autobyteus 1.1.2__py3-none-any.whl → 1.1.4__py3-none-any.whl

autobyteus/agent/agent.py

@@ -4,7 +4,7 @@ import logging
 from typing import AsyncIterator, Optional, List, Any, Dict, TYPE_CHECKING 
 
 from autobyteus.agent.runtime.agent_runtime import AgentRuntime
-from autobyteus.agent.context.phases import AgentOperationalPhase 
+from autobyteus.agent.phases.phase_enum import AgentOperationalPhase 
 from autobyteus.agent.message.agent_input_user_message import AgentInputUserMessage
 from autobyteus.agent.message.inter_agent_message import InterAgentMessage
 from autobyteus.agent.events import UserMessageReceivedEvent, InterAgentMessageReceivedEvent, ToolExecutionApprovalEvent, BaseEvent 

autobyteus/agent/bootstrap_steps/__init__.py

@@ -9,6 +9,7 @@ from .agent_runtime_queue_initialization_step import AgentRuntimeQueueInitializa
 from .workspace_context_initialization_step import WorkspaceContextInitializationStep
 # ToolInitializationStep is no longer a bootstrap step.
 from .system_prompt_processing_step import SystemPromptProcessingStep
+from .mcp_server_prewarming_step import McpServerPrewarmingStep
 # LLMConfigFinalizationStep and LLMInstanceCreationStep removed.
 
 __all__ = [
@@ -16,4 +17,5 @@ __all__ = [
     "AgentRuntimeQueueInitializationStep", # UPDATED
     "WorkspaceContextInitializationStep",
     "SystemPromptProcessingStep",
+    "McpServerPrewarmingStep",
 ]

autobyteus/agent/bootstrap_steps/agent_bootstrapper.py

@@ -6,6 +6,7 @@ from .base_bootstrap_step import BaseBootstrapStep
 from .agent_runtime_queue_initialization_step import AgentRuntimeQueueInitializationStep
 from .workspace_context_initialization_step import WorkspaceContextInitializationStep
 from .system_prompt_processing_step import SystemPromptProcessingStep
+from .mcp_server_prewarming_step import McpServerPrewarmingStep
 from autobyteus.agent.events import AgentReadyEvent
 
 if TYPE_CHECKING:
@@ -31,6 +32,7 @@ class AgentBootstrapper:
             self.bootstrap_steps: List[BaseBootstrapStep] = [
                 AgentRuntimeQueueInitializationStep(),
                 WorkspaceContextInitializationStep(),
+                McpServerPrewarmingStep(),
                 SystemPromptProcessingStep(),
             ]
             logger.debug("AgentBootstrapper initialized with default steps.")

