test_qwen3coder_tool_parser.py 29.4 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project

import json
from collections.abc import Generator
from typing import Optional

import pytest

from vllm.entrypoints.openai.protocol import (ChatCompletionRequest,
                                              ChatCompletionToolsParam,
                                              DeltaMessage, FunctionCall,
                                              ToolCall)
from vllm.entrypoints.openai.tool_parsers.qwen3coder_tool_parser import (
    Qwen3CoderToolParser)
16
17
from vllm.entrypoints.openai.tool_parsers.qwen3xml_tool_parser import (
    Qwen3XMLToolParser)
18
from vllm.transformers_utils.detokenizer_utils import detokenize_incrementally
19
20
from vllm.transformers_utils.tokenizer import AnyTokenizer, get_tokenizer

21
MODEL = "Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8"
22
23
24
25
26
27
28
29
30
31
32
33


@pytest.fixture(scope="module")
def qwen3_tokenizer():
    return get_tokenizer(tokenizer_name=MODEL)


@pytest.fixture
def qwen3_tool_parser(qwen3_tokenizer):
    return Qwen3CoderToolParser(qwen3_tokenizer)


34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
@pytest.fixture
def qwen3_xml_tool_parser(qwen3_tokenizer):
    return Qwen3XMLToolParser(qwen3_tokenizer)


@pytest.fixture(params=["original", "xml"])
def qwen3_tool_parser_parametrized(qwen3_tool_parser, qwen3_xml_tool_parser,
                                   request):
    """Parameterized fixture that provides both parser types for testing"""
    if request.param == "original":
        return qwen3_tool_parser
    else:
        return qwen3_xml_tool_parser


49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
@pytest.fixture
def sample_tools():
    return [
        ChatCompletionToolsParam(type="function",
                                 function={
                                     "name": "get_current_weather",
                                     "description": "Get the current weather",
                                     "parameters": {
                                         "type": "object",
                                         "properties": {
                                             "city": {
                                                 "type": "string",
                                                 "description": "The city name"
                                             },
                                             "state": {
                                                 "type": "string",
                                                 "description":
                                                 "The state code"
                                             },
                                             "unit": {
                                                 "type": "string",
                                                 "enum":
                                                 ["fahrenheit", "celsius"]
                                             }
                                         },
                                         "required": ["city", "state"]
                                     }
                                 }),
        ChatCompletionToolsParam(type="function",
                                 function={
                                     "name": "calculate_area",
                                     "description":
                                     "Calculate area of a shape",
                                     "parameters": {
                                         "type": "object",
                                         "properties": {
                                             "shape": {
                                                 "type": "string"
                                             },
                                             "dimensions": {
                                                 "type": "object"
                                             },
                                             "precision": {
                                                 "type": "integer"
                                             }
                                         }
                                     }
                                 })
    ]


def assert_tool_calls(actual_tool_calls: list[ToolCall],
                      expected_tool_calls: list[ToolCall]):
    assert len(actual_tool_calls) == len(expected_tool_calls)

    for actual_tool_call, expected_tool_call in zip(actual_tool_calls,
                                                    expected_tool_calls):
        # Qwen3 parser doesn't generate IDs during extraction
        assert actual_tool_call.type == "function"
        assert (
            actual_tool_call.function.name == expected_tool_call.function.name)
        assert (json.loads(actual_tool_call.function.arguments) == json.loads(
            expected_tool_call.function.arguments))


def stream_delta_message_generator(
115
    qwen3_tool_parser,
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
    qwen3_tokenizer: AnyTokenizer,
    model_output: str,
    request: Optional[ChatCompletionRequest] = None
) -> Generator[DeltaMessage, None, None]:
    all_token_ids = qwen3_tokenizer.encode(model_output,
                                           add_special_tokens=False)

    previous_text = ""
    previous_tokens = None
    prefix_offset = 0
    read_offset = 0
    for i, delta_token in enumerate(all_token_ids):
        delta_token_ids = [delta_token]
        previous_token_ids = all_token_ids[:i]
        current_token_ids = all_token_ids[:i + 1]

