunEndEvent(data=to_event_data(RunEndEventData(
-                outcome=result.outcome,
-                trace_id=initial_state.trace_id,
-                run_id=initial_state.run_id
-            ))))
+            config.on_event(
+                RunEndEvent(
+                    data=to_event_data(
+                        RunEndEventData(
+                            outcome=result.outcome,
+                            trace_id=initial_state.trace_id,
+                            run_id=initial_state.run_id,
+                        )
+                    )
+                )
+            )
 
         return result
     except Exception as error:
         error_result = RunResult(
             final_state=initial_state,
-            outcome=ErrorOutcome(error=ModelBehaviorError(detail=str(error)))
+            outcome=ErrorOutcome(error=ModelBehaviorError(detail=str(error))),
         )
         if config.on_event:
-            config.on_event(RunEndEvent(data=to_event_data(RunEndEventData(
-                outcome=error_result.outcome,
-                trace_id=initial_state.trace_id,
-                run_id=initial_state.run_id
-            ))))
+            config.on_event(
+                RunEndEvent(
+                    data=to_event_data(
+                        RunEndEventData(
+                            outcome=error_result.outcome,
+                            trace_id=initial_state.trace_id,
+                            run_id=initial_state.run_id,
+                        )
+                    )
+                )
+            )
         return error_result
 
+
 async def _load_conversation_history(state: RunState[Ctx], config: RunConfig[Ctx]) -> RunState[Ctx]:
     """Load conversation history from memory provider."""
     if not (config.memory and config.memory.provider and config.conversation_id):
         return state
 
     if config.on_event:
-        config.on_event(MemoryEvent(data=MemoryEventData(
-            operation='load',
-            conversation_id=config.conversation_id,
-            status='start'
-        )))
+        config.on_event(
+            MemoryEvent(
+                data=MemoryEventData(
+                    operation="load", conversation_id=config.conversation_id, status="start"
+                )
+            )
+        )
 
     result = await config.memory.provider.get_conversation(config.conversation_id)
     if isinstance(result, Failure):
         print(f"[JAF:ENGINE] Warning: Failed to load conversation: {result.error}")
         if config.on_event:
-            config.on_event(MemoryEvent(data=MemoryEventData(
-                operation='load',
-                conversation_id=config.conversation_id,
-                status='fail',
-                error=str(result.error)
-            )))
+            config.on_event(
+                MemoryEvent(
+                    data=MemoryEventData(
+                        operation="load",
+                        conversation_id=config.conversation_id,
+                        status="fail",
+                        error=str(result.error),
+                    )
+                )
+            )
         return state
 
     conversation_data = result.data
@@ -271,17 +306,17 @@ async def _load_conversation_history(state: RunState[Ctx], config: RunConfig[Ctx
         # Filter out halted messages - they're for audit/database only, not for LLM context
         memory_messages = []
         filtered_count = 0
-        
+
         for msg in all_memory_messages:
-            if msg.role not in (ContentRole.TOOL, 'tool'):
+            if msg.role not in (ContentRole.TOOL, "tool"):
                 memory_messages.append(msg)
             else:
                 try:
                     content = json.loads(msg.content)
-                    status = content.get('status')
-                    hitl_status = content.get('hitl_status')
+                    status = content.get("status")
+                    hitl_status = content.get("hitl_status")
                     # Filter out ALL halted/pending approval messages (they're for audit only)
-                    if status == 'halted' or hitl_status == 'pending_approval':
+                    if status == "halted" or hitl_status == "pending_approval":
                         filtered_count += 1
                         continue  # Skip this halted message
                     else:
@@ -301,8 +336,12 @@ async def _load_conversation_history(state: RunState[Ctx], config: RunConfig[Ctx
         approvals_map = state.approvals
 
         # Calculate turn count efficiently
-        memory_assistant_count = sum(1 for msg in memory_messages if msg.role in (ContentRole.ASSISTANT, 'assistant'))
-        current_assistant_count = sum(1 for msg in state.messages if msg.role in (ContentRole.ASSISTANT, 'assistant'))
+        memory_assistant_count = sum(
+            1 for msg in memory_messages if msg.role in (ContentRole.ASSISTANT, "assistant")
+        )
+        current_assistant_count = sum(
+            1 for msg in state.messages if msg.role in (ContentRole.ASSISTANT, "assistant")
+        )
         calculated_turn_count = memory_assistant_count + current_assistant_count
 
         # Use metadata turn_count if available, otherwise calculate from messages
@@ -312,40 +351,54 @@ async def _load_conversation_history(state: RunState[Ctx], config: RunConfig[Ctx
             turn_count = max(metadata_turn_count, calculated_turn_count)
 
         if config.on_event:
-            config.on_event(MemoryEvent(data=MemoryEventData(
-                operation='load',
-                conversation_id=config.conversation_id,
-                status='end',
-                message_count=len(memory_messages)
-            )))
+            config.on_event(
+                MemoryEvent(
+                    data=MemoryEventData(
+                        operation="load",
+                        conversation_id=config.conversation_id,
+                        status="end",
+                        message_count=len(memory_messages),
+                    )
+                )
+            )
 
         if filtered_count > 0:
-            print(f'[JAF:MEMORY] Loaded {len(all_memory_messages)} messages from memory, filtered to {len(memory_messages)} for LLM context (removed {filtered_count} halted messages)')
+            print(
+                f"[JAF:MEMORY] Loaded {len(all_memory_messages)} messages from memory, filtered to {len(memory_messages)} for LLM context (removed {filtered_count} halted messages)"
+            )
         else:
-            print(f'[JAF:MEMORY] Loaded {len(all_memory_messages)} messages from memory')
+            print(f"[JAF:MEMORY] Loaded {len(all_memory_messages)} messages from memory")
 
         return replace(
-            state,
-            messages=combined_messages,
-            turn_count=turn_count,
-            approvals=approvals_map
+            state, messages=combined_messages, turn_count=turn_count, approvals=approvals_map
         )
     return state
 
+
 async def _store_conversation_history(state: RunState[Ctx], config: RunConfig[Ctx]):
     """Store conversation history to memory provider."""
-    if not (config.memory and config.memory.provider and config.conversation_id and config.memory.auto_store):
+    if not (
+        config.memory
+        and config.memory.provider
+        and config.conversation_id
+        and config.memory.auto_store
+    ):
         return
 
     if config.on_event:
-        config.on_event(MemoryEvent(data=MemoryEventData(
-            operation='store',
-            conversation_id=config.conversation_id,
-            status='start'
-        )))
+        config.on_event(
+            MemoryEvent(
+                data=MemoryEventData(
+                    operation="store", conversation_id=config.conversation_id, status="start"
+                )
+            )
+        )
 
     messages_to_store = list(state.messages)
-    if config.memory.compression_threshold and len(messages_to_store) > config.memory.compression_threshold:
+    if (
+        config.memory.compression_threshold
+        and len(messages_to_store) > config.memory.compression_threshold
+    ):
         keep_first = int(config.memory.compression_threshold * 0.2)
         keep_recent = config.memory.compression_threshold - keep_first
         messages_to_store = messages_to_store[:keep_first] + messages_to_store[-keep_recent:]
@@ -355,60 +408,72 @@ async def _store_conversation_history(state: RunState[Ctx], config: RunConfig[Ct
     if state.approvals:
         approval_metadata = {
             "approval_count": len(state.approvals),
-            "approved_tools": [tool_id for tool_id, approval in state.approvals.items() if approval.approved],
-            "rejected_tools": [tool_id for tool_id, approval in state.approvals.items() if not approval.approved],
-            "has_approvals": True
+            "approved_tools": [
+                tool_id for tool_id, approval in state.approvals.items() if approval.approved
+            ],
+            "rejected_tools": [
+                tool_id for tool_id, approval in state.approvals.items() if not approval.approved
+            ],
+            "has_approvals": True,
         }
-    
+
     metadata = {
-        "user_id": getattr(state.context, 'user_id', None),
+        "user_id": getattr(state.context, "user_id", None),
         "trace_id": str(state.trace_id),
         "run_id": str(state.run_id),
         "agent_name": state.current_agent_name,
         "turn_count": state.turn_count,
-        **approval_metadata
+        **approval_metadata,
     }
 
