refactor: cleanup per code audit

- Remove unused _ensure_worker_created method (dead code)
- Change tool call logging from debug to trace level for consistency

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>

Author: smhanan

ccproxy/llms/streaming/processors.py

@@ -424,7 +424,7 @@ async def _process_chunk(
             elif self.tool_calls and self.enable_tool_calls:
                 # Send completed tool calls for both SSE and dict formats
                 # Previous bug: Only sent for SSE format, causing dict format (SDK mode) to miss tool calls
-                logger.debug(
+                logger.trace(
                     "openai_stream_sending_tool_calls",
                     tool_count=len(self.tool_calls),
                     output_format=self.output_format,
@@ -434,7 +434,7 @@ async def _process_chunk(
                 for tool_call_index, (tool_call_id, tool_call) in enumerate(
                     self.tool_calls.items()
                 ):
-                    logger.debug(
+                    logger.trace(
                         "openai_stream_tool_call_yielding",
                         tool_call_id=tool_call_id,
                         tool_name=tool_call["name"],
@@ -462,7 +462,7 @@ async def _process_chunk(
                     )
 
                 # Clear tool_calls after yielding to prevent duplicates
-                logger.debug(
+                logger.trace(
                     "openai_stream_clearing_tool_calls",
                     cleared_count=len(self.tool_calls),
                     category="streaming_conversion",

ccproxy/plugins/claude_sdk/stream_handle.py

@@ -158,33 +158,6 @@ async def create_listener(self) -> AsyncIterator[Any]:
             # Check if we should trigger cleanup
             await self._check_cleanup()
 
-    async def _ensure_worker_created(self) -> None:
-        """Ensure the worker is created (but not started).
-
-        Note: Worker is now started in create_listener() AFTER listener
-        registration to prevent race conditions with fast STDIO tools.
-        """
-        async with self._worker_lock:
-            if self._worker is None:
-                # Create worker (but don't start - that happens after listener registration)
-                worker_id = f"{self.handle_id}-worker"
-                self._worker = StreamWorker(
-                    worker_id=worker_id,
-                    message_iterator=self._message_iterator,
-                    session_id=self.session_id,
-                    request_id=self.request_id,
-                    session_client=self._session_client,
-                    stream_handle=self,  # Pass self for message tracking
-                )
-
-                logger.debug(
-                    "stream_handle_worker_created",
-                    handle_id=self.handle_id,
-                    worker_id=worker_id,
-                    session_id=self.session_id,
-                    category="streaming",
-                )
-
     async def _remove_listener(self, listener_id: str) -> None:
         """Remove a listener and clean it up.