autobyteus 1.2.1__py3-none-any.whl → 1.3.0__py3-none-any.whl

autobyteus/llm/extensions/token_usage_tracking_extension.py

@@ -1,4 +1,4 @@
-from typing import Optional, List, TYPE_CHECKING
+from typing import Optional, List, TYPE_CHECKING, Any
 import logging
 from autobyteus.llm.extensions.base_extension import LLMExtension
 from autobyteus.llm.token_counter.token_counter_factory import get_token_counter
@@ -6,7 +6,6 @@ from autobyteus.llm.utils.token_usage import TokenUsage
 from autobyteus.llm.utils.token_usage_tracker import TokenUsageTracker
 from autobyteus.llm.utils.messages import Message, MessageRole
 from autobyteus.llm.utils.response_types import CompleteResponse
-from autobyteus.llm.user_message import LLMUserMessage
 
 if TYPE_CHECKING:
     from autobyteus.llm.base_llm import BaseLLM
@@ -16,36 +15,54 @@ logger = logging.getLogger(__name__)
 class TokenUsageTrackingExtension(LLMExtension):
     """
     Extension that tracks and monitors token usage and associated costs for LLM interactions.
+    When no token counter is available for the provider, the extension operates in disabled mode
+    and returns zero/None for all metrics.
     """
 
     def __init__(self, llm: "BaseLLM"):
         super().__init__(llm)
         self.token_counter = get_token_counter(llm.model, llm)
-        self.usage_tracker = TokenUsageTracker(llm.model, self.token_counter)
+        if self.token_counter is not None:
+            self.usage_tracker = TokenUsageTracker(llm.model, self.token_counter)
+        else:
+            self.usage_tracker = None
         self._latest_usage: Optional[TokenUsage] = None
 
+    @property
+    def is_enabled(self) -> bool:
+        """Check if token tracking is enabled (token counter is available)."""
+        return self.token_counter is not None
+
     @property
     def latest_token_usage(self) -> Optional[TokenUsage]:
         """Get the latest token usage information."""
         return self._latest_usage
 
     async def before_invoke(
-        self, user_message: LLMUserMessage, **kwargs
+        self, messages: List[Message], rendered_payload: Optional[Any] = None, **kwargs
     ) -> None:
-        pass
+        if not self.is_enabled:
+            return
+        if not messages:
+            logger.warning("TokenUsageTrackingExtension.before_invoke received empty messages list.")
+            return
+        self.usage_tracker.calculate_input_messages(messages)
 
     async def after_invoke(
-        self, user_message: LLMUserMessage, response: CompleteResponse = None, **kwargs
+        self, messages: List[Message], response: CompleteResponse = None, **kwargs
     ) -> None:
         """
         Get the latest usage from tracker and optionally override token counts with provider's usage if available
         """
+        if not self.is_enabled:
+            return
+
         latest_usage = self.usage_tracker.get_latest_usage()
     
         if latest_usage is None:
             logger.warning(
                 "No token usage record found in after_invoke. This may indicate the LLM implementation "
-                "did not call add_user_message. Skipping token usage update for this call."
+                "did not call before_invoke. Skipping token usage update for this call."
             )
             return
 
@@ -54,36 +71,46 @@ class TokenUsageTrackingExtension(LLMExtension):
             latest_usage.prompt_tokens = response.usage.prompt_tokens
             latest_usage.completion_tokens = response.usage.completion_tokens
             latest_usage.total_tokens = response.usage.total_tokens
-            
+        elif isinstance(response, CompleteResponse) and response.content:
+            # Fallback: estimate completion tokens from response content
+            assistant_message = Message(
+                role=MessageRole.ASSISTANT,
+                content=response.content,
+                reasoning_content=response.reasoning,
+            )
+            latest_usage.completion_tokens = self.token_counter.count_output_tokens(assistant_message)
+            latest_usage.total_tokens = latest_usage.prompt_tokens + latest_usage.completion_tokens
+
         # Always calculate costs using current token counts
         latest_usage.prompt_cost = self.usage_tracker.calculate_cost(
             latest_usage.prompt_tokens, True)
         latest_usage.completion_cost = self.usage_tracker.calculate_cost(
             latest_usage.completion_tokens, False)
         latest_usage.total_cost = latest_usage.prompt_cost + latest_usage.completion_cost
-                
-        self._latest_usage = latest_usage
 
-    def on_user_message_added(self, message: Message) -> None:
-        """Track usage whenever a user message is added. Here input message argument is not used, because the input token counts should consider all the input messages"""
-        self.usage_tracker.calculate_input_messages(self.llm.messages)
-
-    def on_assistant_message_added(self, message: Message) -> None:
-        """Track usage whenever an assistant message is added."""
-        self.usage_tracker.calculate_output_message(message)
+        self._latest_usage = latest_usage
 
     def get_total_cost(self) -> float:
+        if not self.is_enabled:
+            return 0.0
         return self.usage_tracker.get_total_cost()
 
     def get_usage_history(self) -> List[TokenUsage]:
+        if not self.is_enabled:
+            return []
         return self.usage_tracker.get_usage_history()
 
     def get_total_input_tokens(self) -> int:
+        if not self.is_enabled:
+            return 0
         return self.usage_tracker.get_total_input_tokens()
 
     def get_total_output_tokens(self) -> int:
+        if not self.is_enabled:
+            return 0
         return self.usage_tracker.get_total_output_tokens()
 
     async def cleanup(self):
-        self.usage_tracker.clear_history()
+        if self.usage_tracker is not None:
+            self.usage_tracker.clear_history()
         self._latest_usage = None