-    result = await config.memory.provider.store_messages(config.conversation_id, messages_to_store, metadata)
+    result = await config.memory.provider.store_messages(
+        config.conversation_id, messages_to_store, metadata
+    )
 
     if isinstance(result, Failure):
         print(f"[JAF:ENGINE] Warning: Failed to store conversation: {result.error}")
         if config.on_event:
-            config.on_event(MemoryEvent(data=MemoryEventData(
-                operation='store',
-                conversation_id=config.conversation_id,
-                status='fail',
-                error=str(result.error)
-            )))
+            config.on_event(
+                MemoryEvent(
+                    data=MemoryEventData(
+                        operation="store",
+                        conversation_id=config.conversation_id,
+                        status="fail",
+                        error=str(result.error),
+                    )
+                )
+            )
     else:
-        print(f"[JAF:ENGINE] Stored {len(messages_to_store)} messages for conversation {config.conversation_id}")
+        print(
+            f"[JAF:ENGINE] Stored {len(messages_to_store)} messages for conversation {config.conversation_id}"
+        )
         if config.on_event:
-            config.on_event(MemoryEvent(data=MemoryEventData(
-                operation='store',
-                conversation_id=config.conversation_id,
-                status='end',
-                message_count=len(messages_to_store)
-            )))
+            config.on_event(
+                MemoryEvent(
+                    data=MemoryEventData(
+                        operation="store",
+                        conversation_id=config.conversation_id,
+                        status="end",
+                        message_count=len(messages_to_store),
+                    )
+                )
+            )
 
     # Removed verbose logging for performance
 
 
-async def _run_internal(
-    state: RunState[Ctx],
-    config: RunConfig[Ctx]
-) -> RunResult[Out]:
+async def _run_internal(state: RunState[Ctx], config: RunConfig[Ctx]) -> RunResult[Out]:
     """Internal run function with recursive execution logic."""
     # Try to resume pending tool calls first
     resumed = await try_resume_pending_tool_calls(state, config)
     if resumed:
         return resumed
-    
+
     # Check max turns
     max_turns = config.max_turns or 50
     if state.turn_count >= max_turns:
         return RunResult(
-            final_state=state,
-            outcome=ErrorOutcome(error=MaxTurnsExceeded(turns=state.turn_count))
+            final_state=state, outcome=ErrorOutcome(error=MaxTurnsExceeded(turns=state.turn_count))
         )
 
     # Get current agent
@@ -416,30 +481,35 @@ async def _run_internal(
     if not current_agent:
         return RunResult(
             final_state=state,
-            outcome=ErrorOutcome(error=AgentNotFound(agent_name=state.current_agent_name))
+            outcome=ErrorOutcome(error=AgentNotFound(agent_name=state.current_agent_name)),
         )
 
     # Determine if agent has advanced guardrails configuration
     has_advanced_guardrails = bool(
-        current_agent.advanced_config and 
-        current_agent.advanced_config.guardrails and
-        (current_agent.advanced_config.guardrails.input_prompt or 
-         current_agent.advanced_config.guardrails.output_prompt or 
-         current_agent.advanced_config.guardrails.require_citations)
+        current_agent.advanced_config
+        and current_agent.advanced_config.guardrails
+        and (
+            current_agent.advanced_config.guardrails.input_prompt
+            or current_agent.advanced_config.guardrails.output_prompt
+            or current_agent.advanced_config.guardrails.require_citations
+        )
+    )
+
+    print(
+        "[JAF:ENGINE] Debug guardrails setup:",
+        {
+            "agent_name": current_agent.name,
+            "has_advanced_config": bool(current_agent.advanced_config),
+            "has_advanced_guardrails": has_advanced_guardrails,
+            "initial_input_guardrails": len(config.initial_input_guardrails or []),
+            "final_output_guardrails": len(config.final_output_guardrails or []),
+        },
     )
-    
-    print('[JAF:ENGINE] Debug guardrails setup:', {
-        'agent_name': current_agent.name,
-        'has_advanced_config': bool(current_agent.advanced_config),
-        'has_advanced_guardrails': has_advanced_guardrails,
-        'initial_input_guardrails': len(config.initial_input_guardrails or []),
-        'final_output_guardrails': len(config.final_output_guardrails or [])
-    })
 
     # Build effective guardrails
     effective_input_guardrails: List[Guardrail] = []
     effective_output_guardrails: List[Guardrail] = []
-    
+
     if has_advanced_guardrails:
         result = await build_effective_guardrails(current_agent, config)
         effective_input_guardrails, effective_output_guardrails = result
@@ -448,35 +518,48 @@ async def _run_internal(
         effective_output_guardrails = list(config.final_output_guardrails or [])
 
     # Execute input guardrails on first turn
-    input_guardrails_to_run = (effective_input_guardrails 
-                              if state.turn_count == 0 and effective_input_guardrails 
-                              else [])
-    
-    print('[JAF:ENGINE] Input guardrails to run:', {
-        'turn_count': state.turn_count,
-        'effective_input_length': len(effective_input_guardrails),
-        'input_guardrails_to_run_length': len(input_guardrails_to_run),
-        'has_advanced_guardrails': has_advanced_guardrails
-    })
+    input_guardrails_to_run = (
+        effective_input_guardrails if state.turn_count == 0 and effective_input_guardrails else []
+    )
+
+    print(
+        "[JAF:ENGINE] Input guardrails to run:",
+        {
+            "turn_count": state.turn_count,
+            "effective_input_length": len(effective_input_guardrails),
+            "input_guardrails_to_run_length": len(input_guardrails_to_run),
+            "has_advanced_guardrails": has_advanced_guardrails,
+        },
+    )
 
     if input_guardrails_to_run and state.turn_count == 0:
-        first_user_message = next((m for m in state.messages if m.role == ContentRole.USER or m.role == 'user'), None)
+        first_user_message = next(
+            (m for m in state.messages if m.role == ContentRole.USER or m.role == "user"), None
+        )
         if first_user_message:
             if has_advanced_guardrails:
-                execution_mode = (current_agent.advanced_config.guardrails.execution_mode 
-                                if current_agent.advanced_config and current_agent.advanced_config.guardrails 
-                                else 'parallel')
-            
-                if execution_mode == 'sequential':
+                execution_mode = (
+                    current_agent.advanced_config.guardrails.execution_mode
+                    if current_agent.advanced_config and current_agent.advanced_config.guardrails
+                    else "parallel"
+                )
+
+                if execution_mode == "sequential":
                     guardrail_result = await execute_input_guardrails_sequential(
                         input_guardrails_to_run, first_user_message, config
                     )
                     if not guardrail_result.is_valid:
                         return RunResult(
                             final_state=state,
-                            outcome=ErrorOutcome(error=InputGuardrailTripwire(
-                                reason=getattr(guardrail_result, 'error_message', 'Input guardrail violation')
-                            ))
+                            outcome=ErrorOutcome(
+                                error=InputGuardrailTripwire(
+                                    reason=getattr(
+                                        guardrail_result,
+                                        "error_message",
+                                        "Input guardrail violation",
+                                    )
+                                )
+                            ),
                         )
                 else:
                     # Parallel execution with LLM call overlap
@@ -484,22 +567,40 @@ async def _run_internal(
                         input_guardrails_to_run, first_user_message, config
                     )
                     if not guardrail_result.is_valid:
-                        print(f"🚨 Input guardrail violation: {getattr(guardrail_result, 'error_message', 'Unknown violation')}")
+                        print(
+                            f"🚨 Input guardrail violation: {getattr(guardrail_result, 'error_message', 'Unknown violation')}"
+                        )
                         return RunResult(
                             final_state=state,
-                            outcome=ErrorOutcome(error=InputGuardrailTripwire(
-                                reason=getattr(guardrail_result, 'error_message', 'Input guardrail violation')
-                            ))
+                            outcome=ErrorOutcome(
+                                error=InputGuardrailTripwire(
+                                    reason=getattr(
+                                        guardrail_result,
+                                        "error_message",
+                                        "Input guardrail violation",
+                                    )
+                                )
+                            ),
                         )
             else:
                 # Legacy guardrails path
-                print('[JAF:ENGINE] Using LEGACY guardrails path with', len(input_guardrails_to_run), 'guardrails')
+                print(
+                    "[JAF:ENGINE] Using LEGACY guardrails path with",
+                    len(input_guardrails_to_run),
+                    "guardrails",
+                )
                 for guardrail in input_guardrails_to_run:
                     if config.on_event:
-                        config.on_event(GuardrailEvent(data=GuardrailEventData(
-                            guardrail_name=getattr(guardrail, '__name__', 'unknown_guardrail'),
-                            content=get_text_content(first_user_message.content)
-                        )))
+                        config.on_event(
+                            GuardrailEvent(
+                                data=GuardrailEventData(
+                                    guardrail_name=getattr(
+                                        guardrail, "__name__", "unknown_guardrail"
+                                    ),
+                                    content=get_text_content(first_user_message.content),
+                                )
+                            )
+                        )
                     if asyncio.iscoroutinefunction(guardrail):
                         result = await guardrail(get_text_content(first_user_message.content))
                     else:
@@ -507,24 +608,34 @@ async def _run_internal(
 