        (new_tokens, delta_text, new_prefix_offset,
         new_read_offset) = detokenize_incrementally(
             tokenizer=qwen3_tokenizer,
             all_input_ids=current_token_ids,
             prev_tokens=previous_tokens,
             prefix_offset=prefix_offset,
             read_offset=read_offset,
             skip_special_tokens=False,
             spaces_between_special_tokens=True,
         )

        current_text = previous_text + delta_text

        delta_message = qwen3_tool_parser.extract_tool_calls_streaming(
            previous_text,
            current_text,
            delta_text,
            previous_token_ids,
            current_token_ids,
            delta_token_ids,
            request=request,
        )
        if delta_message:
            yield delta_message

        previous_text = current_text
        previous_tokens = (previous_tokens +
                           new_tokens if previous_tokens else new_tokens)
        prefix_offset = new_prefix_offset
        read_offset = new_read_offset


164
def test_extract_tool_calls_no_tools(qwen3_tool_parser_parametrized):
165
    model_output = "This is a test response without any tool calls"
166
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
        model_output, request=None)  # type: ignore[arg-type]
    assert not extracted_tool_calls.tools_called
    assert extracted_tool_calls.tool_calls == []
    assert extracted_tool_calls.content == model_output


@pytest.mark.parametrize(
    ids=[
        "single_tool",
        "single_tool_with_content",
        "single_tool_multiline_param",
        "parallel_tools",
        "tool_with_typed_params",
    ],
    argnames=["model_output", "expected_tool_calls", "expected_content"],
    argvalues=[
        ('''<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      })))
        ], None),
        ('''Sure! Let me check the weather for you.<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      })))
        ], "Sure! Let me check the weather for you."),
        ('''<tool_call>
<function=calculate_area>
<parameter=shape>
rectangle
</parameter>
<parameter=dimensions>
{"width": 10, 
 "height": 20}
</parameter>
<parameter=precision>
2
</parameter>
</function>
</tool_call>''', [
            ToolCall(function=FunctionCall(name="calculate_area",
                                           arguments=json.dumps({
                                               "shape": "rectangle",
                                               "dimensions": {
                                                   "width": 10,
                                                   "height": 20
                                               },
                                               "precision": 2
                                           })))
        ], None),
        ('''<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>
<tool_call>
<function=get_current_weather>
<parameter=city>
Orlando
</parameter>
<parameter=state>
FL
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      }))),
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Orlando",
                                          "state": "FL",
                                          "unit": "fahrenheit"
                                      })))
        ], None),
        ('''Let me calculate that area for you.<tool_call>
<function=calculate_area>
<parameter=shape>
circle
</parameter>
<parameter=dimensions>
{"radius": 15.5}
</parameter>
<parameter=precision>
3
</parameter>
</function>
</tool_call>''', [
            ToolCall(function=FunctionCall(name="calculate_area",
                                           arguments=json.dumps({
                                               "shape": "circle",
                                               "dimensions": {
                                                   "radius": 15.5
                                               },
                                               "precision": 3
                                           })))
        ], "Let me calculate that area for you."),
    ],
)
314
315
316
def test_extract_tool_calls(qwen3_tool_parser_parametrized, sample_tools,
                            model_output, expected_tool_calls,
                            expected_content):
317
318
319
    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)
320
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
321
322
323
324
325
326
327
328
        model_output, request=request)
    assert extracted_tool_calls.tools_called

    assert_tool_calls(extracted_tool_calls.tool_calls, expected_tool_calls)

    assert extracted_tool_calls.content == expected_content


329
330
def test_extract_tool_calls_fallback_no_tags(qwen3_tool_parser_parametrized,
                                             sample_tools):
331
332
333
334
335
336
337
338
339
340
341
342
343
    """Test fallback parsing when XML tags are missing"""
    model_output = '''<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
</function>'''

    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)
344
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
345
346
347
348
349
350
351
352
        model_output, request=request)

    assert extracted_tool_calls.tools_called
    assert len(extracted_tool_calls.tool_calls) == 1
    assert (extracted_tool_calls.tool_calls[0].function.name ==
            "get_current_weather")


353
def test_extract_tool_calls_type_conversion(qwen3_tool_parser_parametrized):
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
    """Test parameter type conversion based on tool schema"""
    tools = [
        ChatCompletionToolsParam(type="function",
                                 function={
                                     "name": "test_types",
                                     "parameters": {
                                         "type": "object",
                                         "properties": {
                                             "int_param": {
                                                 "type": "integer"
                                             },
                                             "float_param": {
                                                 "type": "float"
                                             },
                                             "bool_param": {
                                                 "type": "boolean"
                                             },
                                             "str_param": {
                                                 "type": "string"
                                             },
                                             "obj_param": {
                                                 "type": "object"
                                             }
                                         }
                                     }
                                 })
    ]

    model_output = '''<tool_call>
<function=test_types>
<parameter=int_param>
42
</parameter>
<parameter=float_param>
3.14
</parameter>
<parameter=bool_param>
true
</parameter>
<parameter=str_param>
hello world
</parameter>
<parameter=obj_param>
{"key": "value"}
</parameter>
</function>
</tool_call>'''

    request = ChatCompletionRequest(model=MODEL, messages=[], tools=tools)
403
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
        model_output, request=request)

    args = json.loads(extracted_tool_calls.tool_calls[0].function.arguments)
    assert args["int_param"] == 42
    assert args["float_param"] == 3.14
    assert args["bool_param"] is True
    assert args["str_param"] == "hello world"
    assert args["obj_param"] == {"key": "value"}


@pytest.mark.parametrize(
    ids=[
        "no_tools",
        "single_tool",
        "single_tool_with_content",
419
        "single_tool_multiline_param",
420
        "parallel_tools",
421
        "tool_with_typed_params",  # Added this test case
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
    ],
    argnames=["model_output", "expected_tool_calls", "expected_content"],
    argvalues=[
        ("This is a test without tools", [], "This is a test without tools"),
        ('''<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      })))
446
        ], None),
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
        ('''Sure! Let me check the weather for you.<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      })))
        ], "Sure! Let me check the weather for you."),
        ('''<tool_call>
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
<function=calculate_area>
<parameter=shape>
rectangle
</parameter>
<parameter=dimensions>
{"width": 10, 
 "height": 20}
</parameter>
<parameter=precision>
2
</parameter>
</function>
</tool_call>''', [
            ToolCall(function=FunctionCall(name="calculate_area",
                                           arguments=json.dumps({
                                               "shape": "rectangle",
                                               "dimensions": {
                                                   "width": 10,
                                                   "height": 20
                                               },
                                               "precision": 2
                                           })))
        ], None),
        ('''<tool_call>
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>
<tool_call>
<function=get_current_weather>
<parameter=city>
Orlando
</parameter>
<parameter=state>
FL
</parameter>
<parameter=unit>
celsius
</parameter>
</function>
</tool_call>''', [
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Dallas",
                                          "state": "TX",
                                          "unit": "fahrenheit"
                                      }))),
            ToolCall(
                function=FunctionCall(name="get_current_weather",
                                      arguments=json.dumps({
                                          "city": "Orlando",
                                          "state": "FL",
                                          "unit": "celsius"
                                      })))
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
        ], None),
        # Added tool_with_typed_params test case
        ('''Let me calculate that area for you.<tool_call>
<function=calculate_area>
<parameter=shape>
circle
</parameter>
<parameter=dimensions>
{"radius": 15.5}
</parameter>
<parameter=precision>
3
</parameter>
</function>
</tool_call>''', [
            ToolCall(function=FunctionCall(name="calculate_area",
                                           arguments=json.dumps({
                                               "shape": "circle",
                                               "dimensions": {
                                                   "radius": 15.5
                                               },
                                               "precision": 3
                                           })))
        ], "Let me calculate that area for you."),
556
557
    ],
)
558
559
560
561
def test_extract_tool_calls_streaming(qwen3_tool_parser_parametrized,
                                      qwen3_tokenizer, sample_tools,
                                      model_output, expected_tool_calls,
                                      expected_content):
562
    """Test incremental streaming behavior including typed parameters"""
563
564
565
566
567
568
569
570
    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)

    other_content = ''
    tool_states = {}  # Track state per tool index

    for delta_message in stream_delta_message_generator(
571
572
            qwen3_tool_parser_parametrized, qwen3_tokenizer, model_output,
            request):
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
        # role should never be streamed from tool parser
        assert not delta_message.role

        if delta_message.content:
            other_content += delta_message.content

        if delta_message.tool_calls:
            for tool_call in delta_message.tool_calls:
                idx = tool_call.index

