[Tool Call] Stream DeepSeek-V3.2 function call parameters in JSON format. (#16091)

Co-authored-by: Huixxi <uestc.hugo@gmail.com>

python/sglang/srt/function_call/deepseekv32_detector.py

@@ -85,6 +85,7 @@ class DeepSeekV32Detector(BaseFormatDetector):
             r'<｜DSML｜invoke\s+name="([^"]+)"\s*>(.*?)(</｜DSML｜invoke>|$)'
         )
         self.prefix_parameter_end_call = ["</", "｜DSML｜", "parameter"]
+        self.prefix_invoke_end_call = ["</", "｜DSML｜", "inv", "oke"]
         self.current_tool_id = -1
 
     def has_tool_call(self, text: str) -> bool:
@@ -93,9 +94,9 @@ class DeepSeekV32Detector(BaseFormatDetector):
 
     def _parse_parameters_from_xml(
         self, invoke_content: str, allow_partial: bool = False
-    ) -> dict:
+    ) -> str:
         """
-        Parse parameters from either XML-like format or JSON format to dict.
+        Parse parameters from either XML-like format or JSON format to str.
 
         Supports two formats:
         1. XML parameter tags: <｜DSML｜parameter name="..." string="...">value</｜DSML｜parameter>
@@ -103,17 +104,14 @@ class DeepSeekV32Detector(BaseFormatDetector):
         """
         # First, try to parse as direct JSON (new format)
         invoke_content_stripped = invoke_content.strip()
-
+        if invoke_content_stripped.startswith("{"):
-        if invoke_content_stripped.startswith("{") and invoke_content_stripped.endswith(
+            if allow_partial:
-            "}"
+                # Remove incomplete invoke end call prefix in case they are captured by param
-        ):
+                for token in reversed(self.prefix_invoke_end_call):
-            try:
+                    invoke_content_stripped = invoke_content_stripped.rstrip(token)
-                parameters = json.loads(invoke_content_stripped)
+                return invoke_content_stripped
-                if isinstance(parameters, dict):
+            elif invoke_content_stripped.endswith("}"):
-                    return parameters
+                return invoke_content_stripped
-            except (json.JSONDecodeError, ValueError):
-                # If JSON parsing fails, fall through to XML parsing
-                pass
 
         # Fall back to XML parameter tag parsing (original format)
         parameters = {}
@@ -165,7 +163,7 @@ class DeepSeekV32Detector(BaseFormatDetector):
                     except json.JSONDecodeError:
                         parameters[param_name] = param_value.strip()
 
-        return parameters
+        return json.dumps(parameters, ensure_ascii=False)
 
     def detect_and_parse(self, text: str, tools: list[Tool]) -> StreamingParseResult:
         """
@@ -202,7 +200,7 @@ class DeepSeekV32Detector(BaseFormatDetector):
                 # Parse parameters from XML format
                 func_args = self._parse_parameters_from_xml(invoke_content)
                 # construct match_result for parse_base_json
-                match_result = {"name": func_name, "parameters": func_args}
+                match_result = {"name": func_name, "parameters": json.loads(func_args)}
                 calls.extend(self.parse_base_json(match_result, tools))
 
             return StreamingParseResult(normal_text=normal_text, calls=calls)
@@ -288,7 +286,6 @@ class DeepSeekV32Detector(BaseFormatDetector):
                 current_params = self._parse_parameters_from_xml(
                     invoke_content, allow_partial=not is_tool_end
                 )
-                current_args_json = json.dumps(current_params, ensure_ascii=False)
 
                 # 3. Calculate and send incremental arguments
                 sent_len = len(self.streamed_args_for_tool[self.current_tool_id])
@@ -300,12 +297,11 @@ class DeepSeekV32Detector(BaseFormatDetector):
 
                 if is_tool_end:
                     # If complete, send everything remaining
-                    argument_diff = current_args_json[sent_len:]
+                    argument_diff = current_params[sent_len:]
                 elif prev_params is not None:
                     # If partial, send stable prefix diff
-                    prev_args_json = json.dumps(prev_params, ensure_ascii=False)
+                    if current_params != prev_params:
-                    if current_args_json != prev_args_json:
+                        prefix = _find_common_prefix(current_params, prev_params)
-                        prefix = _find_common_prefix(prev_args_json, current_args_json)
                         if len(prefix) > sent_len:
                             argument_diff = prefix[sent_len:]
 
@@ -353,5 +349,5 @@ class DeepSeekV32Detector(BaseFormatDetector):
         return lambda name: StructureInfo(
             begin=f'<｜DSML｜invoke name="{name}">',
             end="</｜DSML｜invoke>",
-            trigger=f"<｜DSML｜invoke",
+            trigger="<｜DSML｜invoke",
         )

test/registered/function_call/test_function_call_parser.py

@@ -1391,9 +1391,11 @@ class TestDeepSeekV32Detector(unittest.TestCase):
 
         tool_calls_by_index = {}
 
+        num_tool_call_chunks = 0
         for chunk in chunks:
             result = self.detector.parse_streaming_increment(chunk, self.tools)
             for call in result.calls:
+                num_tool_call_chunks += 1
                 if call.tool_index is not None:
                     if call.tool_index not in tool_calls_by_index:
                         tool_calls_by_index[call.tool_index] = {
@@ -1408,6 +1410,8 @@ class TestDeepSeekV32Detector(unittest.TestCase):
                             "parameters"
                         ] += call.parameters
 
+        self.assertGreater(num_tool_call_chunks, 8)
+
         self.assertEqual(len(tool_calls_by_index), 1)
         self.assertEqual(tool_calls_by_index[0]["name"], "get_favorite_tourist_spot")
         params = json.loads(tool_calls_by_index[0]["parameters"])
@@ -1422,7 +1426,13 @@ class TestDeepSeekV32Detector(unittest.TestCase):
         text = """<｜DSML｜function_calls>
             <｜DSML｜invoke name="get_favorite_tourist_spot">
             {
-                "city": "San Francisco"
+                "city": "San Francisco",
+                "another_city": "London",
+                "topn": 10,
+                "obj": {
+                    "name": "John",
+                    "age": 30
+                }
             }
             </｜DSML｜invoke>
         </｜DSML｜function_calls>"""
@@ -1436,9 +1446,11 @@ class TestDeepSeekV32Detector(unittest.TestCase):
 
         tool_calls_by_index = {}
 
+        num_tool_call_chunks = 0
         for chunk in chunks:
             result = self.detector.parse_streaming_increment(chunk, self.tools)
             for call in result.calls:
+                num_tool_call_chunks += 1
                 if call.tool_index is not None:
                     if call.tool_index not in tool_calls_by_index:
                         tool_calls_by_index[call.tool_index] = {
@@ -1453,6 +1465,7 @@ class TestDeepSeekV32Detector(unittest.TestCase):
                             "parameters"
                         ] += call.parameters
 
+        self.assertGreater(num_tool_call_chunks, 8)
         self.assertEqual(len(tool_calls_by_index), 1)
         self.assertEqual(tool_calls_by_index[0]["name"], "get_favorite_tourist_spot")