autobyteus/agent/bootstrap_steps/mcp_server_prewarming_step.py

@@ -0,0 +1,71 @@
+# file: autobyteus/autobyteus/agent/bootstrap_steps/mcp_server_prewarming_step.py
+import logging
+from typing import TYPE_CHECKING, Set
+
+from .base_bootstrap_step import BaseBootstrapStep
+from autobyteus.tools.mcp.config_service import McpConfigService
+from autobyteus.tools.mcp.server_instance_manager import McpServerInstanceManager
+from autobyteus.tools.tool_category import ToolCategory
+
+if TYPE_CHECKING:
+    from autobyteus.agent.context import AgentContext
+    from autobyteus.agent.phases import AgentPhaseManager
+
+logger = logging.getLogger(__name__)
+
+class McpServerPrewarmingStep(BaseBootstrapStep):
+    """
+    Bootstrap step to eagerly start all MCP servers associated with an agent's tools.
+    This ensures servers are running and ready before the agent becomes idle.
+    """
+
+    def __init__(self):
+        self._config_service = McpConfigService()
+        self._instance_manager = McpServerInstanceManager()
+        logger.debug("McpServerPrewarmingStep initialized.")
+
+    async def execute(self,
+                      context: 'AgentContext',
+                      phase_manager: 'AgentPhaseManager') -> bool:
+        agent_id = context.agent_id
+        logger.info(f"Agent '{agent_id}': Executing McpServerPrewarmingStep.")
+
+        # 1. Find all unique server IDs by inspecting tool definitions.
+        mcp_server_ids: Set[str] = set()
+        for tool in context.config.tools:
+            # This is the new, superior check. It relies on abstract metadata, not concrete types.
+            if tool.definition and tool.definition.category == ToolCategory.MCP:
+                # This is the new, superior way to get the server_id.
+                # It does not rely on private attributes of the tool instance.
+                server_id = tool.definition.metadata.get("mcp_server_id")
+                if server_id:
+                    mcp_server_ids.add(server_id)
+
+        if not mcp_server_ids:
+            logger.debug(f"Agent '{agent_id}': No MCP tools found. Skipping MCP server pre-warming.")
+            return True
+
+        logger.info(f"Agent '{agent_id}': Found {len(mcp_server_ids)} unique MCP server IDs to pre-warm: {mcp_server_ids}")
+
+        # 2. For each server ID, unconditionally start its server instance for this agent.
+        for server_id in mcp_server_ids:
+            try:
+                config = self._config_service.get_config(server_id)
+                if not config:
+                    logger.warning(f"Agent '{agent_id}': Could not find config for server_id '{server_id}' used by a tool. Cannot pre-warm.")
+                    continue
+
+                logger.info(f"Agent '{agent_id}': Pre-warming MCP server '{server_id}'.")
+                # Get the instance for this agent, which creates it if it doesn't exist.
+                server_instance = self._instance_manager.get_server_instance(agent_id, server_id)
+                # Explicitly connect to start the server process.
+                await server_instance.connect()
+                logger.info(f"Agent '{agent_id}': Successfully connected to pre-warmed MCP server '{server_id}'.")
+
+            except Exception as e:
+                error_message = f"Agent '{agent_id}': Failed to pre-warm MCP server '{server_id}': {e}"
+                logger.error(error_message, exc_info=True)
+                # A failure to pre-warm a server is a critical bootstrap failure.
+                return False
+        
+        return True

autobyteus/agent/bootstrap_steps/system_prompt_processing_step.py

@@ -34,8 +34,10 @@ class SystemPromptProcessingStep(BaseBootstrapStep):
             if not llm_instance:
                 raise ValueError("LLM instance not found in agent state. It must be provided in AgentConfig.")
 
-            current_system_prompt = context.config.system_prompt
-            logger.debug(f"Agent '{agent_id}': Retrieved base system prompt from agent config.")
+            # If a specific system_prompt is not provided in AgentConfig, fall back
+            # to the default system_message from the LLM's own configuration.
+            current_system_prompt = context.config.system_prompt or llm_instance.config.system_message
+            logger.debug(f"Agent '{agent_id}': Retrieved base system prompt.")
             
             processor_instances = context.config.system_prompt_processors
             tool_instances_for_processor = context.tool_instances

autobyteus/agent/context/agent_config.py

@@ -1,5 +1,6 @@
 # file: autobyteus/autobyteus/agent/context/agent_config.py
 import logging
+import copy
 from typing import List, Optional, Union, Tuple, TYPE_CHECKING, Dict, Any
 
