File size: 9,389 Bytes
59b2038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d408a51
 
 
 
 
 
 
 
 
 
 
 
 
 
59b2038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
754345f
59b2038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d408a51
6155b26
d408a51
 
 
 
 
 
 
6155b26
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d408a51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6155b26
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d408a51
 
754345f
d408a51
 
 
 
 
 
 
 
 
6155b26
d408a51
 
 
 
 
 
 
 
 
6155b26
 
 
 
d408a51
 
59b2038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
754345f
59b2038
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
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
115
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
164
165
166
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
from types import SimpleNamespace

import pytest
from litellm import ChatCompletionMessageToolCall, Message

from agent.core import agent_loop
from agent.core.agent_loop import (
    LLMResult,
    _call_llm_streaming,
    _assistant_message_from_result,
    _extract_thinking_state,
)


def test_extract_thinking_state_from_litellm_message():
    message = Message(
        role="assistant",
        content="working",
        thinking_blocks=[{"type": "thinking", "thinking": "reasoned"}],
        reasoning_content="reasoned",
    )

    thinking_blocks, reasoning_content = _extract_thinking_state(message)

    assert thinking_blocks == [{"type": "thinking", "thinking": "reasoned"}]
    assert reasoning_content == "reasoned"


def test_extract_thinking_state_from_provider_fields():
    message = SimpleNamespace(
        provider_specific_fields={
            "thinking_blocks": [{"type": "thinking", "thinking": "reasoned"}],
            "reasoning_content": "reasoned",
        },
    )

    thinking_blocks, reasoning_content = _extract_thinking_state(message)

    assert thinking_blocks == [{"type": "thinking", "thinking": "reasoned"}]
    assert reasoning_content == "reasoned"


def test_assistant_message_from_result_preserves_thinking_with_tool_calls():
    tool_call = ChatCompletionMessageToolCall(
        id="call_1",
        type="function",
        function={"name": "bash", "arguments": '{"command": "date"}'},
    )
    result = LLMResult(
        content=None,
        tool_calls_acc={},
        token_count=12,
        finish_reason="tool_calls",
        thinking_blocks=[{"type": "thinking", "thinking": "reasoned"}],
        reasoning_content="reasoned",
    )

    message = _assistant_message_from_result(
        result,
        model_name="anthropic/claude-opus-4-6",
        tool_calls=[tool_call],
    )

    assert message.tool_calls == [tool_call]
    assert message.thinking_blocks == [{"type": "thinking", "thinking": "reasoned"}]
    assert message.reasoning_content == "reasoned"


def test_assistant_message_from_result_strips_non_anthropic_reasoning_content():
    result = LLMResult(
        content=None,
        tool_calls_acc={},
        token_count=12,
        finish_reason="tool_calls",
        thinking_blocks=[{"type": "thinking", "thinking": "reasoned"}],
        reasoning_content="reasoned",
    )

    message = _assistant_message_from_result(
        result,
        model_name="openai/Qwen/Qwen3-Next-80B-A3B-Instruct",
    )

    assert getattr(message, "thinking_blocks", None) is None
    assert getattr(message, "reasoning_content", None) is None


def test_assistant_message_from_result_omits_absent_thinking_fields():
    result = LLMResult(
        content="done",
        tool_calls_acc={},
        token_count=12,
        finish_reason="stop",
    )

    message = _assistant_message_from_result(
        result,
        model_name="anthropic/claude-opus-4-6",
    )

    assert message.content == "done"
    assert getattr(message, "thinking_blocks", None) is None
    assert getattr(message, "reasoning_content", None) is None


@pytest.mark.asyncio
async def test_streaming_call_rebuilds_anthropic_thinking_state(monkeypatch):
    async def fake_stream():
        yield SimpleNamespace(
            choices=[
                SimpleNamespace(
                    delta=SimpleNamespace(content="done", tool_calls=None),
                    finish_reason="stop",
                )
            ],
        )
        yield SimpleNamespace(choices=[], usage=SimpleNamespace(total_tokens=3))

    async def fake_acompletion(**_kwargs):
        return fake_stream()

    def fake_chunk_builder(chunks, **_kwargs):
        assert len(chunks) == 2
        return SimpleNamespace(
            choices=[
                SimpleNamespace(
                    message=Message(
                        role="assistant",
                        content="done",
                        thinking_blocks=[{"type": "thinking", "thinking": "reasoned"}],
                        reasoning_content="reasoned",
                    )
                )
            ]
        )

    events = []

