fix: get llm token usage add result type

Author: zgqgit

src/backend/bisheng/llm/domain/utils.py

@@ -62,6 +62,11 @@ def parse_token_usage(result: Any) -> tuple[int, int, int, int]:
             output_token += tmp2
             cache_token += tmp3
             total_token += tmp4
+    elif isinstance(result, ChatGenerationChunk):
+        token_usage = result.generation_info.get('token_usage', {})
+        input_token, output_token, cache_token, total_token = get_token_from_usage(token_usage)
+    else:
+        logger.warning(f'unknown result type: {type(result)}')
     return input_token, output_token, cache_token, total_token