Clean up context.py - remove parser error handling, use main branch version

Author: baonudesifeizhai

vllm/entrypoints/context.py

@@ -239,18 +239,11 @@ def append_output(self, output) -> None:
             # beginning of a new message
             self.first_tok_of_message = output.finished
             tok = output.outputs[0].token_ids[0]
-            try:
-                self.parser.process(tok)
-                # Only update these if parsing succeeded
-                self._update_num_output_tokens(output.outputs[0].token_ids)
-                # Check if the current token is part of reasoning content
-                self._update_num_reasoning_tokens([tok])
-                self.last_tok = tok
-            except Exception as e:
-                # Log the error but continue processing to avoid breaking stream
-                logger.warning("Parser error processing token %s: %s", tok, e)
-                # Skip this token to avoid parser state corruption
-                # Don't update last_tok or token counts for failed tokens
+            self.parser.process(tok)
+            self._update_num_output_tokens(output.outputs[0].token_ids)
+            # Check if the current token is part of reasoning content
+            self._update_num_reasoning_tokens([tok])
+            self.last_tok = tok
         else:
             # Handle the case of tool output in direct message format
             assert len(output) == 1, "Tool output should be a single message"