pythonic_bools.diff

diff --git a/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py b/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
index bb91a35af..2947bc8db 100644
--- a/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
+++ b/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
@@ -189,6 +189,8 @@ def _get_parameter_value(val: ast.expr) -> Any:
         }
     elif isinstance(val, ast.List):
         return [_get_parameter_value(v) for v in val.elts]
+    elif isinstance(val, ast.Name) and val.id in ["true", "false"]:
+        return True if val.id == "true" else False
     else:
         raise _UnexpectedAstError("Tool call arguments must be literals")

This change is required to get the remote-vllm provider of Llama Stack passing this integration test:

VLLM_URL="http://localhost:8000/v1" \
INFERENCE_MODEL="RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic" \
LLAMA_STACK_CONFIG=remote-vllm \
python -m pytest -v tests/integration/agents/test_agents.py::test_custom_tool \
  --text-model "RedHatAI/Llama-4-Scout-17B-16E-Instruct-FP8-dynamic"

The function call output it's trying to parse is get_boiling_point(liquid_name='polyjuice', celcius=true). It's the true that messes things up, as it was getting treated as a variable name (because it's lowercase) instead of as the Python boolean True value.