test(vllm): isolate streaming parser fixtures

Mock the request constructor so fake parser tests can run on older
vLLM profiles without ParserManager. Document the streaming version
requirement and fallback behavior.

Signed-off-by: pos-ei-don <1822533+pos-ei-don@users.noreply.github.com>
(cherry picked from commit ff0130fa89)

backend/python/vllm/test.py

@@ -762,6 +762,24 @@ class TestStreamingReasoningSplit(unittest.TestCase):
     THINK = ["Let me ", "count the ", "primes.", "</think>"]
     ANSWER = ["There are ", "37 primes."]
 
+    def setUp(self):
+        from types import SimpleNamespace
+        from unittest.mock import patch
+        import backend
+
+        def parser_request(model, messages, tools=None):
+            return SimpleNamespace(
+                model=model, messages=messages,
+                tools=[SimpleNamespace(function=SimpleNamespace(**tool["function"]))
+                       for tool in tools] if tools else None,
+            )
+
+        # These fake-parser tests also run on backend profiles whose vLLM
+        # predates ParserManager and therefore has no _ParserRequest import.
+        request_patch = patch.object(backend, "_ParserRequest", parser_request, create=True)
+        request_patch.start()
+        self.addCleanup(request_patch.stop)
+
     class _DeltaMsg:
         def __init__(self, content=None, reasoning=None, tool_calls=None):
             self.content = content

docs/content/features/openai-functions.md

@@ -84,6 +84,13 @@ When you import a vLLM model through the LocalAI gallery, the importer looks up
 
 Available tool parsers include `hermes`, `llama3_json`, `llama4_pythonic`, `mistral`, `qwen3_xml`, `deepseek_v3`, `granite4`, `kimi_k2`, `glm45`, and more. Available reasoning parsers include `deepseek_r1`, `qwen3`, `mistral`, `gemma4`, `granite`. See the upstream vLLM documentation for the full list.
 
+With `backend: vllm`, vLLM 0.30 or later, and a configured reasoning parser,
+streamed responses separate `delta.reasoning_content`, `delta.content`, and tool
+calls as tokens arrive. The final chunk does not repeat the generated text.
+If parsing fails during a stream, the backend buffers the remaining output and
+sends any matching, unsent suffixes in the final chunk. Older vLLM versions keep
+the previous streaming behavior and can include reasoning in `delta.content`.
+
 ### MLX / MLX-VLM
 
 MLX backends **auto-detect** the right tool parser by inspecting the model's chat template - you don't need to set anything. Just load an MLX-quantized model that was trained with tool support: