feat: rust sglang server openai apis (#33103)

Co-authored-by: Rain Jiang <rain-jiang@outlook.com>

test/registered/openai_server/function_call/test_openai_function_calling.py

@@ -11,6 +11,7 @@ from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
+    is_rust_server_built,
     popen_launch_server,
 )
 
@@ -62,7 +63,9 @@ class TestOpenAIServerFunctionCalling(CustomTestCase):
     def test_function_calling_format(self):
         """
         Test: Whether the function call format returned by the AI is correct.
-        When returning a tool call, message.content should be None, and tool_calls should be a list.
+        Require a tool call so this tests the response format rather than the
+        model's stochastic decision to call a tool. message.content should be
+        None, and tool_calls should be a list.
         """
         client = openai.Client(api_key=self.api_key, base_url=self.base_url)
 
@@ -102,6 +105,7 @@ class TestOpenAIServerFunctionCalling(CustomTestCase):
             top_p=0.8,
             stream=False,
             tools=tools,
+            tool_choice="required",
         )
 
         tool_calls = response.choices[0].message.tool_calls
@@ -914,6 +918,54 @@ class TestOpenAIPythonicFunctionCalling(CustomTestCase):
         )
 
 
+@unittest.skipUnless(
+    is_rust_server_built(),
+    "embedded rust server extension not built",
+)
+class TestOpenAIFunctionCallingWithRust(TestOpenAIServerFunctionCalling):
+    """Run the registered unary/streaming function-call suite through Rust."""
+
+    @classmethod
+    def setUpClass(cls):
+        cls.model = DEFAULT_SMALL_MODEL_NAME_FOR_TEST
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.api_key = "sk-123456"
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            api_key=cls.api_key,
+            other_args=["--tool-call-parser", "llama3"],
+            env={"SGLANG_RUST_SERVER": "1"},
+        )
+        cls.base_url += "/v1"
+        cls.tokenizer = get_tokenizer(cls.model)
+
+
+@unittest.skipUnless(
+    is_rust_server_built(),
+    "embedded rust server extension not built",
+)
+class TestOpenAIPythonicFunctionCallingWithRust(TestOpenAIPythonicFunctionCalling):
+    """Run Pythonic unary/streaming tool calls through Rust."""
+
+    @classmethod
+    def setUpClass(cls):
+        cls.model = DEFAULT_SMALL_MODEL_NAME_FOR_TEST
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.api_key = "sk-123456"
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            api_key=cls.api_key,
+            other_args=["--tool-call-parser", "pythonic"],
+            env={"SGLANG_RUST_SERVER": "1"},
+        )
+        cls.base_url += "/v1"
+        cls.tokenizer = get_tokenizer(cls.model)
+
+
 # Skip for ci test
 # class TestGLM45ServerFunctionCalling(TestOpenAIServerFunctionCalling):
 #     @classmethod