                # Initialize state for new tool
                if idx not in tool_states:
                    tool_states[idx] = {
                        "id": None,
                        "name": None,
                        "arguments": "",
                        "type": None
                    }

                # First chunk should have id, name, and type
                if tool_call.id:
                    tool_states[idx]["id"] = tool_call.id

                if tool_call.type:
                    assert tool_call.type == "function"
                    tool_states[idx]["type"] = tool_call.type

                if tool_call.function:
                    if tool_call.function.name:
                        # Should only be set once
                        assert tool_states[idx]["name"] is None
                        tool_states[idx]["name"] = tool_call.function.name

                    if tool_call.function.arguments is not None:
                        # Accumulate arguments incrementally
                        tool_states[idx][
                            "arguments"] += tool_call.function.arguments

    # Verify final content
612
    assert other_content == (expected_content or "")  # Handle None case
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631

    # Verify we got all expected tool calls
    assert len(tool_states) == len(expected_tool_calls)

    # Verify each tool call
    for idx, expected_tool in enumerate(expected_tool_calls):
        state = tool_states[idx]
        assert state["id"] is not None
        assert state["type"] == "function"
        assert state["name"] == expected_tool.function.name

        # Parse accumulated arguments
        arguments_str = state["arguments"]
        assert arguments_str is not None
        actual_args = json.loads(arguments_str)
        expected_args = json.loads(expected_tool.function.arguments)
        assert actual_args == expected_args


632
def test_extract_tool_calls_missing_closing_parameter_tag(
633
        qwen3_tool_parser_parametrized, sample_tools):
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
    """Test handling of missing closing </parameter> tag"""
    # Using get_current_weather from sample_tools but with malformed XML
    model_output = '''Let me check the weather for you:
<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>'''

    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)
653
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
        model_output, request=request)

    # The parser should handle the malformed XML gracefully
    assert extracted_tool_calls.tools_called
    assert len(extracted_tool_calls.tool_calls) == 1

    # Verify the function name is correct
    assert extracted_tool_calls.tool_calls[
        0].function.name == "get_current_weather"

    # Verify the arguments are parsed despite the missing closing tag
    args = json.loads(extracted_tool_calls.tool_calls[0].function.arguments)
    assert "city" in args
    assert args["city"] == "Dallas"
    assert args["state"] == "TX"
    assert args["unit"] == "fahrenheit"

    # Check that content before the tool call is preserved
    assert "Let me check the weather for you:" in extracted_tool_calls.content


def test_extract_tool_calls_streaming_missing_closing_tag(
676
        qwen3_tool_parser_parametrized, qwen3_tokenizer, sample_tools):
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
    """Test streaming with missing closing </parameter> tag"""
    # Using get_current_weather from sample_tools but with malformed XML
    model_output = '''Let me check the weather for you:
<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
<parameter=state>
TX
</parameter>
<parameter=unit>
fahrenheit
</parameter>
</function>
</tool_call>'''

    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)

    other_content = ''
    tool_states = {}

    for delta_message in stream_delta_message_generator(
701
702
            qwen3_tool_parser_parametrized, qwen3_tokenizer, model_output,
            request):
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751

        if delta_message.content:
            other_content += delta_message.content

        if delta_message.tool_calls:
            for tool_call in delta_message.tool_calls:
                idx = tool_call.index

                if idx not in tool_states:
                    tool_states[idx] = {
                        "id": None,
                        "name": None,
                        "arguments": "",
                        "type": None
                    }

                if tool_call.id:
                    tool_states[idx]["id"] = tool_call.id

                if tool_call.type:
                    assert tool_call.type == "function"
                    tool_states[idx]["type"] = tool_call.type

                if tool_call.function:
                    if tool_call.function.name:
                        tool_states[idx]["name"] = tool_call.function.name

                    if tool_call.function.arguments is not None:
                        tool_states[idx][
                            "arguments"] += tool_call.function.arguments