 # Correctly import the new master processor and the base class
@@ -10,6 +11,7 @@ from autobyteus.agent.llm_response_processor import ProviderAwareToolUsageProces
 if TYPE_CHECKING:
     from autobyteus.tools.base_tool import BaseTool
     from autobyteus.agent.input_processor import BaseAgentUserInputMessageProcessor
+    from autobyteus.agent.tool_execution_result_processor import BaseToolExecutionResultProcessor
     from autobyteus.llm.base_llm import BaseLLM
     from autobyteus.agent.workspace.base_workspace import BaseAgentWorkspace
     from autobyteus.agent.hooks.base_phase_hook import BasePhaseHook
@@ -32,13 +34,14 @@ class AgentConfig:
                  role: str,
                  description: str,
                  llm_instance: 'BaseLLM',
-                 system_prompt: str,
-                 tools: List['BaseTool'],
+                 system_prompt: Optional[str] = None,
+                 tools: Optional[List['BaseTool']] = None,
                  auto_execute_tools: bool = True,
                  use_xml_tool_format: bool = True,
                  input_processors: Optional[List['BaseAgentUserInputMessageProcessor']] = None,
                  llm_response_processors: Optional[List['BaseLLMResponseProcessor']] = None,
                  system_prompt_processors: Optional[List['BaseSystemPromptProcessor']] = None,
+                 tool_execution_result_processors: Optional[List['BaseToolExecutionResultProcessor']] = None,
                  workspace: Optional['BaseAgentWorkspace'] = None,
                  phase_hooks: Optional[List['BasePhaseHook']] = None,
                  initial_custom_data: Optional[Dict[str, Any]] = None):
@@ -51,13 +54,15 @@ class AgentConfig:
             description: A description of the agent.
             llm_instance: A pre-initialized LLM instance (subclass of BaseLLM).
                           The user is responsible for creating and configuring this instance.
-            system_prompt: The base system prompt.
-            tools: A list of pre-initialized tool instances (subclasses of BaseTool).
+            system_prompt: The base system prompt. If None, the system_message from the
+                           llm_instance's config will be used as the base.
+            tools: An optional list of pre-initialized tool instances (subclasses of BaseTool).
             auto_execute_tools: If True, the agent will execute tools without approval.
             use_xml_tool_format: Whether to use XML for tool descriptions and examples.
             input_processors: A list of input processor instances.
             llm_response_processors: A list of LLM response processor instances.
             system_prompt_processors: A list of system prompt processor instances.
+            tool_execution_result_processors: A list of tool execution result processor instances.
             workspace: An optional pre-initialized workspace instance for the agent.
             phase_hooks: An optional list of phase transition hook instances.
             initial_custom_data: An optional dictionary of data to pre-populate
@@ -68,17 +73,43 @@ class AgentConfig:
         self.description = description
         self.llm_instance = llm_instance
         self.system_prompt = system_prompt
-        self.tools = tools
+        self.tools = tools or []
         self.workspace = workspace
         self.auto_execute_tools = auto_execute_tools
         self.use_xml_tool_format = use_xml_tool_format
         self.input_processors = input_processors or []
         self.llm_response_processors = llm_response_processors if llm_response_processors is not None else list(self.DEFAULT_LLM_RESPONSE_PROCESSORS)
         self.system_prompt_processors = system_prompt_processors if system_prompt_processors is not None else list(self.DEFAULT_SYSTEM_PROMPT_PROCESSORS)
+        self.tool_execution_result_processors = tool_execution_result_processors or []
         self.phase_hooks = phase_hooks or []
         self.initial_custom_data = initial_custom_data
 
         logger.debug(f"AgentConfig created for name '{self.name}', role '{self.role}'.")
 
+    def copy(self) -> 'AgentConfig':
+        """
+        Creates a copy of this AgentConfig. It avoids deep-copying complex objects
+        like tools, workspaces, and processors that may contain un-pickleable state.
+        Instead, it creates shallow copies of the lists, allowing the lists themselves
+        to be modified independently while sharing the object instances within them.
+        """
+        return AgentConfig(
+            name=self.name,
+            role=self.role,
+            description=self.description,
+            llm_instance=self.llm_instance,  # Keep reference, do not copy
+            system_prompt=self.system_prompt,
+            tools=self.tools.copy(),  # Shallow copy the list, but reference the original tool instances
+            auto_execute_tools=self.auto_execute_tools,
+            use_xml_tool_format=self.use_xml_tool_format,
+            input_processors=self.input_processors.copy(), # Shallow copy the list
+            llm_response_processors=self.llm_response_processors.copy(), # Shallow copy the list
+            system_prompt_processors=self.system_prompt_processors.copy(), # Shallow copy the list
+            tool_execution_result_processors=self.tool_execution_result_processors.copy(), # Shallow copy the list
+            workspace=self.workspace,  # Pass by reference, do not copy
+            phase_hooks=self.phase_hooks.copy(), # Shallow copy the list
+            initial_custom_data=copy.deepcopy(self.initial_custom_data) # Deep copy for simple data
+        )
+
     def __repr__(self) -> str:
         return (f"AgentConfig(name='{self.name}', role='{self.role}', llm_instance='{self.llm_instance.__class__.__name__}', workspace_configured={self.workspace is not None})")

autobyteus/agent/events/worker_event_dispatcher.py

@@ -21,8 +21,7 @@ from autobyteus.agent.events.agent_events import ( # Updated relative import pat
 if TYPE_CHECKING:
     from autobyteus.agent.context import AgentContext
     from autobyteus.agent.handlers import EventHandlerRegistry
-    from autobyteus.agent.context.agent_phase_manager import AgentPhaseManager
-
+    from autobyteus.agent.phases import AgentPhaseManager
 logger = logging.getLogger(__name__)
 
 class WorkerEventDispatcher:

autobyteus/agent/handlers/inter_agent_message_event_handler.py

@@ -43,7 +43,7 @@ class InterAgentMessageReceivedEventHandler(AgentEventHandler):
         logger.info(
             f"Agent '{context.agent_id}' handling InterAgentMessageReceivedEvent from sender "
             f"'{inter_agent_msg.sender_agent_id}', type '{inter_agent_msg.message_type.value}'. "
-            f"Content: '{inter_agent_msg.content[:100]}...'"
+            f"Content: '{inter_agent_msg.content}'"
         )
 
         content_for_llm = (

autobyteus/agent/handlers/llm_user_message_ready_event_handler.py

@@ -45,7 +45,7 @@ class LLMUserMessageReadyEventHandler(AgentEventHandler):
             raise RuntimeError(error_msg) 
 
         llm_user_message: LLMUserMessage = event.llm_user_message
-        logger.info(f"Agent '{agent_id}' handling LLMUserMessageReadyEvent: '{llm_user_message.content[:100]}...'") 
+        logger.info(f"Agent '{agent_id}' handling LLMUserMessageReadyEvent: '{llm_user_message.content}'") 
         logger.debug(f"Agent '{agent_id}' preparing to send full message to LLM:\n---\n{llm_user_message.content}\n---")
         
         context.state.add_message_to_history({"role": "user", "content": llm_user_message.content})
@@ -137,4 +137,4 @@ class LLMUserMessageReadyEventHandler(AgentEventHandler):
             complete_response=complete_response_obj
         )
         await context.input_event_queues.enqueue_internal_system_event(llm_complete_event)
-        logger.info(f"Agent '{agent_id}' enqueued LLMCompleteResponseReceivedEvent from LLMUserMessageReadyEventHandler.")
+        logger.info(f"Agent '{agent_id}' enqueued LLMCompleteResponseReceivedEvent from LLMUserMessageReadyEventHandler.")

autobyteus/agent/handlers/tool_result_event_handler.py

@@ -6,6 +6,7 @@ from typing import TYPE_CHECKING, Optional
 from autobyteus.agent.handlers.base_event_handler import AgentEventHandler 
 from autobyteus.agent.events import ToolResultEvent, LLMUserMessageReadyEvent 
 from autobyteus.llm.user_message import LLMUserMessage 
+from autobyteus.agent.tool_execution_result_processor import BaseToolExecutionResultProcessor
 
 if TYPE_CHECKING:
     from autobyteus.agent.context import AgentContext 
@@ -29,10 +30,37 @@ class ToolResultEventHandler(AgentEventHandler):
             logger.warning(f"ToolResultEventHandler received non-ToolResultEvent: {type(event)}. Skipping.")
             return
 
-        agent_id = context.agent_id 
-        tool_invocation_id = event.tool_invocation_id if event.tool_invocation_id else 'N/A'
+        agent_id = context.agent_id
+        processed_event = event
 
-        logger.info(f"Agent '{agent_id}' handling ToolResultEvent from tool: '{event.tool_name}' (Invocation ID: {tool_invocation_id}). Error: {event.error is not None}")
+        # --- New: Apply Tool Execution Result Processors ---
+        processor_instances = context.config.tool_execution_result_processors
+        if processor_instances:
+            processor_names = [p.get_name() for p in processor_instances]
+            logger.debug(f"Agent '{agent_id}': Applying tool execution result processors: {processor_names}")
+            for processor_instance in processor_instances:
+                processor_name_for_log = "unknown"
+                try:
+                    if not isinstance(processor_instance, BaseToolExecutionResultProcessor):
+                        logger.error(f"Agent '{agent_id}': Invalid tool result processor type: {type(processor_instance)}. Skipping.")
+                        continue
+                    
+                    processor_name_for_log = processor_instance.get_name()
+                    logger.debug(f"Agent '{agent_id}': Applying tool result processor '{processor_name_for_log}'.")
+                    
+                    event_before_proc = processed_event
+                    processed_event = await processor_instance.process(event_before_proc, context)
+                    logger.info(f"Agent '{agent_id}': Tool result processor '{processor_name_for_log}' applied successfully.")
+
+                except Exception as e:
+                    logger.error(f"Agent '{agent_id}': Error applying tool result processor '{processor_name_for_log}': {e}. "
+                                 f"Skipping and continuing with result from before this processor.", exc_info=True)
+                    processed_event = event_before_proc
+        # --- End New ---
+
+        tool_invocation_id = processed_event.tool_invocation_id if processed_event.tool_invocation_id else 'N/A'
+
+        logger.info(f"Agent '{agent_id}' handling processed ToolResultEvent from tool: '{processed_event.tool_name}' (Invocation ID: {tool_invocation_id}). Error: {processed_event.error is not None}")
         
