File size: 7,009 Bytes
447ebeb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import json
import os
import sys
from datetime import datetime
from unittest.mock import AsyncMock, patch, MagicMock

sys.path.insert(
    0, os.path.abspath("../..")
)  # Adds the parent directory to the system path


import httpx
import pytest
from respx import MockRouter

import litellm
from litellm import Choices, Message, ModelResponse
from base_llm_unit_tests import BaseLLMChatTest, BaseOSeriesModelsTest


@pytest.mark.parametrize("model", ["o1-preview", "o1-mini", "o1"])
@pytest.mark.asyncio
async def test_o1_handle_system_role(model):
    """
    Tests that:
    - max_tokens is translated to 'max_completion_tokens'
    - role 'system' is translated to 'user'
    """
    from openai import AsyncOpenAI
    from litellm.utils import supports_system_messages

    os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
    litellm.model_cost = litellm.get_model_cost_map(url="")

    litellm.set_verbose = True

    client = AsyncOpenAI(api_key="fake-api-key")

    with patch.object(
        client.chat.completions.with_raw_response, "create"
    ) as mock_client:
        try:
            await litellm.acompletion(
                model=model,
                max_tokens=10,
                messages=[{"role": "system", "content": "Be a good bot!"}],
                client=client,
            )
        except Exception as e:
            print(f"Error: {e}")

        mock_client.assert_called_once()
        request_body = mock_client.call_args.kwargs

        print("request_body: ", request_body)

        assert request_body["model"] == model
        assert request_body["max_completion_tokens"] == 10
        if supports_system_messages(model, "openai"):
            assert request_body["messages"] == [
                {"role": "system", "content": "Be a good bot!"}
            ]
        else:
            assert request_body["messages"] == [
                {"role": "user", "content": "Be a good bot!"}
            ]


@pytest.mark.parametrize(
    "model, expected_tool_calling_support",
    [("o1-preview", False), ("o1-mini", False), ("o1", True)],
)
@pytest.mark.asyncio
async def test_o1_handle_tool_calling_optional_params(
    model, expected_tool_calling_support
):
    """
    Tests that:
    - max_tokens is translated to 'max_completion_tokens'
    - role 'system' is translated to 'user'
    """
    from openai import AsyncOpenAI
    from litellm.utils import ProviderConfigManager
    from litellm.types.utils import LlmProviders

    os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
    litellm.model_cost = litellm.get_model_cost_map(url="")

    config = ProviderConfigManager.get_provider_chat_config(
        model=model, provider=LlmProviders.OPENAI
    )

    supported_params = config.get_supported_openai_params(model=model)

    assert expected_tool_calling_support == ("tools" in supported_params)


@pytest.mark.asyncio
@pytest.mark.parametrize("model", ["gpt-4", "gpt-4-0314", "gpt-4-32k", "o1-preview"])
async def test_o1_max_completion_tokens(model: str):
    """
    Tests that:
    - max_completion_tokens is passed directly to OpenAI chat completion models
    """
    from openai import AsyncOpenAI

    litellm.set_verbose = True

    client = AsyncOpenAI(api_key="fake-api-key")

    with patch.object(
        client.chat.completions.with_raw_response, "create"
    ) as mock_client:
        try:
            await litellm.acompletion(
                model=model,
                max_completion_tokens=10,
                messages=[{"role": "user", "content": "Hello!"}],
                client=client,
            )
        except Exception as e:
            print(f"Error: {e}")

        mock_client.assert_called_once()
        request_body = mock_client.call_args.kwargs

        print("request_body: ", request_body)

        assert request_body["model"] == model
        assert request_body["max_completion_tokens"] == 10
        assert request_body["messages"] == [{"role": "user", "content": "Hello!"}]


def test_litellm_responses():
    """
    ensures that type of completion_tokens_details is correctly handled / returned
    """
    from litellm import ModelResponse
    from litellm.types.utils import CompletionTokensDetails

    response = ModelResponse(
        usage={
            "completion_tokens": 436,
            "prompt_tokens": 14,
            "total_tokens": 450,
            "completion_tokens_details": {"reasoning_tokens": 0},
        }
    )

    print("response: ", response)

    assert isinstance(response.usage.completion_tokens_details, CompletionTokensDetails)


class TestOpenAIO1(BaseOSeriesModelsTest, BaseLLMChatTest):
    def get_base_completion_call_args(self):
        return {
            "model": "o1",
        }

    def get_client(self):
        from openai import OpenAI

        return OpenAI(api_key="fake-api-key")

    def test_tool_call_no_arguments(self, tool_call_no_arguments):
        """Test that tool calls with no arguments is translated correctly. Relevant issue: https://github.com/BerriAI/litellm/issues/6833"""
        pass

    def test_prompt_caching(self):
        """Temporary override. o1 prompt caching is not working."""
        pass


class TestOpenAIO3(BaseOSeriesModelsTest, BaseLLMChatTest):
    def get_base_completion_call_args(self):
        return {
            "model": "o3-mini",
        }

    def get_client(self):
        from openai import OpenAI

        return OpenAI(api_key="fake-api-key")

    def test_tool_call_no_arguments(self, tool_call_no_arguments):
        """Test that tool calls with no arguments is translated correctly. Relevant issue: https://github.com/BerriAI/litellm/issues/6833"""
        pass

    def test_prompt_caching(self):
        """Override, as o3 prompt caching is flaky"""
        pass


def test_o1_supports_vision():
    """Test that o1 supports vision"""
    os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
    litellm.model_cost = litellm.get_model_cost_map(url="")
    for k, v in litellm.model_cost.items():
        if k.startswith("o1") and v.get("litellm_provider") == "openai":
            assert v.get("supports_vision") is True, f"{k} does not support vision"


def test_o3_reasoning_effort():
    resp = litellm.completion(
        model="o3-mini",
        messages=[{"role": "user", "content": "Hello!"}],
        reasoning_effort="high",
    )
    assert resp.choices[0].message.content is not None


@pytest.mark.parametrize("model", ["o1-preview", "o1-mini", "o1", "o3-mini"])
def test_streaming_response(model):
    """Test that streaming response is returned correctly"""
    from litellm import completion

    response = completion(
        model=model,
        messages=[
            {"role": "system", "content": "Be a good bot!"},
            {"role": "user", "content": "Hello!"},
        ],
        stream=True,
    )

    assert response is not None

    chunks = []
    for chunk in response:
        chunks.append(chunk)

    resp = litellm.stream_chunk_builder(chunks=chunks)
    print(resp)