diff --git a/tests/parser/test_streaming.py b/tests/parser/test_streaming.py index 6830d694e1..2ba2392f8e 100644 --- a/tests/parser/test_streaming.py +++ b/tests/parser/test_streaming.py @@ -58,7 +58,11 @@ def stream_text(parser, tokenizer, text, request, prompt_token_ids=None): for tid in token_ids: delta_text = tokenizer.decode([tid]) result = parser.parse_delta( - delta_text, [tid], request, prompt_token_ids=prompt_token_ids + delta_text, + [tid], + request, + prompt_token_ids=prompt_token_ids, + finished=False, ) prompt_token_ids = None results.append(result) @@ -146,7 +150,11 @@ def stream_chunks(parser, tokenizer, chunks, request_obj): for chunk in chunks: delta_text = tokenizer.decode(chunk) result = parser.parse_delta( - delta_text, chunk, request_obj, prompt_token_ids=prompt_token_ids + delta_text, + chunk, + request_obj, + prompt_token_ids=prompt_token_ids, + finished=False, ) prompt_token_ids = None results.append(result) diff --git a/vllm/entrypoints/openai/responses/serving.py b/vllm/entrypoints/openai/responses/serving.py index bb700cd7dd..eee02707a9 100644 --- a/vllm/entrypoints/openai/responses/serving.py +++ b/vllm/entrypoints/openai/responses/serving.py @@ -1408,6 +1408,7 @@ class OpenAIServingResponses(OpenAIServing): delta_token_ids=delta_token_ids, request=request, prompt_token_ids=ctx.last_output.prompt_token_ids, + finished=output.finish_reason is not None, ) else: delta_message = DeltaMessage(content=output.text) diff --git a/vllm/parser/abstract_parser.py b/vllm/parser/abstract_parser.py index 317f486e79..9e4d1830b4 100644 --- a/vllm/parser/abstract_parser.py +++ b/vllm/parser/abstract_parser.py @@ -344,7 +344,8 @@ class Parser: delta_token_ids: list[int], request: ChatCompletionRequest | ResponsesRequest, prompt_token_ids: list[int] | None = None, - finished: bool = False, + *, + finished: bool, ) -> DeltaMessage | None: """Parse a single streaming delta, orchestrating reasoning then tool call extraction via internal stream state. @@ -809,7 +810,8 @@ class DelegatingParser(Parser): delta_token_ids: list[int], request: ChatCompletionRequest | ResponsesRequest, prompt_token_ids: list[int] | None = None, - finished: bool = False, + *, + finished: bool, ) -> DeltaMessage | None: state = self._stream_state