llama4_multi_python_tags.diff

diff --git a/tests/entrypoints/openai/tool_parsers/test_pythonic_tool_parser.py b/tests/entrypoints/openai/tool_parsers/test_pythonic_tool_parser.py
index fbbbc1fb2..1f953706b 100644
--- a/tests/entrypoints/openai/tool_parsers/test_pythonic_tool_parser.py
+++ b/tests/entrypoints/openai/tool_parsers/test_pythonic_tool_parser.py
@@ -52,6 +52,27 @@ ESCAPED_STRING_FUNCTION_CALL = FunctionCall(
     name="get_weather",
     arguments='{"city": "Martha\'s Vineyard", "metric": "\\"cool units\\""}',
 )
+PYTHON_TAGS_FUNCTION_OUTPUT="<|python_start|>[get_weather(city='San Francisco', metric='celsius')]<|python_end|>"
+PYTHON_TAGS_FUNCTION_CALL = FunctionCall(
+    name="get_weather",
+    arguments='{"city": "San Francisco", "metric": "celsius"}',
+)
+PYTHON_TAGS_MULTI_FUNCTION_OUTPUT='<|python_start|>get_boiling_point(liquid_name="polyjuice", celcius=true); get_boiling_point(liquid_name="polyjuice", celcius=false)<|python_end|>'
+PYTHON_TAGS_MULTI_FUNCTION_CALL = [
+    FunctionCall(
+        name="get_boiling_point",
+        arguments='{"liquid_name": "polyjuice", "celcius": true}',
+    ),
+    FunctionCall(
+        name="get_boiling_point",
+        arguments='{"liquid_name": "polyjuice", "celcius": false}',
+    ),
+]
+LOWERCASE_BOOL_FUNCTION_OUTPUT = "has_things(thing1=true, thing2=false)"
+LOWERCASE_BOOL_FUNCTION_CALL = FunctionCall(
+    name="has_things",
+    arguments='{"thing1": true, "thing2": false}',
+)
 
 
 @pytest.mark.parametrize("streaming", [True, False])
@@ -118,6 +139,28 @@ TEST_CASES = [
                  f"[{SIMPLE_FUNCTION_OUTPUT}, {MORE_TYPES_FUNCTION_OUTPUT}]",
                  [SIMPLE_FUNCTION_CALL, MORE_TYPES_FUNCTION_CALL],
                  id="parallel_calls_nonstreaming"),
+    pytest.param(True,
+                 PYTHON_TAGS_FUNCTION_OUTPUT, [PYTHON_TAGS_FUNCTION_CALL],
+                 id="python_tags_streaming"),
+    pytest.param(False,
+                 PYTHON_TAGS_FUNCTION_OUTPUT, [PYTHON_TAGS_FUNCTION_CALL],
+                 id="python_tags_nonstreaming"),
+    pytest.param(True,
+                 PYTHON_TAGS_MULTI_FUNCTION_OUTPUT,
+                 PYTHON_TAGS_MULTI_FUNCTION_CALL,
+                 id="python_tags_multi_streaming"),
+    pytest.param(False,
+                 PYTHON_TAGS_MULTI_FUNCTION_OUTPUT,
+                 PYTHON_TAGS_MULTI_FUNCTION_CALL,
+                 id="python_tags_multi_nonstreaming"),
+    pytest.param(True,
+                 f"[{LOWERCASE_BOOL_FUNCTION_OUTPUT}]",
+                 [LOWERCASE_BOOL_FUNCTION_CALL],
+                 id="lowercase_bool_streaming"),
+    pytest.param(False,
+                 f"[{LOWERCASE_BOOL_FUNCTION_OUTPUT}]",
+                 [LOWERCASE_BOOL_FUNCTION_CALL],
+                 id="lowercase_bool_nonstreaming"),
 ]
 
 
diff --git a/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py b/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
index bb91a35af..717a2ea21 100644
--- a/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
+++ b/vllm/entrypoints/openai/tool_parsers/pythonic_tool_parser.py
@@ -62,6 +62,17 @@ class PythonicToolParser(ToolParser):
         Extract the tool calls from a complete model response.
         """
 
+        # remove <|python_start|> and <|python_end|>
+        # as Llama 4 model sometime will output those tokens
+        if model_output.startswith("<|python_start|>"):
+            model_output = model_output[len("<|python_start|>"):]
+            if not model_output.startswith("["):
+                model_output = "[" + model_output
+            model_output = model_output.replace(");", "),")
+            model_output = model_output.replace("<|python_end|>", "")
+            if not model_output.endswith("]"):
+                model_output += "]"
+
         if not (self.TOOL_CALL_REGEX.match(model_output)):
             return ExtractedToolCallInformation(tools_called=False,
                                                 tool_calls=[],
@@ -100,6 +111,36 @@ class PythonicToolParser(ToolParser):
         request: ChatCompletionRequest,
     ) -> Union[DeltaMessage, None]:
 
+        start_python_tag = "<|python_start|>"
+        end_python_tag = "<|python_end|>"
+
+        # If we start with the entire start python tag, remove it
+        if current_text.startswith(start_python_tag):
+            current_text = current_text[len(start_python_tag):]
+            if not current_text.startswith("["):
+                current_text = "[" + current_text
+            current_text = current_text.replace(");", "),")
+
+        # If we end with the entire end python tag, remove it
+        if current_text.endswith(end_python_tag):
+            current_text = current_text[:-len(end_python_tag)]
+
+        # If we start with part of the start python tag, remove it
+        for i in range(1, len(start_python_tag)):
+            start_tag_substr = start_python_tag[:-i]
+            if current_text.startswith(start_tag_substr):
+                current_text = current_text[len(start_tag_substr):]
+
+        # If we end with part of the end python tag, remove it
+        for i in range(1, len(end_python_tag)):
+            end_tag_substr = end_python_tag[:i]
+            if current_text.endswith(end_tag_substr):
+                current_text = current_text[:-len(end_tag_substr)]
+
+        # If there's nothing left after removing python tags, stop parsing
+        if not current_text:
+            return None
+
         if not current_text.startswith("["):
             return DeltaMessage(content=delta_text)
 
@@ -189,6 +230,8 @@ 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")