    # Verify content was streamed
    assert "Let me check the weather for you:" in other_content

    # Verify we got the tool call
    assert len(tool_states) == 1
    state = tool_states[0]
    assert state["id"] is not None
    assert state["type"] == "function"
    assert state["name"] == "get_current_weather"

    # Verify arguments were parsed correctly despite missing closing tag
    assert state["arguments"] is not None
    args = json.loads(state["arguments"])
    assert args["city"] == "Dallas"
    assert args["state"] == "TX"
    assert args["unit"] == "fahrenheit"


752
753
def test_extract_tool_calls_streaming_incremental(
        qwen3_tool_parser_parametrized, qwen3_tokenizer, sample_tools):
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
    """Test that streaming is truly incremental"""
    model_output = '''I'll check the weather.<tool_call>
<function=get_current_weather>
<parameter=city>
Dallas
</parameter>
<parameter=state>
TX
</parameter>
</function>
</tool_call>'''

    request = ChatCompletionRequest(model=MODEL,
                                    messages=[],
                                    tools=sample_tools)

    chunks = []
    for delta_message in stream_delta_message_generator(
772
773
            qwen3_tool_parser_parametrized, qwen3_tokenizer, model_output,
            request):
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
        chunks.append(delta_message)

    # Should have multiple chunks
    assert len(chunks) > 3

    # First chunk(s) should be content
    assert chunks[0].content is not None
    assert chunks[0].tool_calls is None or chunks[0].tool_calls == []

    # Should have a chunk with tool header (id, name, type)
    header_found = False
    for chunk in chunks:
        if chunk.tool_calls and chunk.tool_calls[0].id:
            header_found = True
            assert (chunk.tool_calls[0].function.name == "get_current_weather")
            assert chunk.tool_calls[0].type == "function"
            # Empty initially
            assert chunk.tool_calls[0].function.arguments == ""
            break
    assert header_found

    # Should have chunks with incremental arguments
    arg_chunks = []
    for chunk in chunks:
        if chunk.tool_calls and chunk.tool_calls[0].function.arguments:
            arg_chunks.append(chunk.tool_calls[0].function.arguments)

    # Arguments should be streamed incrementally
    assert len(arg_chunks) > 1

    # Concatenated arguments should form valid JSON
    full_args = "".join(arg_chunks)
    parsed_args = json.loads(full_args)
    assert parsed_args["city"] == "Dallas"
    assert parsed_args["state"] == "TX"
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854


def test_extract_tool_calls_complex_type_with_single_quote(
        qwen3_tool_parser_parametrized):
    """Test parameter type conversion based on tool schema"""
    tools = [
        ChatCompletionToolsParam(type="function",
                                 function={
                                     "name": "test_types",
                                     "parameters": {
                                         "type": "object",
                                         "properties": {
                                             "int_param": {
                                                 "type": "integer"
                                             },
                                             "float_param": {
                                                 "type": "float"
                                             },
                                             "bool_param": {
                                                 "type": "boolean"
                                             },
                                             "str_param": {
                                                 "type": "string"
                                             },
                                             "obj_param": {
                                                 "type": "object"
                                             }
                                         }
                                     }
                                 })
    ]

    model_output = '''<tool_call>
<function=test_types>
<parameter=obj_param>
{'key': 'value'}
</parameter>
</function>
</tool_call>'''

    request = ChatCompletionRequest(model=MODEL, messages=[], tools=tools)
    extracted_tool_calls = qwen3_tool_parser_parametrized.extract_tool_calls(
        model_output, request=request)

    args = json.loads(extracted_tool_calls.tool_calls[0].function.arguments)
    assert args["obj_param"] == {"key": "value"}