         notifier: Optional['AgentExternalEventNotifier'] = None
         if context.phase_manager:
@@ -41,61 +69,61 @@ class ToolResultEventHandler(AgentEventHandler):
         if not notifier: # pragma: no cover
             logger.error(f"Agent '{agent_id}': Notifier not available in ToolResultEventHandler. Tool result processing logs will not be emitted.")
 
-        if event.error:
-            logger.debug(f"Agent '{agent_id}' tool '{event.tool_name}' (ID: {tool_invocation_id}) raw error details: {event.error}")
+        if processed_event.error:
+            logger.debug(f"Agent '{agent_id}' tool '{processed_event.tool_name}' (ID: {tool_invocation_id}) raw error details: {processed_event.error}")
         else:
             try:
-                raw_result_str_for_debug_log = json.dumps(event.result, indent=2)
+                raw_result_str_for_debug_log = json.dumps(processed_event.result, indent=2)
             except TypeError: # pragma: no cover
-                raw_result_str_for_debug_log = str(event.result)
-            logger.debug(f"Agent '{agent_id}' tool '{event.tool_name}' (ID: {tool_invocation_id}) raw result:\n---\n{raw_result_str_for_debug_log}\n---")
+                raw_result_str_for_debug_log = str(processed_event.result)
+            logger.debug(f"Agent '{agent_id}' tool '{processed_event.tool_name}' (ID: {tool_invocation_id}) raw result:\n---\n{raw_result_str_for_debug_log}\n---")
 
 
         content_for_llm: str
-        if event.error:
+        if processed_event.error:
             content_for_llm = (
-                f"The tool '{event.tool_name}' (invocation ID: {tool_invocation_id}) encountered an error.\n"
-                f"Error details: {event.error}\n"
+                f"The tool '{processed_event.tool_name}' (invocation ID: {tool_invocation_id}) encountered an error.\n"
+                f"Error details: {processed_event.error}\n"
                 f"Please analyze this error and decide the next course of action."
             )
-            log_msg_error_processed = f"[TOOL_RESULT_ERROR_PROCESSED] Agent_ID: {agent_id}, Tool: {event.tool_name}, Invocation_ID: {tool_invocation_id}, Error: {event.error}"
+            log_msg_error_processed = f"[TOOL_RESULT_ERROR_PROCESSED] Agent_ID: {agent_id}, Tool: {processed_event.tool_name}, Invocation_ID: {tool_invocation_id}, Error: {processed_event.error}"
             if notifier:
                 try:
                     log_data = {
                         "log_entry": log_msg_error_processed,
                         "tool_invocation_id": tool_invocation_id,
-                        "tool_name": event.tool_name,
+                        "tool_name": processed_event.tool_name,
                     }
                     notifier.notify_agent_data_tool_log(log_data)
                 except Exception as e_notify: 
                     logger.error(f"Agent '{agent_id}': Error notifying tool result error log: {e_notify}", exc_info=True)
         else:
             try:
-                result_str_for_llm = json.dumps(event.result, indent=2) if not isinstance(event.result, str) else event.result
+                result_str_for_llm = json.dumps(processed_event.result, indent=2) if not isinstance(processed_event.result, str) else processed_event.result
             except TypeError: # pragma: no cover
-                result_str_for_llm = str(event.result)
+                result_str_for_llm = str(processed_event.result)
 