                     if not result.is_valid:
                         if config.on_event:
-                            config.on_event(GuardrailViolationEvent(data=GuardrailViolationEventData(
-                                stage='input',
-                                reason=getattr(result, 'error_message', 'Input guardrail failed')
-                            )))
+                            config.on_event(
+                                GuardrailViolationEvent(
+                                    data=GuardrailViolationEventData(
+                                        stage="input",
+                                        reason=getattr(
+                                            result, "error_message", "Input guardrail failed"
+                                        ),
+                                    )
+                                )
+                            )
                         return RunResult(
                             final_state=state,
-                            outcome=ErrorOutcome(error=InputGuardrailTripwire(
-                                reason=getattr(result, 'error_message', 'Input guardrail failed')
-                            ))
+                            outcome=ErrorOutcome(
+                                error=InputGuardrailTripwire(
+                                    reason=getattr(
+                                        result, "error_message", "Input guardrail failed"
+                                    )
+                                )
+                            ),
                         )
 
     # Agent debugging logs removed for performance
 
     # Get model name
     model = (
-        config.model_override or
-        (current_agent.model_config.name if current_agent.model_config else None) or
-        "gpt-4o"
+        config.model_override
+        or (current_agent.model_config.name if current_agent.model_config else None)
+        or "gpt-4o"
     )
 
     # Apply before_llm_call callback if provided
@@ -540,24 +651,30 @@ async def _run_internal(
 
     # Emit LLM call start event
     if config.on_event:
-        config.on_event(LLMCallStartEvent(data=to_event_data(LLMCallStartEventData(
-            agent_name=current_agent.name,
-            model=model,
-            trace_id=state.trace_id,
-            run_id=state.run_id,
-            context=state.context,
-            messages=state.messages
-        ))))
+        config.on_event(
+            LLMCallStartEvent(
+                data=to_event_data(
+                    LLMCallStartEventData(
+                        agent_name=current_agent.name,
+                        model=model,
+                        trace_id=state.trace_id,
+                        run_id=state.run_id,
+                        context=state.context,
+                        messages=state.messages,
+                    )
+                )
+            )
+        )
 
     # Retry logic for empty LLM responses
     llm_response: Dict[str, Any]
     assistant_event_streamed = False
-    
+
     for retry_attempt in range(config.max_empty_response_retries + 1):
         # Get completion from model provider
         # Check if streaming should be used based on configuration and availability
         get_stream = getattr(config.model_provider, "get_completion_stream", None)
-        use_streaming = (config.prefer_streaming != False and callable(get_stream))
+        use_streaming = config.prefer_streaming != False and callable(get_stream)
 
         if use_streaming:
             try:
@@ -577,11 +694,13 @@ async def _run_internal(
                         idx = getattr(tcd, "index", 0) or 0
                         # Ensure slot exists
                         while len(partial_tool_calls) <= idx:
-                            partial_tool_calls.append({
-                                "id": None,
-                                "type": "function",
-                                "function": {"name": None, "arguments": ""}
-                            })
+                            partial_tool_calls.append(
+                                {
+                                    "id": None,
+                                    "type": "function",
+                                    "function": {"name": None, "arguments": ""},
+                                }
+                            )
                         target = partial_tool_calls[idx]
                         # id
                         tc_id = getattr(tcd, "id", None)
@@ -608,34 +727,45 @@ async def _run_internal(
                                 arguments = tc["function"]["arguments"]
                                 if isinstance(arguments, str):
                                     arguments = _normalize_tool_call_arguments(arguments)
-                                message_tool_calls.append({
-                                    "id": tc["id"] or f"call_{i}",
-                                    "type": "function",
-                                    "function": {
-                                        "name": tc["function"]["name"] or "",
-                                        "arguments": arguments
+                                message_tool_calls.append(
+                                    {
+                                        "id": tc["id"] or f"call_{i}",
+                                        "type": "function",
+                                        "function": {
+                                            "name": tc["function"]["name"] or "",
+                                            "arguments": arguments,
+                                        },
                                     }
-                                })
+                                )
 
                         partial_msg = Message(
                             role=ContentRole.ASSISTANT,
                             content=aggregated_text or "",
-                            tool_calls=None if not message_tool_calls else [
+                            tool_calls=None
+                            if not message_tool_calls
+                            else [
                                 ToolCall(
                                     id=mc["id"],
                                     type="function",
                                     function=ToolCallFunction(
                                         name=mc["function"]["name"],
-                                        arguments=_normalize_tool_call_arguments(mc["function"]["arguments"])
+                                        arguments=_normalize_tool_call_arguments(
+                                            mc["function"]["arguments"]
+                                        ),
                                     ),
-                                ) for mc in message_tool_calls
+                                )
+                                for mc in message_tool_calls
                             ],
                         )
                         try:
                             if config.on_event:
-                                config.on_event(AssistantMessageEvent(data=to_event_data(
-                                    AssistantMessageEventData(message=partial_msg)
-                                )))
+                                config.on_event(
+                                    AssistantMessageEvent(
+                                        data=to_event_data(
+                                            AssistantMessageEventData(message=partial_msg)
+                                        )
+                                    )
+                                )
                         except Exception as _e:
                             # Do not fail the run on callback errors
                             pass
@@ -648,50 +778,61 @@ async def _run_internal(
                         arguments = tc["function"]["arguments"]
                         if isinstance(arguments, str):
                             arguments = _normalize_tool_call_arguments(arguments)
-                        final_tool_calls.append({
-                            "id": tc["id"] or f"call_{i}",
-                            "type": "function",
-                            "function": {
-                                "name": tc["function"]["name"] or "",
-                                "arguments": arguments
+                        final_tool_calls.append(
+                            {
+                                "id": tc["id"] or f"call_{i}",
+                                "type": "function",
+                                "function": {
+                                    "name": tc["function"]["name"] or "",
+                                    "arguments": arguments,
+                                },
                             }
-                        })
+                        )
 
                 llm_response = {
-                    "message": {
-                        "content": aggregated_text or None,
-                        "tool_calls": final_tool_calls
-                    }
+                    "message": {"content": aggregated_text or None, "tool_calls": final_tool_calls}
                 }
             except Exception:
                 # Fallback to non-streaming on error
                 assistant_event_streamed = False
-                llm_response = await config.model_provider.get_completion(state, current_agent, config)
+                llm_response = await config.model_provider.get_completion(
+                    state, current_agent, config
+                )
         else:
             llm_response = await config.model_provider.get_completion(state, current_agent, config)
-        
+
         # Check if response has meaningful content
-        has_content = llm_response.get('message', {}).get('content')
-        has_tool_calls = llm_response.get('message', {}).get('tool_calls')
-        
+        has_content = llm_response.get("message", {}).get("content")
+        has_tool_calls = llm_response.get("message", {}).get("tool_calls")
+
         # If we got a valid response, break out of retry loop
         if has_content or has_tool_calls:
             break
-        
+
         # If this is not the last attempt, retry with exponential backoff
         if retry_attempt < config.max_empty_response_retries:
-            delay = config.empty_response_retry_delay * (2 ** retry_attempt)
+            delay = config.empty_response_retry_delay * (2**retry_attempt)
             if config.log_empty_responses:
-                print(f"[JAF:ENGINE] Empty LLM response on attempt {retry_attempt + 1}/{config.max_empty_response_retries + 1}, retrying in {delay:.1f}s...")
-                print(f"[JAF:ENGINE] Response had message: {bool(llm_response.get('message'))}, content: {bool(has_content)}, tool_calls: {bool(has_tool_calls)}")
+                print(
+                    f"[JAF:ENGINE] Empty LLM response on attempt {retry_attempt + 1}/{config.max_empty_response_retries + 1}, retrying in {delay:.1f}s..."
+                )
+                print(
+                    f"[JAF:ENGINE] Response had message: {bool(llm_response.get('message'))}, content: {bool(has_content)}, tool_calls: {bool(has_tool_calls)}"
+                )
             await asyncio.sleep(delay)
         else:
             # Last attempt failed, log detailed diagnostic info
             if config.log_empty_responses:
-                print(f"[JAF:ENGINE] Empty LLM response after {config.max_empty_response_retries + 1} attempts")
+                print(
+                    f"[JAF:ENGINE] Empty LLM response after {config.max_empty_response_retries + 1} attempts"
+                )
                 print(f"[JAF:ENGINE] Agent: {current_agent.name}, Model: {model}")
-                print(f"[JAF:ENGINE] Message count: {len(state.messages)}, Turn: {state.turn_count}")
-                print(f"[JAF:ENGINE] Response structure: {json.dumps(llm_response, indent=2)[:1000]}")
+                print(
+                    f"[JAF:ENGINE] Message count: {len(state.messages)}, Turn: {state.turn_count}"
+                )
+                print(
+                    f"[JAF:ENGINE] Response structure: {json.dumps(llm_response, indent=2)[:1000]}"
+                )
 
     # Apply after_llm_call callback if provided
     if config.after_llm_call:
@@ -706,30 +847,34 @@ async def _run_internal(
 
     # Emit LLM call end event
     if config.on_event:
-        config.on_event(LLMCallEndEvent(data=to_event_data(LLMCallEndEventData(
-            choice=llm_response,
-            trace_id=state.trace_id,
-            run_id=state.run_id,
-            usage=llm_response.get("usage")
-        ))))
+        config.on_event(
+            LLMCallEndEvent(
+                data=to_event_data(
+                    LLMCallEndEventData(
+                        choice=llm_response,
+                        trace_id=state.trace_id,
+                        run_id=state.run_id,
+                        usage=llm_response.get("usage"),
+                    )
+                )
+            )
+        )
 
     # Check if response has message
-    if not llm_response.get('message'):
+    if not llm_response.get("message"):
         if config.log_empty_responses:
             print(f"[JAF:ENGINE] ERROR: No message in LLM response")
             print(f"[JAF:ENGINE] Response structure: {json.dumps(llm_response, indent=2)[:500]}")
         return RunResult(
             final_state=state,
-            outcome=ErrorOutcome(error=ModelBehaviorError(
-                detail='No message in model response'
-            ))
+            outcome=ErrorOutcome(error=ModelBehaviorError(detail="No message in model response")),
         )
 
     # Create assistant message
     assistant_message = Message(
         role=ContentRole.ASSISTANT,
-        content=llm_response['message'].get('content') or '',
-        tool_calls=_convert_tool_calls(llm_response['message'].get('tool_calls'))
+        content=llm_response["message"].get("content") or "",
+        tool_calls=_convert_tool_calls(llm_response["message"].get("tool_calls")),
     )
 
     new_messages = list(state.messages) + [assistant_message]
@@ -737,83 +882,97 @@ async def _run_internal(
     # Handle tool calls
     if assistant_message.tool_calls:
         tool_results = await _execute_tool_calls(
-            assistant_message.tool_calls,
-            current_agent,
-            state,
-            config
+            assistant_message.tool_calls, current_agent, state, config
         )
 
         # Check for interruptions
-        interruptions = [r.get('interruption') for r in tool_results if r.get('interruption')]
+        interruptions = [r.get("interruption") for r in tool_results if r.get("interruption")]
         if interruptions:
             # Separate completed tool results from interrupted ones
-            completed_results = [r for r in tool_results if not r.get('interruption')]
-            approval_required_results = [r for r in tool_results if r.get('interruption')]
-            
+            completed_results = [r for r in tool_results if not r.get("interruption")]
+            approval_required_results = [r for r in tool_results if r.get("interruption")]
+
             # Add pending approvals to state.approvals
             updated_approvals = dict(state.approvals)
             for interruption in interruptions:
-                if interruption.type == 'tool_approval':
+                if interruption.type == "tool_approval":
                     updated_approvals[interruption.tool_call.id] = ApprovalValue(
-                        status='pending',
+                        status="pending",
                         approved=False,
-                        additional_context={'status': 'pending', 'timestamp': str(int(time.time() * 1000))}
+                        additional_context={
+                            "status": "pending",
+                            "timestamp": str(int(time.time() * 1000)),
+                        },
                     )
 
             # Create state with only completed tool results (for LLM context)
             interrupted_state = replace(
                 state,
-                messages=new_messages + [r['message'] for r in completed_results],
+                messages=new_messages + [r["message"] for r in completed_results],
                 turn_count=state.turn_count + 1,
-                approvals=updated_approvals
+                approvals=updated_approvals,
             )
-            
+
             # Store conversation state with ALL messages including approval-required (for database records)
             if config.memory and config.memory.auto_store and config.conversation_id:
-                print(f'[JAF:ENGINE] Storing conversation state due to interruption for {config.conversation_id}')
+                print(
+                    f"[JAF:ENGINE] Storing conversation state due to interruption for {config.conversation_id}"
+                )
                 state_for_storage = replace(
                     interrupted_state,
-                    messages=interrupted_state.messages + [r['message'] for r in approval_required_results]
+                    messages=interrupted_state.messages
+                    + [r["message"] for r in approval_required_results],
                 )
                 await _store_conversation_history(state_for_storage, config)
-            
+
             return RunResult(
                 final_state=interrupted_state,
-                outcome=InterruptedOutcome(interruptions=interruptions)
+                outcome=InterruptedOutcome(interruptions=interruptions),
             )
 
         # Check for handoffs
-        handoff_result = next((r for r in tool_results if r.get('is_handoff')), None)
+        handoff_result = next((r for r in tool_results if r.get("is_handoff")), None)
         if handoff_result:
-            target_agent = handoff_result['target_agent']
+            target_agent = handoff_result["target_agent"]
 
             # Validate handoff permission
             if not current_agent.handoffs or target_agent not in current_agent.handoffs:
                 return RunResult(
                     final_state=replace(state, messages=new_messages),
-                    outcome=ErrorOutcome(error=HandoffError(
-                        detail=f"Agent {current_agent.name} cannot handoff to {target_agent}"
-                    ))
+                    outcome=ErrorOutcome(
+                        error=HandoffError(
+                            detail=f"Agent {current_agent.name} cannot handoff to {target_agent}"
+                        )
+                    ),
                 )
 
             # Emit handoff event
             if config.on_event:
-                config.on_event(HandoffEvent(data=to_event_data(HandoffEventData(
-                    from_=current_agent.name,
-                    to=target_agent
-                ))))
+                config.on_event(
+                    HandoffEvent(
+                        data=to_event_data(
+                            HandoffEventData(from_=current_agent.name, to=target_agent)
+                        )
+                    )
+                )
 
             # Remove any halted messages that are being replaced by actual execution results
             cleaned_new_messages = []
             for msg in new_messages:
-                if msg.role not in (ContentRole.TOOL, 'tool'):
+                if msg.role not in (ContentRole.TOOL, "tool"):
                     cleaned_new_messages.append(msg)
                 else:
                     try:
                         content = json.loads(msg.content)
-                        if content.get('status') == 'halted' or content.get('hitl_status') == 'pending_approval':
+                        if (
+                            content.get("status") == "halted"
+                            or content.get("hitl_status") == "pending_approval"
+                        ):
                             # Remove this halted message if we have a new result for the same tool_call_id
-                            if not any(result['message'].tool_call_id == msg.tool_call_id for result in tool_results):
+                            if not any(
+                                result["message"].tool_call_id == msg.tool_call_id
+                                for result in tool_results
+                            ):
                                 cleaned_new_messages.append(msg)
                         else:
                             cleaned_new_messages.append(msg)
@@ -823,10 +982,10 @@ async def _run_internal(
             # Continue with new agent
             next_state = replace(
                 state,
-                messages=cleaned_new_messages + [r['message'] for r in tool_results],
+                messages=cleaned_new_messages + [r["message"] for r in tool_results],
                 current_agent_name=target_agent,
                 turn_count=state.turn_count + 1,
-                approvals=state.approvals
+                approvals=state.approvals,
             )
 
             return await _run_internal(next_state, config)
@@ -834,14 +993,20 @@ async def _run_internal(
         # Remove any halted messages that are being replaced by actual execution results
         cleaned_new_messages = []
         for msg in new_messages:
-            if msg.role not in (ContentRole.TOOL, 'tool'):
+            if msg.role not in (ContentRole.TOOL, "tool"):
                 cleaned_new_messages.append(msg)
             else:
                 try:
                     content = json.loads(msg.content)
-                    if content.get('status') == 'halted' or content.get('hitl_status') == 'pending_approval':
+                    if (
+                        content.get("status") == "halted"
+                        or content.get("hitl_status") == "pending_approval"
+                    ):
                         # Remove this halted message if we have a new result for the same tool_call_id
-                        if not any(result['message'].tool_call_id == msg.tool_call_id for result in tool_results):
+                        if not any(
+                            result["message"].tool_call_id == msg.tool_call_id
+                            for result in tool_results
+                        ):
                             cleaned_new_messages.append(msg)
                     else:
                         cleaned_new_messages.append(msg)
@@ -851,9 +1016,9 @@ async def _run_internal(
         # Continue with tool results
         next_state = replace(
             state,
-            messages=cleaned_new_messages + [r['message'] for r in tool_results],
+            messages=cleaned_new_messages + [r["message"] for r in tool_results],
             turn_count=state.turn_count + 1,
-            approvals=state.approvals
+            approvals=state.approvals,
         )
 
         return await _run_internal(next_state, config)
@@ -863,19 +1028,26 @@ async def _run_internal(
         if current_agent.output_codec:
             # Parse with output codec
             if config.on_event:
-                config.on_event(OutputParseEvent(data=OutputParseEventData(
-                    content=get_text_content(assistant_message.content),
-                    status='start'
-                )))
+                config.on_event(
+                    OutputParseEvent(
+                        data=OutputParseEventData(
+                            content=get_text_content(assistant_message.content), status="start"
+                        )
+                    )
+                )
             try:
                 parsed_content = _try_parse_json(get_text_content(assistant_message.content))
                 output_data = current_agent.output_codec.model_validate(parsed_content)
                 if config.on_event:
-                    config.on_event(OutputParseEvent(data=OutputParseEventData(
-                        content=get_text_content(assistant_message.content),
-                        status='end',
-                        parsed_output=output_data
-                    )))
+                    config.on_event(
+                        OutputParseEvent(
+                            data=OutputParseEventData(
+                                content=get_text_content(assistant_message.content),
+                                status="end",
+                                parsed_output=output_data,
+                            )
+                        )
+                    )
 
                 # Check final output guardrails
                 if has_advanced_guardrails:
@@ -886,19 +1058,31 @@ async def _run_internal(
                     if not output_guardrail_result.is_valid:
                         return RunResult(
                             final_state=replace(state, messages=new_messages),
-                            outcome=ErrorOutcome(error=OutputGuardrailTripwire(
-                                reason=getattr(output_guardrail_result, 'error_message', 'Output guardrail violation')
-                            ))
+                            outcome=ErrorOutcome(
+                                error=OutputGuardrailTripwire(
+                                    reason=getattr(
+                                        output_guardrail_result,
+                                        "error_message",
+                                        "Output guardrail violation",
+                                    )
+                                )
+                            ),
                         )
                 else:
                     # Legacy system
                     if effective_output_guardrails:
                         for guardrail in effective_output_guardrails:
                             if config.on_event:
-                                config.on_event(GuardrailEvent(data=GuardrailEventData(
-                                    guardrail_name=getattr(guardrail, '__name__', 'unknown_guardrail'),
-                                    content=output_data
-                                )))
+                                config.on_event(
+                                    GuardrailEvent(
+                                        data=GuardrailEventData(
+                                            guardrail_name=getattr(
+                                                guardrail, "__name__", "unknown_guardrail"
+                                            ),
+                                            content=output_data,
+                                        )
+                                    )
+                                )
                         if asyncio.iscoroutinefunction(guardrail):
                             result = await guardrail(output_data)
                         else:
@@ -906,34 +1090,55 @@ async def _run_internal(
 
                         if not result.is_valid:
                             if config.on_event:
-                                config.on_event(GuardrailViolationEvent(data=GuardrailViolationEventData(
-                                    stage='output',
-                                    reason=getattr(result, 'error_message', 'Output guardrail failed')
-                                )))
+                                config.on_event(
+                                    GuardrailViolationEvent(
+                                        data=GuardrailViolationEventData(
+                                            stage="output",
+                                            reason=getattr(
+                                                result, "error_message", "Output guardrail failed"
+                                            ),
+                                        )
+                                    )
+                                )
                             return RunResult(
-                                final_state=replace(state, messages=new_messages, approvals=state.approvals),
-                                outcome=ErrorOutcome(error=OutputGuardrailTripwire(
-                                    reason=getattr(result, 'error_message', 'Output guardrail failed')
-                                ))
+                                final_state=replace(
+                                    state, messages=new_messages, approvals=state.approvals
+                                ),
+                                outcome=ErrorOutcome(
+                                    error=OutputGuardrailTripwire(
+                                        reason=getattr(
+                                            result, "error_message", "Output guardrail failed"
+                                        )
+                                    )
+                                ),
                             )
 
                 return RunResult(
-                    final_state=replace(state, messages=new_messages, turn_count=state.turn_count + 1, approvals=state.approvals),
-                    outcome=CompletedOutcome(output=output_data)
+                    final_state=replace(
+                        state,
+                        messages=new_messages,
+                        turn_count=state.turn_count + 1,
+                        approvals=state.approvals,
+                    ),
+                    outcome=CompletedOutcome(output=output_data),
                 )
 
             except ValidationError as e:
                 if config.on_event:
-                    config.on_event(OutputParseEvent(data=OutputParseEventData(
-                        content=get_text_content(assistant_message.content),
-                        status='fail',
-                        error=str(e)
-                    )))
+                    config.on_event(
+                        OutputParseEvent(
+                            data=OutputParseEventData(
+                                content=get_text_content(assistant_message.content),
+                                status="fail",
+                                error=str(e),
+                            )
+                        )
+                    )
                 return RunResult(
                     final_state=replace(state, messages=new_messages, approvals=state.approvals),
-                    outcome=ErrorOutcome(error=DecodeError(
-                        errors=[{'message': str(e), 'details': e.errors()}]
-                    ))
+                    outcome=ErrorOutcome(
+                        error=DecodeError(errors=[{"message": str(e), "details": e.errors()}])
+                    ),
                 )
         else:
             # No output codec, return content as string
@@ -945,19 +1150,31 @@ async def _run_internal(
                 if not output_guardrail_result.is_valid:
                     return RunResult(
                         final_state=replace(state, messages=new_messages),
-                        outcome=ErrorOutcome(error=OutputGuardrailTripwire(
-                            reason=getattr(output_guardrail_result, 'error_message', 'Output guardrail violation')
-                        ))
+                        outcome=ErrorOutcome(
+                            error=OutputGuardrailTripwire(
+                                reason=getattr(
+                                    output_guardrail_result,
+                                    "error_message",
+                                    "Output guardrail violation",
+                                )
+                            )
+                        ),
                     )
             else:
                 # Legacy system
                 if effective_output_guardrails:
                     for guardrail in effective_output_guardrails:
                         if config.on_event:
-                            config.on_event(GuardrailEvent(data=GuardrailEventData(
-                                guardrail_name=getattr(guardrail, '__name__', 'unknown_guardrail'),
-                                content=get_text_content(assistant_message.content)
-                            )))
+                            config.on_event(
+                                GuardrailEvent(
+                                    data=GuardrailEventData(
+                                        guardrail_name=getattr(
+                                            guardrail, "__name__", "unknown_guardrail"
+                                        ),
+                                        content=get_text_content(assistant_message.content),
+                                    )
+                                )
+                            )
                         if asyncio.iscoroutinefunction(guardrail):
                             result = await guardrail(get_text_content(assistant_message.content))
                         else:
@@ -965,30 +1182,48 @@ async def _run_internal(
 
                         if not result.is_valid:
                             if config.on_event:
-                                config.on_event(GuardrailViolationEvent(data=GuardrailViolationEventData(
-                                    stage='output',
-                                    reason=getattr(result, 'error_message', 'Output guardrail failed')
-                                )))
+                                config.on_event(
+                                    GuardrailViolationEvent(
+                                        data=GuardrailViolationEventData(
+                                            stage="output",
+                                            reason=getattr(
+                                                result, "error_message", "Output guardrail failed"
+                                            ),
+                                        )
+                                    )
+                                )
                             return RunResult(
-                                final_state=replace(state, messages=new_messages, approvals=state.approvals),
-                                outcome=ErrorOutcome(error=OutputGuardrailTripwire(
-                                    reason=getattr(result, 'error_message', 'Output guardrail failed')
-                                ))
+                                final_state=replace(
+                                    state, messages=new_messages, approvals=state.approvals
+                                ),
+                                outcome=ErrorOutcome(
+                                    error=OutputGuardrailTripwire(
+                                        reason=getattr(
+                                            result, "error_message", "Output guardrail failed"
+                                        )
+                                    )
+                                ),
                             )
 
             return RunResult(
-                final_state=replace(state, messages=new_messages, turn_count=state.turn_count + 1, approvals=state.approvals),
-                outcome=CompletedOutcome(output=get_text_content(assistant_message.content))
+                final_state=replace(
+                    state,
+                    messages=new_messages,
+                    turn_count=state.turn_count + 1,
+                    approvals=state.approvals,
+                ),
+                outcome=CompletedOutcome(output=get_text_content(assistant_message.content)),
             )
 
     # Model produced neither content nor tool calls
     return RunResult(
         final_state=replace(state, messages=new_messages, approvals=state.approvals),
-        outcome=ErrorOutcome(error=ModelBehaviorError(
-            detail='Model produced neither content nor tool calls'
-        ))
+        outcome=ErrorOutcome(
+            error=ModelBehaviorError(detail="Model produced neither content nor tool calls")
+        ),
     )
 
+
 def _convert_tool_calls(tool_calls: Optional[List[Dict[str, Any]]]) -> Optional[List[ToolCall]]:
     """Convert API tool calls to internal format."""
     if not tool_calls:
@@ -996,12 +1231,12 @@ def _convert_tool_calls(tool_calls: Optional[List[Dict[str, Any]]]) -> Optional[
 
     return [
         ToolCall(
-            id=tc['id'],
-            type='function',
+            id=tc["id"],
+            type="function",
             function=ToolCallFunction(
-                name=tc['function']['name'],
-                arguments=_normalize_tool_call_arguments(tc['function']['arguments'])
-            )
+                name=tc["function"]["name"],
+                arguments=_normalize_tool_call_arguments(tc["function"]["arguments"]),
+            ),
         )
         for tc in tool_calls
     ]
@@ -1027,24 +1262,28 @@ def _normalize_tool_call_arguments(arguments: Any) -> Any:
 
     return arguments
 
+
 async def _execute_tool_calls(
-    tool_calls: List[ToolCall],
-    agent: Agent[Ctx, Any],
-    state: RunState[Ctx],
-    config: RunConfig[Ctx]
+    tool_calls: List[ToolCall], agent: Agent[Ctx, Any], state: RunState[Ctx], config: RunConfig[Ctx]
 ) -> List[Dict[str, Any]]:
     """Execute tool calls and return results."""
 
     async def execute_single_tool_call(tool_call: ToolCall) -> Dict[str, Any]:
-        print(f'[JAF:TOOL-EXEC] Starting execute_single_tool_call for {tool_call.function.name}')
+        print(f"[JAF:TOOL-EXEC] Starting execute_single_tool_call for {tool_call.function.name}")
         if config.on_event:
-            config.on_event(ToolCallStartEvent(data=to_event_data(ToolCallStartEventData(
-                tool_name=tool_call.function.name,
-                args=_try_parse_json(tool_call.function.arguments),
-                trace_id=state.trace_id,
-                run_id=state.run_id,
-                call_id=tool_call.id
-            ))))
+            config.on_event(
+                ToolCallStartEvent(
+                    data=to_event_data(
+                        ToolCallStartEventData(
+                            tool_name=tool_call.function.name,
+                            args=_try_parse_json(tool_call.function.arguments),
+                            trace_id=state.trace_id,
+                            run_id=state.run_id,
+                            call_id=tool_call.id,
+                        )
+                    )
+                )
+            )
 
         try:
             # Find the tool
@@ -1056,28 +1295,34 @@ async def _execute_tool_calls(
                         break
 
             if not tool:
-                error_result = json.dumps({
-                    'hitl_status': 'tool_not_found',  # HITL workflow status
-                    'message': f'Tool {tool_call.function.name} not found',
-                    'tool_name': tool_call.function.name,
-                })
+                error_result = json.dumps(
+                    {
+                        "hitl_status": "tool_not_found",  # HITL workflow status
+                        "message": f"Tool {tool_call.function.name} not found",
+                        "tool_name": tool_call.function.name,
+                    }
+                )
 
                 if config.on_event:
-                    config.on_event(ToolCallEndEvent(data=to_event_data(ToolCallEndEventData(
-                        tool_name=tool_call.function.name,
-                        result=error_result,
-                        trace_id=state.trace_id,
-                        run_id=state.run_id,
-                        execution_status='error',  # Tool execution failed
-                        tool_result={'error': 'tool_not_found'},
-                        call_id=tool_call.id
-                    ))))
+                    config.on_event(
+                        ToolCallEndEvent(
+                            data=to_event_data(
+                                ToolCallEndEventData(
+                                    tool_name=tool_call.function.name,
+                                    result=error_result,
+                                    trace_id=state.trace_id,
+                                    run_id=state.run_id,
+                                    execution_status="error",  # Tool execution failed
+                                    tool_result={"error": "tool_not_found"},
+                                    call_id=tool_call.id,
+                                )
+                            )
+                        )
+                    )
 
                 return {
-                    'message': Message(
-                        role=ContentRole.TOOL,
-                        content=error_result,
-                        tool_call_id=tool_call.id
+                    "message": Message(
+                        role=ContentRole.TOOL, content=error_result, tool_call_id=tool_call.id
                     )
                 }
 
@@ -1085,54 +1330,66 @@ async def _execute_tool_calls(
             raw_args = _try_parse_json(tool_call.function.arguments)
             try:
                 # Assuming the tool schema parameters is a Pydantic model
-                if hasattr(tool.schema.parameters, 'model_validate'):
+                if hasattr(tool.schema.parameters, "model_validate"):
                     validated_args = tool.schema.parameters.model_validate(raw_args)
                 else:
                     validated_args = raw_args
             except ValidationError as e:
-                error_result = json.dumps({
-                    'hitl_status': 'validation_error',  # HITL workflow status
-                    'message': f'Invalid arguments for {tool_call.function.name}: {e!s}',
-                    'tool_name': tool_call.function.name,
-                    'validation_errors': e.errors()
-                })
+                error_result = json.dumps(
+                    {
+                        "hitl_status": "validation_error",  # HITL workflow status
+                        "message": f"Invalid arguments for {tool_call.function.name}: {e!s}",
+                        "tool_name": tool_call.function.name,
+                        "validation_errors": e.errors(),
+                    }
+                )
 
                 if config.on_event:
-                    config.on_event(ToolCallEndEvent(data=to_event_data(ToolCallEndEventData(
-                        tool_name=tool_call.function.name,
-                        result=error_result,
-                        trace_id=state.trace_id,
-                        run_id=state.run_id,
-                        execution_status='error',  # Tool execution failed due to validation
-                        tool_result={'error': 'validation_error', 'details': e.errors()},
-                        call_id=tool_call.id
-                    ))))
+                    config.on_event(
+                        ToolCallEndEvent(
+                            data=to_event_data(
+                                ToolCallEndEventData(
+                                    tool_name=tool_call.function.name,
+                                    result=error_result,
+                                    trace_id=state.trace_id,
+                                    run_id=state.run_id,
+                                    execution_status="error",  # Tool execution failed due to validation
+                                    tool_result={
+                                        "error": "validation_error",
+                                        "details": e.errors(),
+                                    },
+                                    call_id=tool_call.id,
+                                )
+                            )
+                        )
+                    )
 
                 return {
-                    'message': Message(
-                        role=ContentRole.TOOL,
-                        content=error_result,
-                        tool_call_id=tool_call.id
+                    "message": Message(
+                        role=ContentRole.TOOL, content=error_result, tool_call_id=tool_call.id
                     )
                 }
 
             # Check if tool needs approval
             needs_approval = False
-            approval_func = getattr(tool, 'needs_approval', False)
+            approval_func = getattr(tool, "needs_approval", False)
             if callable(approval_func):
                 needs_approval = await approval_func(state.context, validated_args)
             else:
                 needs_approval = bool(approval_func)
-            
+
             # Check approval status - first by ID, then by signature for cross-session matching
             approval_status = state.approvals.get(tool_call.id)
             if not approval_status:
                 signature = f"{tool_call.function.name}:{tool_call.function.arguments}"
                 for _, approval in state.approvals.items():
-                    if approval.additional_context and approval.additional_context.get('signature') == signature:
+                    if (
+                        approval.additional_context
+                        and approval.additional_context.get("signature") == signature
+                    ):
                         approval_status = approval
                         break
-            
+
             derived_status = None
             if approval_status:
                 # Use explicit status if available
@@ -1140,76 +1397,93 @@ async def _execute_tool_calls(
                     derived_status = approval_status.status
                 # Fall back to approved boolean if status not set
                 elif approval_status.approved is True:
-                    derived_status = 'approved'
+                    derived_status = "approved"
                 elif approval_status.approved is False:
-                    if approval_status.additional_context and approval_status.additional_context.get('status') == 'pending':
-                        derived_status = 'pending'
+                    if (
+                        approval_status.additional_context
+                        and approval_status.additional_context.get("status") == "pending"
+                    ):
+                        derived_status = "pending"
                     else:
-                        derived_status = 'rejected'
+                        derived_status = "rejected"
 
-            is_pending = derived_status == 'pending'
+            is_pending = derived_status == "pending"
 
             # If approval needed and not yet decided, create interruption
             if needs_approval and (approval_status is None or is_pending):
                 interruption = ToolApprovalInterruption(
-                    type='tool_approval',
+                    type="tool_approval",
                     tool_call=tool_call,
                     agent=agent,
-                    session_id=str(state.run_id)
+                    session_id=str(state.run_id),
                 )
-                
+
                 # Return interrupted result with halted message
-                halted_result = json.dumps({
-                    'hitl_status': 'pending_approval',  # HITL workflow status: waiting for approval
-                    'message': f'Tool {tool_call.function.name} requires approval.',
-                })
-                
+                halted_result = json.dumps(
+                    {
+                        "hitl_status": "pending_approval",  # HITL workflow status: waiting for approval
+                        "message": f"Tool {tool_call.function.name} requires approval.",
+                    }
+                )
+
                 return {
-                    'message': Message(
-                        role=ContentRole.TOOL,
-                        content=halted_result,
-                        tool_call_id=tool_call.id
+                    "message": Message(
+                        role=ContentRole.TOOL, content=halted_result, tool_call_id=tool_call.id
                     ),
-                    'interruption': interruption
+                    "interruption": interruption,
                 }
 
             # If approval was explicitly rejected, return rejection message
-            if derived_status == 'rejected':
-                rejection_reason = approval_status.additional_context.get('rejection_reason', 'User declined the action') if approval_status.additional_context else 'User declined the action'
-                rejection_result = json.dumps({
-                    'hitl_status': 'rejected',  # HITL workflow status: user rejected the action
-                    'message': f'Action was not approved. {rejection_reason}. Please ask if you can help with something else or suggest an alternative approach.',
-                    'tool_name': tool_call.function.name,
-                    'rejection_reason': rejection_reason,
-                    'additional_context': approval_status.additional_context if approval_status else None
-                })
-                
+            if derived_status == "rejected":
+                rejection_reason = (
+                    approval_status.additional_context.get(
+                        "rejection_reason", "User declined the action"
+                    )
+                    if approval_status.additional_context
+                    else "User declined the action"
+                )
+                rejection_result = json.dumps(
+                    {
+                        "hitl_status": "rejected",  # HITL workflow status: user rejected the action
+                        "message": f"Action was not approved. {rejection_reason}. Please ask if you can help with something else or suggest an alternative approach.",
+                        "tool_name": tool_call.function.name,
+                        "rejection_reason": rejection_reason,
+                        "additional_context": approval_status.additional_context
+                        if approval_status
+                        else None,
+                    }
+                )
+
                 return {
-                    'message': Message(
-                        role=ContentRole.TOOL,
-                        content=rejection_result,
-                        tool_call_id=tool_call.id
+                    "message": Message(
+                        role=ContentRole.TOOL, content=rejection_result, tool_call_id=tool_call.id
                     )
                 }
 
             # Determine timeout for this tool
             # Priority: tool-specific timeout > RunConfig default > 30 seconds global default
-            if tool and hasattr(tool, 'schema'):
-                timeout = getattr(tool.schema, 'timeout', None)
+            if tool and hasattr(tool, "schema"):
+                timeout = getattr(tool.schema, "timeout", None)
             else:
                 timeout = None
             if timeout is None:
-                timeout = config.default_tool_timeout if config.default_tool_timeout is not None else 300.0
+                timeout = (
+                    config.default_tool_timeout
+                    if config.default_tool_timeout is not None
+                    else 300.0
+                )
 
             # Merge additional context if provided through approval
             additional_context = approval_status.additional_context if approval_status else None
             context_with_additional = state.context
             if additional_context:
                 # Create a copy of context with additional fields from approval
-                if hasattr(state.context, '__dict__'):
+                if hasattr(state.context, "__dict__"):
                     # For dataclass contexts, add additional context as attributes
                     context_dict = {**state.context.__dict__, **additional_context}
-                    context_with_additional = type(state.context)(**{k: v for k, v in context_dict.items() if k in state.context.__dict__})
+                    context_with_additional = type(state.context)(
+                        **{k: v for k, v in context_dict.items() if k in state.context.__dict__}
+                    )
                     # Add any extra fields as attributes
                     for key, value in additional_context.items():
                         if not hasattr(context_with_additional, key):
@@ -1217,143 +1491,167 @@ async def _execute_tool_calls(
                 else:
                     # For dict contexts, merge normally
                     context_with_additional = {**state.context, **additional_context}
-            
-            print(f'[JAF:ENGINE] About to execute tool: {tool_call.function.name}')
-            print(f'[JAF:ENGINE] Tool args:', validated_args)
-            print(f'[JAF:ENGINE] Tool context:', state.context)
-            
+
+            print(f"[JAF:ENGINE] About to execute tool: {tool_call.function.name}")
+            print(f"[JAF:ENGINE] Tool args:", validated_args)
+            print(f"[JAF:ENGINE] Tool context:", state.context)
+
             # Execute the tool with timeout
             try:
                 tool_result = await asyncio.wait_for(
-                    tool.execute(validated_args, context_with_additional),
-                    timeout=timeout
+                    tool.execute(validated_args, context_with_additional), timeout=timeout
                 )
             except asyncio.TimeoutError:
-                timeout_error_result = json.dumps({
-                    'hitl_status': 'execution_timeout',  # HITL workflow status
-                    'message': f'Tool {tool_call.function.name} timed out after {timeout} seconds',
-                    'tool_name': tool_call.function.name,
-                    'timeout_seconds': timeout
-                })
+                timeout_error_result = json.dumps(
+                    {
+                        "hitl_status": "execution_timeout",  # HITL workflow status
+                        "message": f"Tool {tool_call.function.name} timed out after {timeout} seconds",
+                        "tool_name": tool_call.function.name,
+                        "timeout_seconds": timeout,
+                    }
+                )
 
                 if config.on_event:
-                    config.on_event(ToolCallEndEvent(data=to_event_data(ToolCallEndEventData(
-                        tool_name=tool_call.function.name,
-                        result=timeout_error_result,
-                        trace_id=state.trace_id,
-                        run_id=state.run_id,
-                        execution_status='timeout',  # Tool execution timed out
-                        tool_result={'error': 'timeout'},
-                        call_id=tool_call.id
-                    ))))
+                    config.on_event(
+                        ToolCallEndEvent(
+                            data=to_event_data(
+                                ToolCallEndEventData(
+                                    tool_name=tool_call.function.name,
+                                    result=timeout_error_result,
+                                    trace_id=state.trace_id,
+                                    run_id=state.run_id,
+                                    execution_status="timeout",  # Tool execution timed out
+                                    tool_result={"error": "timeout"},
+                                    call_id=tool_call.id,
+                                )
+                            )
+                        )
+                    )
 
                 return {
-                    'message': Message(
+                    "message": Message(
                         role=ContentRole.TOOL,
                         content=timeout_error_result,
-                        tool_call_id=tool_call.id
+                        tool_call_id=tool_call.id,
                     )
                 }
 
             # Handle both string and ToolResult formats
             if isinstance(tool_result, str):
                 result_string = tool_result
-                print(f'[JAF:ENGINE] Tool {tool_call.function.name} returned string:', result_string)
+                print(
+                    f"[JAF:ENGINE] Tool {tool_call.function.name} returned string:", result_string
+                )
             else:
                 # It's a ToolResult object
                 result_string = tool_result_to_string(tool_result)
-                print(f'[JAF:ENGINE] Tool {tool_call.function.name} returned ToolResult:', tool_result)
-                print(f'[JAF:ENGINE] Converted to string:', result_string)
+                print(
+                    f"[JAF:ENGINE] Tool {tool_call.function.name} returned ToolResult:", tool_result
+                )
+                print(f"[JAF:ENGINE] Converted to string:", result_string)
 
             # Wrap tool result with status information for approval context
             if approval_status and approval_status.additional_context:
-                final_content = json.dumps({
-                    'hitl_status': 'approved_and_executed',  # HITL workflow status: approved by user and executed
-                    'result': result_string,
-                    'tool_name': tool_call.function.name,
-                    'approval_context': approval_status.additional_context,
-                    'message': 'Tool was approved and executed successfully with additional context.'
-                })
+                final_content = json.dumps(
+                    {
+                        "hitl_status": "approved_and_executed",  # HITL workflow status: approved by user and executed
+                        "result": result_string,
+                        "tool_name": tool_call.function.name,
+                        "approval_context": approval_status.additional_context,
+                        "message": "Tool was approved and executed successfully with additional context.",
+                    }
+                )
             elif needs_approval:
-                final_content = json.dumps({
-                    'hitl_status': 'approved_and_executed',  # HITL workflow status: approved by user and executed
-                    'result': result_string,
-                    'tool_name': tool_call.function.name,
-                    'message': 'Tool was approved and executed successfully.'
-                })
+                final_content = json.dumps(
+                    {
+                        "hitl_status": "approved_and_executed",  # HITL workflow status: approved by user and executed
+                        "result": result_string,
+                        "tool_name": tool_call.function.name,
+                        "message": "Tool was approved and executed successfully.",
+                    }
+                )
             else:
-                final_content = json.dumps({
-                    'hitl_status': 'executed',  # HITL workflow status: executed normally (no approval needed)
-                    'result': result_string,
-                    'tool_name': tool_call.function.name,
-                    'message': 'Tool executed successfully.'
-                })
+                final_content = json.dumps(
+                    {
+                        "hitl_status": "executed",  # HITL workflow status: executed normally (no approval needed)
+                        "result": result_string,
+                        "tool_name": tool_call.function.name,
+                        "message": "Tool executed successfully.",
+                    }
+                )
 
             if config.on_event:
-                config.on_event(ToolCallEndEvent(data=to_event_data(ToolCallEndEventData(
-                    tool_name=tool_call.function.name,
-                    result=final_content,
-                    trace_id=state.trace_id,
-                    run_id=state.run_id,
-                    tool_result=tool_result,
-                    execution_status='success',  # Tool execution succeeded
-                    call_id=tool_call.id
-                ))))
+                config.on_event(
+                    ToolCallEndEvent(
+                        data=to_event_data(
+                            ToolCallEndEventData(
+                                tool_name=tool_call.function.name,
+                                result=final_content,
+                                trace_id=state.trace_id,
+                                run_id=state.run_id,
+                                tool_result=tool_result,
+                                execution_status="success",  # Tool execution succeeded
+                                call_id=tool_call.id,
+                            )
+                        )
+                    )
+                )
 
             # Check for handoff
             handoff_check = _try_parse_json(result_string)
-            if isinstance(handoff_check, dict) and 'handoff_to' in handoff_check:
+            if isinstance(handoff_check, dict) and "handoff_to" in handoff_check:
                 return {
-                    'message': Message(
-                        role=ContentRole.TOOL,
-                        content=final_content,
-                        tool_call_id=tool_call.id
+                    "message": Message(
+                        role=ContentRole.TOOL, content=final_content, tool_call_id=tool_call.id
                     ),
-                    'is_handoff': True,
-                    'target_agent': handoff_check['handoff_to']
+                    "is_handoff": True,
+                    "target_agent": handoff_check["handoff_to"],
                 }
 
             return {
-                'message': Message(
-                    role=ContentRole.TOOL,
-                    content=final_content,
-                    tool_call_id=tool_call.id
+                "message": Message(
+                    role=ContentRole.TOOL, content=final_content, tool_call_id=tool_call.id
                 )
             }
 
         except Exception as error:
-            error_result = json.dumps({
-                'hitl_status': 'execution_error',  # HITL workflow status
-                'message': str(error),
-                'tool_name': tool_call.function.name,
-            })
+            error_result = json.dumps(
+                {
+                    "hitl_status": "execution_error",  # HITL workflow status
+                    "message": str(error),
+                    "tool_name": tool_call.function.name,
+                }
+            )
 
             if config.on_event:
-                config.on_event(ToolCallEndEvent(data=to_event_data(ToolCallEndEventData(
-                    tool_name=tool_call.function.name,
-                    result=error_result,
-                    trace_id=state.trace_id,
-                    run_id=state.run_id,
-                    execution_status='error',  # Tool execution failed with exception
-                    tool_result={'error': 'execution_error', 'detail': str(error)},
-                    call_id=tool_call.id
-                ))))
+                config.on_event(
+                    ToolCallEndEvent(
+                        data=to_event_data(
+                            ToolCallEndEventData(
+                                tool_name=tool_call.function.name,
+                                result=error_result,
+                                trace_id=state.trace_id,
+                                run_id=state.run_id,
+                                execution_status="error",  # Tool execution failed with exception
+                                tool_result={"error": "execution_error", "detail": str(error)},
+                                call_id=tool_call.id,
+                            )
+                        )
+                    )
+                )
 
             return {
-                'message': Message(
-                    role=ContentRole.TOOL,
-                    content=error_result,
-                    tool_call_id=tool_call.id
+                "message": Message(
+                    role=ContentRole.TOOL, content=error_result, tool_call_id=tool_call.id
                 )
             }
 
     # Execute all tool calls in parallel
-    results = await asyncio.gather(*[
-        execute_single_tool_call(tc) for tc in tool_calls
-    ])
+    results = await asyncio.gather(*[execute_single_tool_call(tc) for tc in tool_calls])
 
     return results
 
+
 def _try_parse_json(text: str) -> Any:
     """Try to parse JSON, return original string if it fails."""
     if not text or not isinstance(text, str):