File size: 6,833 Bytes
a0ebf39
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import { describe, expect, it } from 'vitest';

import { getProvider } from '@/lib/ai/providers';
import {
  getDefaultThinkingConfig,
  getThinkingDisplayValue,
  normalizeThinkingConfig,
  supportsConfigurableThinking,
} from '@/lib/ai/thinking-config';
import type { ProviderId } from '@/lib/types/provider';

function getThinking(providerId: ProviderId, modelId: string) {
  const model = getProvider(providerId)?.models.find((item) => item.id === modelId);
  return model?.capabilities?.thinking;
}

describe('thinking config metadata', () => {
  it('marks configurable models with adapter-backed thinking capabilities', () => {
    const thinking = getThinking('qwen', 'qwen3.6-plus');

    expect(supportsConfigurableThinking(thinking)).toBe(true);
    expect(thinking?.control).toBe('toggle-budget');
    expect(thinking?.requestAdapter).toBe('qwen');
  });

  it('does not expose fixed thinking models as configurable', () => {
    const thinking = getThinking('grok', 'grok-4.20-reasoning');
    const minimaxThinking = getThinking('minimax', 'MiniMax-M2.7');

    expect(thinking?.control).toBe('none');
    expect(supportsConfigurableThinking(thinking)).toBe(false);
    expect(minimaxThinking?.control).toBe('none');
    expect(supportsConfigurableThinking(minimaxThinking)).toBe(false);
  });

  it('exposes Claude Haiku 4.5 thinking as budget-only, not effort', () => {
    const thinking = getThinking('anthropic', 'claude-haiku-4-5');

    expect(supportsConfigurableThinking(thinking)).toBe(true);
    expect(thinking?.control).toBe('toggle-budget');
    expect(thinking?.requestAdapter).toBe('anthropic');
    expect(thinking?.effortValues).toBeUndefined();
    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'disabled',
      budgetTokens: 1024,
    });
    expect(normalizeThinkingConfig(thinking, { mode: 'enabled', budgetTokens: 4096 })).toEqual({
      mode: 'enabled',
      budgetTokens: 4096,
    });
  });

  it('removes deprecated and legacy models from the built-in catalog', () => {
    const openaiModels = getProvider('openai')?.models.map((item) => item.id);
    const glmModels = getProvider('glm')?.models.map((item) => item.id);
    const googleModels = getProvider('google')?.models.map((item) => item.id);
    const deepseekModels = getProvider('deepseek')?.models.map((item) => item.id);
    const hunyuanModels = getProvider('tencent-hunyuan')?.models.map((item) => item.id);
    const minimaxModels = getProvider('minimax')?.models.map((item) => item.id);
    const siliconflowModels = getProvider('siliconflow')?.models.map((item) => item.id);

    expect(openaiModels).not.toContain('o3-mini');
    expect(openaiModels).not.toContain('o3');
    expect(openaiModels).not.toContain('o4-mini');
    expect(openaiModels).not.toContain('gpt-5.2');
    expect(openaiModels).not.toContain('gpt-5.1');
    expect(openaiModels).not.toContain('gpt-5');
    expect(openaiModels).not.toContain('gpt-4o');
    expect(glmModels).not.toContain('glm-4.5-air');
    expect(glmModels).not.toContain('glm-4.5-airx');
    expect(glmModels).not.toContain('glm-4.5-flash');
    expect(googleModels).toContain('gemini-3.1-pro-preview');
    expect(googleModels).not.toContain('gemini-3-pro-preview');
    expect(deepseekModels).toEqual(['deepseek-v4-pro', 'deepseek-v4-flash']);
    expect(hunyuanModels).toEqual(['hy3-preview']);
    expect(minimaxModels).toEqual(['MiniMax-M2.7']);
    expect(siliconflowModels).not.toContain('MiniMaxAI/MiniMax-M2');
  });
});

describe('thinking config normalization', () => {
  it('normalizes OpenAI effort defaults and selected effort values', () => {
    const thinking = getThinking('openai', 'gpt-5.4');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'disabled',
      effort: 'none',
    });
    expect(normalizeThinkingConfig(thinking, { effort: 'high' })).toEqual({
      mode: 'enabled',
      effort: 'high',
    });
  });

  it('normalizes GPT-5.5 as non-toggleable effort levels', () => {
    const thinking = getThinking('openai', 'gpt-5.5');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'enabled',
      effort: 'medium',
    });
    expect(normalizeThinkingConfig(thinking, { mode: 'disabled' })).toEqual({
      mode: 'enabled',
      effort: 'low',
    });
    expect(thinking?.effortValues).toEqual(['low', 'medium', 'high', 'xhigh']);
  });

  it('normalizes Claude 4.5+ thinking as effort levels', () => {
    const thinking = getThinking('anthropic', 'claude-sonnet-4-6');
    const opus47Thinking = getThinking('anthropic', 'claude-opus-4-7');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'enabled',
      effort: 'medium',
    });
    expect(normalizeThinkingConfig(thinking, { effort: 'max' })).toEqual({
      mode: 'enabled',
      effort: 'max',
    });
    expect(normalizeThinkingConfig(thinking, { mode: 'disabled' })).toEqual({
      mode: 'disabled',
      effort: 'none',
    });
    expect(opus47Thinking?.effortValues).toEqual(['none', 'low', 'medium', 'high', 'xhigh', 'max']);
  });

  it('normalizes DeepSeek V4 thinking as high/max effort levels', () => {
    const thinking = getThinking('deepseek', 'deepseek-v4-pro');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'enabled',
      effort: 'high',
    });
    expect(normalizeThinkingConfig(thinking, { effort: 'max' })).toEqual({
      mode: 'enabled',
      effort: 'max',
    });
  });

  it('normalizes Tencent HY3 thinking as no_think/low/high effort levels', () => {
    const thinking = getThinking('tencent-hunyuan', 'hy3-preview');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'disabled',
      effort: 'none',
    });
    expect(normalizeThinkingConfig(thinking, { effort: 'high' })).toEqual({
      mode: 'enabled',
      effort: 'high',
    });
    expect(thinking?.effortValues).toEqual(['none', 'low', 'high']);
  });

  it('normalizes Doubao Seed 2.0 thinking as reasoning effort levels', () => {
    const thinking = getThinking('doubao', 'doubao-seed-2-0-pro-260215');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'enabled',
      effort: 'medium',
    });
    expect(normalizeThinkingConfig(thinking, { effort: 'high' })).toEqual({
      mode: 'enabled',
      effort: 'high',
    });
    expect(thinking?.effortValues).toEqual(['minimal', 'low', 'medium', 'high']);
  });

  it('preserves dynamic Gemini budgets and display labels', () => {
    const thinking = getThinking('google', 'gemini-2.5-flash');

    expect(getDefaultThinkingConfig(thinking)).toEqual({
      mode: 'enabled',
      budgetTokens: -1,
    });
    expect(getThinkingDisplayValue(thinking, undefined)).toBe('auto');
    expect(getThinkingDisplayValue(thinking, { mode: 'enabled', budgetTokens: 8192 })).toBe('8192');
  });
});