             content_for_llm = (
-                f"The tool '{event.tool_name}' (invocation ID: {tool_invocation_id}) has executed.\n"
+                f"The tool '{processed_event.tool_name}' (invocation ID: {tool_invocation_id}) has executed.\n"
                 f"Result:\n{result_str_for_llm}\n" 
                 f"Based on this result, what is the next step or final answer?"
             )
-            log_msg_success_processed = f"[TOOL_RESULT_SUCCESS_PROCESSED] Agent_ID: {agent_id}, Tool: {event.tool_name}, Invocation_ID: {tool_invocation_id}, Result (first 200 chars of stringified): {str(event.result)[:200]}"
+            log_msg_success_processed = f"[TOOL_RESULT_SUCCESS_PROCESSED] Agent_ID: {agent_id}, Tool: {processed_event.tool_name}, Invocation_ID: {tool_invocation_id}, Result: {str(processed_event.result)}"
             if notifier:
                 try:
                     log_data = {
                         "log_entry": log_msg_success_processed,
                         "tool_invocation_id": tool_invocation_id,
-                        "tool_name": event.tool_name,
+                        "tool_name": processed_event.tool_name,
                     }
                     notifier.notify_agent_data_tool_log(log_data)
                 except Exception as e_notify: 
                     logger.error(f"Agent '{agent_id}': Error notifying tool result success log: {e_notify}", exc_info=True)
         
-        logger.debug(f"Agent '{agent_id}' preparing message for LLM based on tool '{event.tool_name}' (ID: {tool_invocation_id}) result:\n---\n{content_for_llm}\n---")
+        logger.debug(f"Agent '{agent_id}' preparing message for LLM based on tool '{processed_event.tool_name}' (ID: {tool_invocation_id}) result:\n---\n{content_for_llm}\n---")
         llm_user_message = LLMUserMessage(content=content_for_llm)
         
         next_event = LLMUserMessageReadyEvent(llm_user_message=llm_user_message) 
         await context.input_event_queues.enqueue_internal_system_event(next_event)
         
-        logger.info(f"Agent '{agent_id}' enqueued LLMUserMessageReadyEvent for LLM based on tool '{event.tool_name}' (ID: {tool_invocation_id}) result summary.") 
+        logger.info(f"Agent '{agent_id}' enqueued LLMUserMessageReadyEvent for LLM based on tool '{processed_event.tool_name}' (ID: {tool_invocation_id}) result summary.")

autobyteus/agent/handlers/user_input_message_event_handler.py

@@ -35,7 +35,7 @@ class UserInputMessageEventHandler(AgentEventHandler):
         original_agent_input_user_msg: AgentInputUserMessage = event.agent_input_user_message 
         processed_agent_input_user_msg: AgentInputUserMessage = original_agent_input_user_msg 
         
-        logger.info(f"Agent '{context.agent_id}' handling UserMessageReceivedEvent: '{original_agent_input_user_msg.content[:100]}...'") 
+        logger.info(f"Agent '{context.agent_id}' handling UserMessageReceivedEvent: '{original_agent_input_user_msg.content}'") 
         
         processor_instances = context.config.input_processors
         if processor_instances:

autobyteus/agent/input_processor/__init__.py

@@ -4,15 +4,9 @@ Components for pre-processing AgentUserMessage objects.
 """
 from .base_user_input_processor import BaseAgentUserInputMessageProcessor
 
-# Import concrete processors to make them easily accessible for instantiation
-from .passthrough_input_processor import PassthroughInputProcessor
-from .metadata_appending_input_processor import MetadataAppendingInputProcessor
-from .content_prefixing_input_processor import ContentPrefixingInputProcessor
+# Concrete processors have been removed. Users can define their own.
 
 
 __all__ = [
     "BaseAgentUserInputMessageProcessor",
-    "PassthroughInputProcessor",
-    "MetadataAppendingInputProcessor",
-    "ContentPrefixingInputProcessor",
 ]

autobyteus/agent/llm_response_processor/provider_aware_tool_usage_processor.py

@@ -1,11 +1,12 @@
 # file: autobyteus/autobyteus/agent/llm_response_processor/provider_aware_tool_usage_processor.py
 import logging
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, List
 
 from .base_processor import BaseLLMResponseProcessor
+from autobyteus.agent.events import PendingToolInvocationEvent
+from autobyteus.agent.tool_invocation import ToolInvocation
 from autobyteus.tools.usage.parsers import ProviderAwareToolUsageParser
 from autobyteus.tools.usage.parsers.exceptions import ToolUsageParseException
-from autobyteus.agent.events import PendingToolInvocationEvent
 
 if TYPE_CHECKING:
     from autobyteus.agent.context import AgentContext
@@ -17,21 +18,20 @@ logger = logging.getLogger(__name__)
 class ProviderAwareToolUsageProcessor(BaseLLMResponseProcessor):
     """
     A "master" tool usage processor that uses a high-level parser from the
-    `tools` module to extract tool invocations, and then enqueues the
-    necessary agent events based on the parsed results.
+    `tools` module to extract tool invocations. It then ensures each invocation
+    has a session-unique ID before enqueuing the necessary agent events.
     """
+    INVOCATION_COUNTS_KEY = "agent_tool_invocation_counts"
+
     def __init__(self):
         self._parser = ProviderAwareToolUsageParser()
         logger.debug("ProviderAwareToolUsageProcessor initialized.")
 