    async def send_event(event):
        events.append(event)

    session = SimpleNamespace(
        config=SimpleNamespace(model_name="anthropic/claude-opus-4-6"),
        is_cancelled=False,
        send_event=send_event,
    )
    monkeypatch.setattr(agent_loop, "acompletion", fake_acompletion)
    monkeypatch.setattr(agent_loop, "stream_chunk_builder", fake_chunk_builder)

    result = await _call_llm_streaming(
        session,
        messages=[Message(role="user", content="hi")],
        tools=[],
        llm_params={"model": "anthropic/claude-opus-4-6"},
    )

    assert result.content == "done"
    assert result.thinking_blocks == [{"type": "thinking", "thinking": "reasoned"}]
    assert result.reasoning_content == "reasoned"


@pytest.mark.asyncio
async def test_streaming_call_rebuilds_anthropic_delta_thinking_state(monkeypatch):
    async def fake_stream():
        yield SimpleNamespace(
            choices=[
                SimpleNamespace(
                    delta=SimpleNamespace(
                        content=None,
                        tool_calls=None,
                        thinking_blocks=[
                            {
                                "type": "thinking",
                                "thinking": "reasoned",
                                "signature": "",
                            }
                        ],
                    ),
                    finish_reason=None,
                )
            ],
        )
        yield SimpleNamespace(
            choices=[
                SimpleNamespace(
                    delta=SimpleNamespace(
                        content=None,
                        tool_calls=None,
                        thinking_blocks=[
                            {
                                "type": "thinking",
                                "thinking": "",
                                "signature": "signed",
                            }
                        ],
                    ),
                    finish_reason=None,
                )
            ],
        )
        yield SimpleNamespace(
            choices=[
                SimpleNamespace(
                    delta=SimpleNamespace(content="done", tool_calls=None),
                    finish_reason="stop",
                )
            ],
        )
        yield SimpleNamespace(choices=[], usage=SimpleNamespace(total_tokens=3))

    async def fake_acompletion(**_kwargs):
        return fake_stream()

    def fake_chunk_builder(chunks, **_kwargs):
        assert len(chunks) == 4
        return SimpleNamespace(
            choices=[
                SimpleNamespace(
                    message=Message(
                        role="assistant",
                        content="done",
                        thinking_blocks=[
                            {
                                "type": "thinking",
                                "thinking": "reasoned",
                                "signature": "signed",
                            }
                        ],
                        reasoning_content="reasoned",
                    )
                )
            ]
        )

    events = []

    async def send_event(event):
        events.append(event)

    session = SimpleNamespace(
        config=SimpleNamespace(model_name="anthropic/claude-opus-4-7"),
        is_cancelled=False,
        send_event=send_event,
    )
    monkeypatch.setattr(agent_loop, "acompletion", fake_acompletion)
    monkeypatch.setattr(agent_loop, "stream_chunk_builder", fake_chunk_builder)

    result = await _call_llm_streaming(
        session,
        messages=[Message(role="user", content="hi")],
        tools=[],
        llm_params={"model": "anthropic/claude-opus-4-7"},
    )

    assert result.content == "done"
    assert result.thinking_blocks == [
        {"type": "thinking", "thinking": "reasoned", "signature": "signed"}
    ]
    assert result.reasoning_content == "reasoned"


@pytest.mark.asyncio
async def test_streaming_call_skips_chunk_rebuild_for_non_anthropic(monkeypatch):
    async def fake_stream():
        yield SimpleNamespace(
            choices=[
                SimpleNamespace(
                    delta=SimpleNamespace(content="done", tool_calls=None),
                    finish_reason="stop",
                )
            ],
        )

    async def fake_acompletion(**_kwargs):
        return fake_stream()

    def fail_chunk_builder(*_args, **_kwargs):
        raise AssertionError("stream_chunk_builder should not run")

    events = []

    async def send_event(event):
        events.append(event)

    session = SimpleNamespace(
        config=SimpleNamespace(model_name="openai/Qwen/Qwen3"),
        is_cancelled=False,
        send_event=send_event,
    )
    monkeypatch.setattr(agent_loop, "acompletion", fake_acompletion)
    monkeypatch.setattr(agent_loop, "stream_chunk_builder", fail_chunk_builder)

    result = await _call_llm_streaming(
        session,
        messages=[Message(role="user", content="hi")],
        tools=[],
        llm_params={"model": "openai/Qwen/Qwen3"},
    )

    assert result.content == "done"
    assert result.thinking_blocks is None
    assert result.reasoning_content is None