-    @classmethod
-    def get_name(cls) -> str:
-        return "provider_aware_tool_usage"
-
     async def process_response(self, response: 'CompleteResponse', context: 'AgentContext', triggering_event: 'LLMCompleteResponseReceivedEvent') -> bool:
         """
-        Uses a ProviderAwareToolUsageParser to get a list of tool invocations,
-        and then enqueues a PendingToolInvocationEvent for each one.
+        Uses a ProviderAwareToolUsageParser to get tool invocations, makes their
+        IDs unique within the agent's session, and then enqueues a
+        PendingToolInvocationEvent for each one.
         Propagates ToolUsageParseException if parsing fails.
         """
         try:
@@ -44,9 +44,38 @@ class ProviderAwareToolUsageProcessor(BaseLLMResponseProcessor):
         if not tool_invocations:
             return False
 
-        logger.info(f"Agent '{context.agent_id}': Parsed {len(tool_invocations)} tool invocations. Enqueuing events.")
+        # --- NEW LOGIC FOR UNIQUE ID GENERATION ---
+        
+        # Ensure the counter map exists in the agent's state's custom data
+        if self.INVOCATION_COUNTS_KEY not in context.custom_data:
+            context.custom_data[self.INVOCATION_COUNTS_KEY] = {}
+        
+        invocation_counts = context.custom_data[self.INVOCATION_COUNTS_KEY]
+        
+        processed_invocations: List[ToolInvocation] = []
+
         for invocation in tool_invocations:
-            logger.info(f"Agent '{context.agent_id}' ({self.get_name()}) identified tool invocation: {invocation.name}. Enqueuing event.")
+            base_id = invocation.id
+            
+            # Get the current count for this base ID, default to 0
+            count = invocation_counts.get(base_id, 0)
+            
+            # Create the new session-unique ID
+            unique_id = f"{base_id}_{count}"
+            
+            # Update the invocation's ID in-place
+            invocation.id = unique_id
+            
+            # Increment the counter for the next time this base ID is seen
+            invocation_counts[base_id] = count + 1
+            
+            processed_invocations.append(invocation)
+
+        # --- END NEW LOGIC ---
+
+        logger.info(f"Agent '{context.agent_id}': Parsed {len(processed_invocations)} tool invocations. Enqueuing events with unique IDs.")
+        for invocation in processed_invocations:
+            logger.info(f"Agent '{context.agent_id}' ({self.get_name()}) identified tool invocation: {invocation.name} with unique ID {invocation.id}. Enqueuing event.")
             await context.input_event_queues.enqueue_tool_invocation_request(
                 PendingToolInvocationEvent(tool_invocation=invocation)
             )

autobyteus/agent/message/context_file_type.py

@@ -17,6 +17,8 @@ class ContextFileType(str, Enum):
     HTML = "html"          # .html, .htm
     PYTHON = "python"      # .py
     JAVASCRIPT = "javascript" # .js
+    AUDIO = "audio"        # .mp3, .wav, .m4a, .flac, .ogg
+    VIDEO = "video"        # .mp4, .mov, .avi, .mkv, .webm
     IMAGE = "image"        # .png, .jpg, .jpeg, .gif, .webp (when image is for contextual analysis, not direct LLM vision input)
     UNKNOWN = "unknown"    # Fallback for unrecognized types
 
@@ -54,6 +56,10 @@ class ContextFileType(str, Enum):
             return cls.PYTHON
         elif extension == ".js":
             return cls.JAVASCRIPT
+        elif extension in [".mp3", ".wav", ".m4a", ".flac", ".ogg"]:
+            return cls.AUDIO
+        elif extension in [".mp4", ".mov", ".avi", ".mkv", ".webm"]:
+            return cls.VIDEO
         elif extension in [".png", ".jpg", ".jpeg", ".gif", ".webp"]:
             return cls.IMAGE 
         else: