@lobehub/chat 1.40.1 → 1.40.3

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (118) hide show
  1. package/CHANGELOG.md +50 -0
  2. package/changelog/v1.json +18 -0
  3. package/package.json +1 -1
  4. package/src/app/(main)/changelog/page.tsx +3 -1
  5. package/src/app/(main)/chat/(workspace)/features/ChangelogModal.tsx +11 -0
  6. package/src/app/(main)/chat/(workspace)/page.tsx +8 -5
  7. package/src/app/(main)/discover/(detail)/model/[...slugs]/features/Actions.tsx +1 -1
  8. package/src/app/(main)/discover/(detail)/model/[...slugs]/features/Header.tsx +1 -1
  9. package/src/app/(main)/discover/(detail)/model/[...slugs]/features/InfoSidebar/SuggestionItem.tsx +2 -2
  10. package/src/app/(main)/discover/(detail)/model/[...slugs]/features/ProviderList/ProviderItem.tsx +1 -1
  11. package/src/app/(main)/discover/(detail)/provider/[slug]/features/ModelList/ModelItem.tsx +3 -3
  12. package/src/app/(main)/discover/(list)/models/features/Card.tsx +6 -2
  13. package/src/app/(main)/settings/llm/ProviderList/Cloudflare/index.tsx +1 -5
  14. package/src/app/(main)/settings/llm/components/ProviderModelList/ModelFetcher.tsx +2 -2
  15. package/src/app/@modal/(.)changelog/modal/page.tsx +3 -1
  16. package/src/components/Loading/BrandTextLoading/index.tsx +5 -0
  17. package/src/components/ModelSelect/index.tsx +7 -4
  18. package/src/config/__tests__/app.test.ts +6 -2
  19. package/src/config/app.ts +1 -2
  20. package/src/config/featureFlags/schema.ts +3 -0
  21. package/src/config/modelProviders/ai21.ts +2 -2
  22. package/src/config/modelProviders/ai360.ts +4 -4
  23. package/src/config/modelProviders/anthropic.ts +8 -8
  24. package/src/config/modelProviders/azure.ts +5 -5
  25. package/src/config/modelProviders/baichuan.ts +6 -6
  26. package/src/config/modelProviders/bedrock.ts +14 -14
  27. package/src/config/modelProviders/cloudflare.ts +12 -11
  28. package/src/config/modelProviders/deepseek.ts +1 -1
  29. package/src/config/modelProviders/fireworksai.ts +29 -27
  30. package/src/config/modelProviders/giteeai.ts +7 -7
  31. package/src/config/modelProviders/github.ts +29 -28
  32. package/src/config/modelProviders/google.ts +18 -19
  33. package/src/config/modelProviders/groq.ts +13 -13
  34. package/src/config/modelProviders/higress.ts +195 -194
  35. package/src/config/modelProviders/huggingface.ts +7 -7
  36. package/src/config/modelProviders/hunyuan.ts +25 -17
  37. package/src/config/modelProviders/internlm.ts +6 -4
  38. package/src/config/modelProviders/minimax.ts +5 -5
  39. package/src/config/modelProviders/mistral.ts +14 -16
  40. package/src/config/modelProviders/moonshot.ts +3 -3
  41. package/src/config/modelProviders/novita.ts +15 -15
  42. package/src/config/modelProviders/ollama.ts +46 -46
  43. package/src/config/modelProviders/openai.ts +23 -22
  44. package/src/config/modelProviders/openrouter.ts +20 -18
  45. package/src/config/modelProviders/perplexity.ts +7 -7
  46. package/src/config/modelProviders/qwen.ts +23 -25
  47. package/src/config/modelProviders/sensenova.ts +8 -8
  48. package/src/config/modelProviders/siliconcloud.ts +138 -92
  49. package/src/config/modelProviders/spark.ts +6 -6
  50. package/src/config/modelProviders/stepfun.ts +9 -9
  51. package/src/config/modelProviders/taichu.ts +2 -3
  52. package/src/config/modelProviders/togetherai.ts +57 -48
  53. package/src/config/modelProviders/upstage.ts +3 -3
  54. package/src/config/modelProviders/wenxin.ts +12 -12
  55. package/src/config/modelProviders/xai.ts +4 -4
  56. package/src/config/modelProviders/zeroone.ts +11 -11
  57. package/src/config/modelProviders/zhipu.ts +17 -16
  58. package/src/database/_deprecated/core/model.ts +1 -1
  59. package/src/database/_deprecated/models/sessionGroup.ts +4 -1
  60. package/src/database/client/migrations.json +2 -5
  61. package/src/database/migrations/meta/0012_snapshot.json +176 -518
  62. package/src/database/schemas/agent.ts +1 -1
  63. package/src/database/schemas/message.ts +1 -0
  64. package/src/database/schemas/session.ts +1 -0
  65. package/src/database/server/models/topic.ts +19 -17
  66. package/src/features/ChangelogModal/index.tsx +8 -2
  67. package/src/features/DebugUI/Content.tsx +0 -1
  68. package/src/features/PluginStore/index.tsx +2 -2
  69. package/src/features/User/UserPanel/useMenu.tsx +1 -1
  70. package/src/layout/GlobalProvider/AntdV5MonkeyPatch.tsx +4 -4
  71. package/src/libs/agent-runtime/google/index.ts +4 -3
  72. package/src/libs/agent-runtime/higress/index.ts +1 -1
  73. package/src/libs/agent-runtime/huggingface/index.ts +2 -4
  74. package/src/libs/agent-runtime/minimax/index.ts +5 -10
  75. package/src/libs/agent-runtime/mistral/index.ts +3 -6
  76. package/src/libs/agent-runtime/moonshot/index.ts +3 -6
  77. package/src/libs/agent-runtime/novita/__snapshots__/index.test.ts.snap +18 -18
  78. package/src/libs/agent-runtime/novita/index.ts +1 -1
  79. package/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap +10 -10
  80. package/src/libs/agent-runtime/openrouter/__snapshots__/index.test.ts.snap +168 -168
  81. package/src/libs/agent-runtime/openrouter/index.ts +1 -1
  82. package/src/libs/agent-runtime/perplexity/index.ts +4 -4
  83. package/src/libs/agent-runtime/sensenova/index.ts +9 -3
  84. package/src/libs/agent-runtime/taichu/index.ts +4 -10
  85. package/src/libs/agent-runtime/utils/streams/minimax.test.ts +5 -2
  86. package/src/libs/agent-runtime/utils/streams/minimax.ts +4 -1
  87. package/src/libs/agent-runtime/zhipu/index.ts +12 -13
  88. package/src/libs/langchain/loaders/index.ts +2 -2
  89. package/src/libs/langchain/types.ts +9 -1
  90. package/src/locales/default/modelProvider.ts +1 -1
  91. package/src/migrations/FromV3ToV4/fixtures/ollama-output-v4.json +1 -1
  92. package/src/migrations/FromV6ToV7/types/v7.ts +0 -2
  93. package/src/server/globalConfig/genServerLLMConfig.test.ts +4 -4
  94. package/src/server/globalConfig/genServerLLMConfig.ts +29 -24
  95. package/src/server/globalConfig/index.ts +1 -2
  96. package/src/server/routers/edge/config/__snapshots__/index.test.ts.snap +9 -9
  97. package/src/server/routers/lambda/_template.ts +1 -1
  98. package/src/server/routers/lambda/knowledgeBase.ts +1 -1
  99. package/src/server/routers/lambda/session.ts +1 -1
  100. package/src/server/routers/lambda/sessionGroup.ts +1 -1
  101. package/src/server/routers/lambda/thread.ts +1 -1
  102. package/src/server/services/changelog/index.test.ts +4 -2
  103. package/src/server/services/changelog/index.ts +10 -2
  104. package/src/server/services/nextAuthUser/index.ts +1 -1
  105. package/src/store/serverConfig/selectors.test.ts +1 -0
  106. package/src/store/user/slices/modelList/__snapshots__/action.test.ts.snap +1 -1
  107. package/src/store/user/slices/modelList/action.test.ts +4 -4
  108. package/src/store/user/slices/modelList/reducers/customModelCard.test.ts +6 -6
  109. package/src/store/user/slices/modelList/selectors/modelProvider.ts +3 -2
  110. package/src/tools/dalle/Render/Item/index.tsx +1 -1
  111. package/src/types/files/index.ts +0 -1
  112. package/src/types/llm.ts +4 -5
  113. package/src/utils/__snapshots__/parseModels.test.ts.snap +2 -2
  114. package/src/utils/genUserLLMConfig.test.ts +4 -4
  115. package/src/utils/genUserLLMConfig.ts +6 -4
  116. package/src/utils/parseModels.test.ts +16 -16
  117. package/src/utils/parseModels.ts +1 -1
  118. package/src/utils/server/jwt.ts +2 -6
@@ -3,45 +3,45 @@ import { ModelProviderCard } from '@/types/llm';
3
3
  const HuggingFace: ModelProviderCard = {
4
4
  chatModels: [
5
5
  {
6
+ contextWindowTokens: 32_768,
6
7
  description: 'Mistral AI的指令调优模型',
7
8
  displayName: 'Mistral 7B Instruct v0.3',
8
9
  enabled: true,
9
10
  id: 'mistralai/Mistral-7B-Instruct-v0.3',
10
- tokens: 32_768,
11
11
  },
12
12
  {
13
+ contextWindowTokens: 8192,
13
14
  description: 'Google的轻量级指令调优模型',
14
15
  displayName: 'Gemma 2 2B Instruct',
15
16
  id: 'google/gemma-2-2b-it',
16
- tokens: 8192,
17
17
  },
18
18
  {
19
+ contextWindowTokens: 32_768,
19
20
  description: '阿里云通义千问团队开发的大型语言模型',
20
21
  displayName: 'Qwen 2.5 72B Instruct',
21
22
  id: 'Qwen/Qwen2.5-72B-Instruct',
22
- tokens: 32_768,
23
23
  },
24
24
  {
25
+ contextWindowTokens: 32_768,
25
26
  description: 'Qwen2.5-Coder 专注于代码编写',
26
27
  displayName: 'Qwen 2.5 Coder 32B Instruct',
27
28
  id: 'Qwen/Qwen2.5-Coder-32B-Instruct',
28
- tokens: 32_768,
29
29
  },
30
30
  {
31
+ contextWindowTokens: 32_768,
31
32
  description: 'Qwen QwQ 是由 Qwen 团队开发的实验研究模型,专注于提升AI推理能力。',
32
33
  displayName: 'QwQ 32B Preview',
33
34
  id: 'Qwen/QwQ-32B-Preview',
34
- tokens: 32_768,
35
35
  },
36
36
  {
37
+ contextWindowTokens: 32_768,
37
38
  displayName: 'Phi 3.5 mini instruct',
38
39
  id: 'microsoft/Phi-3.5-mini-instruct',
39
- tokens: 32_768,
40
40
  },
41
41
  {
42
+ contextWindowTokens: 16_384,
42
43
  displayName: 'Hermes 3 Llama 3.1 8B',
43
44
  id: 'NousResearch/Hermes-3-Llama-3.1-8B',
44
- tokens: 16_384,
45
45
  },
46
46
  ],
47
47
  checkModel: 'mistralai/Mistral-7B-Instruct-v0.2',
@@ -4,7 +4,9 @@ import { ModelProviderCard } from '@/types/llm';
4
4
  const Hunyuan: ModelProviderCard = {
5
5
  chatModels: [
6
6
  {
7
- description: '升级为 MOE 结构,上下文窗口为 256k ,在 NLP,代码,数学,行业等多项评测集上领先众多开源模型。',
7
+ contextWindowTokens: 256_000,
8
+ description:
9
+ '升级为 MOE 结构,上下文窗口为 256k ,在 NLP,代码,数学,行业等多项评测集上领先众多开源模型。',
8
10
  displayName: 'Hunyuan Lite',
9
11
  enabled: true,
10
12
  id: 'hunyuan-lite',
@@ -14,10 +16,11 @@ const Hunyuan: ModelProviderCard = {
14
16
  input: 0,
15
17
  output: 0,
16
18
  },
17
- tokens: 256_000,
18
19
  },
19
20
  {
20
- description: '采用更优的路由策略,同时缓解了负载均衡和专家趋同的问题。长文方面,大海捞针指标达到99.9%。MOE-32K 性价比相对更高,在平衡效果、价格的同时,可对实现对长文本输入的处理。',
21
+ contextWindowTokens: 32_000,
22
+ description:
23
+ '采用更优的路由策略,同时缓解了负载均衡和专家趋同的问题。长文方面,大海捞针指标达到99.9%。MOE-32K 性价比相对更高,在平衡效果、价格的同时,可对实现对长文本输入的处理。',
21
24
  displayName: 'Hunyuan Standard',
22
25
  enabled: true,
23
26
  id: 'hunyuan-standard',
@@ -27,10 +30,11 @@ const Hunyuan: ModelProviderCard = {
27
30
  input: 4.5,
28
31
  output: 5,
29
32
  },
30
- tokens: 32_000,
31
33
  },
32
34
  {
33
- description: '采用更优的路由策略,同时缓解了负载均衡和专家趋同的问题。长文方面,大海捞针指标达到99.9%。MOE-256K 在长度和效果上进一步突破,极大的扩展了可输入长度。',
35
+ contextWindowTokens: 256_000,
36
+ description:
37
+ '采用更优的路由策略,同时缓解了负载均衡和专家趋同的问题。长文方面,大海捞针指标达到99.9%。MOE-256K 在长度和效果上进一步突破,极大的扩展了可输入长度。',
34
38
  displayName: 'Hunyuan Standard 256K',
35
39
  enabled: true,
36
40
  id: 'hunyuan-standard-256K',
@@ -40,10 +44,11 @@ const Hunyuan: ModelProviderCard = {
40
44
  input: 15,
41
45
  output: 60,
42
46
  },
43
- tokens: 256_000,
44
47
  },
45
48
  {
46
- description: '混元全新一代大语言模型的预览版,采用全新的混合专家模型(MoE)结构,相比hunyuan-pro推理效率更快,效果表现更强。',
49
+ contextWindowTokens: 32_000,
50
+ description:
51
+ '混元全新一代大语言模型的预览版,采用全新的混合专家模型(MoE)结构,相比hunyuan-pro推理效率更快,效果表现更强。',
47
52
  displayName: 'Hunyuan Turbo',
48
53
  enabled: true,
49
54
  functionCall: true,
@@ -54,10 +59,11 @@ const Hunyuan: ModelProviderCard = {
54
59
  input: 15,
55
60
  output: 50,
56
61
  },
57
- tokens: 32_000,
58
62
  },
59
63
  {
60
- description: '万亿级参数规模 MOE-32K 长文模型。在各种 benchmark 上达到绝对领先的水平,复杂指令和推理,具备复杂数学能力,支持 functioncall,在多语言翻译、金融法律医疗等领域应用重点优化。',
64
+ contextWindowTokens: 32_000,
65
+ description:
66
+ '万亿级参数规模 MOE-32K 长文模型。在各种 benchmark 上达到绝对领先的水平,复杂指令和推理,具备复杂数学能力,支持 functioncall,在多语言翻译、金融法律医疗等领域应用重点优化。',
61
67
  displayName: 'Hunyuan Pro',
62
68
  enabled: true,
63
69
  functionCall: true,
@@ -68,9 +74,9 @@ const Hunyuan: ModelProviderCard = {
68
74
  input: 30,
69
75
  output: 100,
70
76
  },
71
- tokens: 32_000,
72
77
  },
73
78
  {
79
+ contextWindowTokens: 8000,
74
80
  description: '混元最新多模态模型,支持图片+文本输入生成文本内容。',
75
81
  displayName: 'Hunyuan Vision',
76
82
  enabled: true,
@@ -81,11 +87,12 @@ const Hunyuan: ModelProviderCard = {
81
87
  input: 18,
82
88
  output: 18,
83
89
  },
84
- tokens: 8000,
85
90
  vision: true,
86
91
  },
87
92
  {
88
- description: '混元最新代码生成模型,经过 200B 高质量代码数据增训基座模型,迭代半年高质量 SFT 数据训练,上下文长窗口长度增大到 8K,五大语言代码生成自动评测指标上位居前列;五大语言10项考量各方面综合代码任务人工高质量评测上,性能处于第一梯队',
93
+ contextWindowTokens: 8000,
94
+ description:
95
+ '混元最新代码生成模型,经过 200B 高质量代码数据增训基座模型,迭代半年高质量 SFT 数据训练,上下文长窗口长度增大到 8K,五大语言代码生成自动评测指标上位居前列;五大语言10项考量各方面综合代码任务人工高质量评测上,性能处于第一梯队',
89
96
  displayName: 'Hunyuan Code',
90
97
  id: 'hunyuan-code',
91
98
  maxOutput: 4000,
@@ -94,10 +101,11 @@ const Hunyuan: ModelProviderCard = {
94
101
  input: 4,
95
102
  output: 8,
96
103
  },
97
- tokens: 8000,
98
104
  },
99
105
  {
100
- description: '混元最新 MOE 架构 FunctionCall 模型,经过高质量的 FunctionCall 数据训练,上下文窗口达 32K,在多个维度的评测指标上处于领先。',
106
+ contextWindowTokens: 32_000,
107
+ description:
108
+ '混元最新 MOE 架构 FunctionCall 模型,经过高质量的 FunctionCall 数据训练,上下文窗口达 32K,在多个维度的评测指标上处于领先。',
101
109
  displayName: 'Hunyuan FunctionCall',
102
110
  functionCall: true,
103
111
  id: 'hunyuan-functioncall',
@@ -107,10 +115,11 @@ const Hunyuan: ModelProviderCard = {
107
115
  input: 4,
108
116
  output: 8,
109
117
  },
110
- tokens: 32_000,
111
118
  },
112
119
  {
113
- description: '混元最新版角色扮演模型,混元官方精调训练推出的角色扮演模型,基于混元模型结合角色扮演场景数据集进行增训,在角色扮演场景具有更好的基础效果。',
120
+ contextWindowTokens: 8000,
121
+ description:
122
+ '混元最新版角色扮演模型,混元官方精调训练推出的角色扮演模型,基于混元模型结合角色扮演场景数据集进行增训,在角色扮演场景具有更好的基础效果。',
114
123
  displayName: 'Hunyuan Role',
115
124
  id: 'hunyuan-role',
116
125
  maxOutput: 4000,
@@ -119,7 +128,6 @@ const Hunyuan: ModelProviderCard = {
119
128
  input: 4,
120
129
  output: 8,
121
130
  },
122
- tokens: 8000,
123
131
  },
124
132
  ],
125
133
  checkModel: 'hunyuan-lite',
@@ -3,7 +3,9 @@ import { ModelProviderCard } from '@/types/llm';
3
3
  const InternLM: ModelProviderCard = {
4
4
  chatModels: [
5
5
  {
6
- description: '我们最新的模型系列,有着卓越的推理性能,支持 1M 的上下文长度以及更强的指令跟随和工具调用能力。',
6
+ contextWindowTokens: 32_768,
7
+ description:
8
+ '我们最新的模型系列,有着卓越的推理性能,支持 1M 的上下文长度以及更强的指令跟随和工具调用能力。',
7
9
  displayName: 'InternLM2.5',
8
10
  enabled: true,
9
11
  functionCall: true,
@@ -13,9 +15,9 @@ const InternLM: ModelProviderCard = {
13
15
  input: 0,
14
16
  output: 0,
15
17
  },
16
- tokens: 32_768,
17
18
  },
18
19
  {
20
+ contextWindowTokens: 32_768,
19
21
  description: '我们仍在维护的老版本模型,有 7B、20B 多种模型参数量可选。',
20
22
  displayName: 'InternLM2 Pro Chat',
21
23
  functionCall: true,
@@ -25,7 +27,6 @@ const InternLM: ModelProviderCard = {
25
27
  input: 0,
26
28
  output: 0,
27
29
  },
28
- tokens: 32_768,
29
30
  },
30
31
  ],
31
32
  checkModel: 'internlm2.5-latest',
@@ -34,7 +35,8 @@ const InternLM: ModelProviderCard = {
34
35
  disableBrowserRequest: true,
35
36
  id: 'internlm',
36
37
  modelList: { showModelFetcher: true },
37
- modelsUrl: 'https://internlm.intern-ai.org.cn/doc/docs/Models#%E8%8E%B7%E5%8F%96%E6%A8%A1%E5%9E%8B%E5%88%97%E8%A1%A8',
38
+ modelsUrl:
39
+ 'https://internlm.intern-ai.org.cn/doc/docs/Models#%E8%8E%B7%E5%8F%96%E6%A8%A1%E5%9E%8B%E5%88%97%E8%A1%A8',
38
40
  name: 'InternLM',
39
41
  url: 'https://internlm.intern-ai.org.cn',
40
42
  };
@@ -4,40 +4,40 @@ import { ModelProviderCard } from '@/types/llm';
4
4
  const Minimax: ModelProviderCard = {
5
5
  chatModels: [
6
6
  {
7
+ contextWindowTokens: 245_760,
7
8
  description: '适用于广泛的自然语言处理任务,包括文本生成、对话系统等。',
8
9
  displayName: 'abab6.5s',
9
10
  enabled: true,
10
11
  functionCall: true,
11
12
  id: 'abab6.5s-chat',
12
- tokens: 245_760,
13
13
  },
14
14
  {
15
+ contextWindowTokens: 8192,
15
16
  description: '专为多语种人设对话设计,支持英文及其他多种语言的高质量对话生成。',
16
17
  displayName: 'abab6.5g',
17
18
  enabled: true,
18
19
  functionCall: true,
19
20
  id: 'abab6.5g-chat',
20
- tokens: 8192,
21
21
  },
22
22
  {
23
+ contextWindowTokens: 8192,
23
24
  description: '针对中文人设对话场景优化,提供流畅且符合中文表达习惯的对话生成能力。',
24
25
  displayName: 'abab6.5t',
25
26
  enabled: true,
26
27
  functionCall: true,
27
28
  id: 'abab6.5t-chat',
28
- tokens: 8192,
29
29
  },
30
30
  {
31
+ contextWindowTokens: 16_384,
31
32
  description: '面向生产力场景,支持复杂任务处理和高效文本生成,适用于专业领域应用。',
32
33
  displayName: 'abab5.5',
33
34
  id: 'abab5.5-chat',
34
- tokens: 16_384,
35
35
  },
36
36
  {
37
+ contextWindowTokens: 8192,
37
38
  description: '专为中文人设对话场景设计,提供高质量的中文对话生成能力,适用于多种应用场景。',
38
39
  displayName: 'abab5.5s',
39
40
  id: 'abab5.5s-chat',
40
- tokens: 8192,
41
41
  },
42
42
  ],
43
43
  checkModel: 'abab6.5s-chat',
@@ -5,6 +5,7 @@ import { ModelProviderCard } from '@/types/llm';
5
5
  const Mistral: ModelProviderCard = {
6
6
  chatModels: [
7
7
  {
8
+ contextWindowTokens: 128_000,
8
9
  description:
9
10
  'Mistral Nemo是一个与Nvidia合作开发的12B模型,提供出色的推理和编码性能,易于集成和替换。',
10
11
  displayName: 'Mistral Nemo',
@@ -15,9 +16,9 @@ const Mistral: ModelProviderCard = {
15
16
  input: 0.15,
16
17
  output: 0.15,
17
18
  },
18
- tokens: 128_000,
19
19
  },
20
20
  {
21
+ contextWindowTokens: 128_000,
21
22
  description:
22
23
  'Mistral Small是成本效益高、快速且可靠的选项,适用于翻译、摘要和情感分析等用例。',
23
24
  displayName: 'Mistral Small',
@@ -28,9 +29,9 @@ const Mistral: ModelProviderCard = {
28
29
  input: 0.2,
29
30
  output: 0.6,
30
31
  },
31
- tokens: 128_000,
32
32
  },
33
33
  {
34
+ contextWindowTokens: 128_000,
34
35
  description:
35
36
  'Mistral Large是旗舰大模型,擅长多语言任务、复杂推理和代码生成,是高端应用的理想选择。',
36
37
  displayName: 'Mistral Large',
@@ -41,9 +42,9 @@ const Mistral: ModelProviderCard = {
41
42
  input: 2,
42
43
  output: 6,
43
44
  },
44
- tokens: 128_000,
45
45
  },
46
46
  {
47
+ contextWindowTokens: 32_768,
47
48
  description: 'Codestral是专注于代码生成的尖端生成模型,优化了中间填充和代码补全任务。',
48
49
  displayName: 'Codestral',
49
50
  id: 'codestral-latest',
@@ -51,9 +52,9 @@ const Mistral: ModelProviderCard = {
51
52
  input: 0.2,
52
53
  output: 0.6,
53
54
  },
54
- tokens: 32_768,
55
55
  },
56
56
  {
57
+ contextWindowTokens: 128_000,
57
58
  description:
58
59
  'Pixtral Large 是一款拥有 1240 亿参数的开源多模态模型,基于 Mistral Large 2 构建。这是我们多模态家族中的第二款模型,展现了前沿水平的图像理解能力。',
59
60
  displayName: 'Pixtral Large',
@@ -64,10 +65,10 @@ const Mistral: ModelProviderCard = {
64
65
  input: 2,
65
66
  output: 6,
66
67
  },
67
- tokens: 128_000,
68
68
  vision: true,
69
69
  },
70
70
  {
71
+ contextWindowTokens: 128_000,
71
72
  description:
72
73
  'Pixtral 模型在图表和图理解、文档问答、多模态推理和指令遵循等任务上表现出强大的能力,能够以自然分辨率和宽高比摄入图像,还能够在长达 128K 令牌的长上下文窗口中处理任意数量的图像。',
73
74
  displayName: 'Pixtral 12B',
@@ -77,32 +78,30 @@ const Mistral: ModelProviderCard = {
77
78
  input: 0.15,
78
79
  output: 0.15,
79
80
  },
80
- tokens: 128_000,
81
81
  vision: true,
82
82
  },
83
83
  {
84
- description:
85
- 'Ministral 3B 是Mistral的世界顶级边缘模型。',
84
+ contextWindowTokens: 128_000,
85
+ description: 'Ministral 3B 是Mistral的世界顶级边缘模型。',
86
86
  displayName: 'Ministral 3B',
87
87
  id: 'ministral-3b-latest',
88
88
  pricing: {
89
89
  input: 0.04,
90
90
  output: 0.04,
91
91
  },
92
- tokens: 128_000,
93
92
  },
94
93
  {
95
- description:
96
- 'Ministral 8B 是Mistral的性价比极高的边缘模型。',
94
+ contextWindowTokens: 128_000,
95
+ description: 'Ministral 8B 是Mistral的性价比极高的边缘模型。',
97
96
  displayName: 'Ministral 8B',
98
97
  id: 'ministral-8b-latest',
99
98
  pricing: {
100
99
  input: 0.1,
101
100
  output: 0.1,
102
101
  },
103
- tokens: 128_000,
104
102
  },
105
103
  {
104
+ contextWindowTokens: 32_768,
106
105
  description:
107
106
  'Mistral 7B是一款紧凑但高性能的模型,擅长批量处理和简单任务,如分类和文本生成,具有良好的推理能力。',
108
107
  displayName: 'Mistral 7B',
@@ -111,9 +110,9 @@ const Mistral: ModelProviderCard = {
111
110
  input: 0.25,
112
111
  output: 0.25,
113
112
  },
114
- tokens: 32_768,
115
113
  },
116
114
  {
115
+ contextWindowTokens: 32_768,
117
116
  description:
118
117
  'Mixtral 8x7B是一个稀疏专家模型,利用多个参数提高推理速度,适合处理多语言和代码生成任务。',
119
118
  displayName: 'Mixtral 8x7B',
@@ -122,9 +121,9 @@ const Mistral: ModelProviderCard = {
122
121
  input: 0.7,
123
122
  output: 0.7,
124
123
  },
125
- tokens: 32_768,
126
124
  },
127
125
  {
126
+ contextWindowTokens: 65_536,
128
127
  description:
129
128
  'Mixtral 8x22B是一个更大的专家模型,专注于复杂任务,提供出色的推理能力和更高的吞吐量。',
130
129
  displayName: 'Mixtral 8x22B',
@@ -134,9 +133,9 @@ const Mistral: ModelProviderCard = {
134
133
  input: 2,
135
134
  output: 6,
136
135
  },
137
- tokens: 65_536,
138
136
  },
139
137
  {
138
+ contextWindowTokens: 256_000,
140
139
  description:
141
140
  'Codestral Mamba是专注于代码生成的Mamba 2语言模型,为先进的代码和推理任务提供强力支持。',
142
141
  displayName: 'Codestral Mamba',
@@ -145,7 +144,6 @@ const Mistral: ModelProviderCard = {
145
144
  input: 0.15,
146
145
  output: 0.15,
147
146
  },
148
- tokens: 256_000,
149
147
  },
150
148
  ],
151
149
  checkModel: 'ministral-3b-latest',
@@ -4,31 +4,31 @@ import { ModelProviderCard } from '@/types/llm';
4
4
  const Moonshot: ModelProviderCard = {
5
5
  chatModels: [
6
6
  {
7
+ contextWindowTokens: 8192,
7
8
  description:
8
9
  'Moonshot V1 8K 专为生成短文本任务设计,具有高效的处理性能,能够处理8,192个tokens,非常适合简短对话、速记和快速内容生成。',
9
10
  displayName: 'Moonshot V1 8K',
10
11
  enabled: true,
11
12
  functionCall: true,
12
13
  id: 'moonshot-v1-8k',
13
- tokens: 8192,
14
14
  },
15
15
  {
16
+ contextWindowTokens: 32_768,
16
17
  description:
17
18
  'Moonshot V1 32K 提供中等长度的上下文处理能力,能够处理32,768个tokens,特别适合生成各种长文档和复杂对话,应用于内容创作、报告生成和对话系统等领域。',
18
19
  displayName: 'Moonshot V1 32K',
19
20
  enabled: true,
20
21
  functionCall: true,
21
22
  id: 'moonshot-v1-32k',
22
- tokens: 32_768,
23
23
  },
24
24
  {
25
+ contextWindowTokens: 128_000,
25
26
  description:
26
27
  'Moonshot V1 128K 是一款拥有超长上下文处理能力的模型,适用于生成超长文本,满足复杂的生成任务需求,能够处理多达128,000个tokens的内容,非常适合科研、学术和大型文档生成等应用场景。',
27
28
  displayName: 'Moonshot V1 128K',
28
29
  enabled: true,
29
30
  functionCall: true,
30
31
  id: 'moonshot-v1-128k',
31
- tokens: 128_000,
32
32
  },
33
33
  ],
34
34
  checkModel: 'moonshot-v1-8k',
@@ -4,107 +4,107 @@ import { ModelProviderCard } from '@/types/llm';
4
4
  const Novita: ModelProviderCard = {
5
5
  chatModels: [
6
6
  {
7
+ contextWindowTokens: 8192,
7
8
  description:
8
9
  'Llama 3.1 8B Instruct 是 Meta 推出的最新版本,优化了高质量对话场景,表现优于许多领先的闭源模型。',
9
10
  displayName: 'Llama 3.1 8B Instruct',
10
11
  enabled: true,
11
12
  id: 'meta-llama/llama-3.1-8b-instruct',
12
- tokens: 8192,
13
13
  },
14
14
  {
15
+ contextWindowTokens: 131_072,
15
16
  description:
16
17
  'Llama 3.1 70B Instruct 专为高质量对话而设计,在人类评估中表现突出,特别适合高交互场景。',
17
18
  displayName: 'Llama 3.1 70B Instruct',
18
19
  enabled: true,
19
20
  id: 'meta-llama/llama-3.1-70b-instruct',
20
- tokens: 131_072,
21
21
  },
22
22
  {
23
+ contextWindowTokens: 32_768,
23
24
  description:
24
25
  'Llama 3.1 405B Instruct 是 Meta最新推出的版本,优化用于生成高质量对话,超越了许多领导闭源模型。',
25
26
  displayName: 'Llama 3.1 405B Instruct',
26
27
  enabled: true,
27
28
  id: 'meta-llama/llama-3.1-405b-instruct',
28
- tokens: 32_768,
29
29
  },
30
30
  {
31
+ contextWindowTokens: 8192,
31
32
  description: 'Llama 3 8B Instruct 优化了高质量对话场景,性能优于许多闭源模型。',
32
33
  displayName: 'Llama 3 8B Instruct',
33
34
  id: 'meta-llama/llama-3-8b-instruct',
34
- tokens: 8192,
35
35
  },
36
36
  {
37
+ contextWindowTokens: 8192,
37
38
  description: 'Llama 3 70B Instruct 优化用于高质量对话场景,在各类人类评估中表现优异。',
38
39
  displayName: 'Llama 3 70B Instruct',
39
40
  id: 'meta-llama/llama-3-70b-instruct',
40
- tokens: 8192,
41
41
  },
42
42
  {
43
+ contextWindowTokens: 8192,
43
44
  description: 'Gemma 2 9B 是谷歌的一款开源语言模型,以其在效率和性能方面设立了新的标准。',
44
45
  displayName: 'Gemma 2 9B',
45
46
  enabled: true,
46
47
  id: 'google/gemma-2-9b-it',
47
- tokens: 8192,
48
48
  },
49
49
  {
50
+ contextWindowTokens: 32_768,
50
51
  description: 'Mistral Nemo 是多语言支持和高性能编程的7.3B参数模型。',
51
52
  displayName: 'Mistral Nemo',
52
53
  enabled: true,
53
54
  id: 'mistralai/mistral-nemo',
54
- tokens: 32_768,
55
55
  },
56
56
  {
57
+ contextWindowTokens: 32_768,
57
58
  description: 'Mistral 7B Instruct 是一款兼有速度优化和长上下文支持的高性能行业标准模型。',
58
59
  displayName: 'Mistral 7B Instruct',
59
60
  enabled: true,
60
61
  id: 'mistralai/mistral-7b-instruct',
61
- tokens: 32_768,
62
62
  },
63
63
  {
64
+ contextWindowTokens: 32_768,
64
65
  description: 'WizardLM 2 7B 是微软AI最新的快速轻量化模型,性能接近于现有开源领导模型的10倍。',
65
66
  displayName: 'WizardLM 2 7B',
66
67
  enabled: true,
67
68
  id: 'microsoft/wizardlm 2-7b',
68
- tokens: 32_768,
69
69
  },
70
70
  {
71
+ contextWindowTokens: 65_535,
71
72
  description: 'WizardLM-2 8x22B 是微软AI最先进的Wizard模型,显示出极其竞争力的表现。',
72
73
  displayName: 'WizardLM-2 8x22B',
73
74
  enabled: true,
74
75
  id: 'microsoft/wizardlm-2-8x22b',
75
- tokens: 65_535,
76
76
  },
77
77
  {
78
+ contextWindowTokens: 16_000,
78
79
  description: 'Dolphin Mixtral 8x22B 是一款为指令遵循、对话和编程设计的模型。',
79
80
  displayName: 'Dolphin Mixtral 8x22B',
80
81
  id: 'cognitivecomputations/dolphin-mixtral-8x22b',
81
- tokens: 16_000,
82
82
  },
83
83
  {
84
+ contextWindowTokens: 8192,
84
85
  description:
85
86
  'Hermes 2 Pro Llama 3 8B 是 Nous Hermes 2的升级版本,包含最新的内部开发的数据集。',
86
87
  displayName: 'Hermes 2 Pro Llama 3 8B',
87
88
  id: 'nousresearch/hermes-2-pro-llama-3-8b',
88
- tokens: 8192,
89
89
  },
90
90
  {
91
+ contextWindowTokens: 32_768,
91
92
  description:
92
93
  'Hermes 2 Mixtral 8x7B DPO 是一款高度灵活的多模型合并,旨在提供卓越的创造性体验。',
93
94
  displayName: 'Hermes 2 Mixtral 8x7B DPO',
94
95
  id: 'Nous-Hermes-2-Mixtral-8x7B-DPO',
95
- tokens: 32_768,
96
96
  },
97
97
  {
98
+ contextWindowTokens: 4096,
98
99
  description: 'MythoMax l2 13B 是一款合并了多个顶尖模型的创意与智能相结合的语言模型。',
99
100
  displayName: 'MythoMax l2 13B',
100
101
  id: 'gryphe/mythomax-l2-13b',
101
- tokens: 4096,
102
102
  },
103
103
  {
104
+ contextWindowTokens: 4096,
104
105
  description: 'OpenChat 7B 是经过“C-RLFT(条件强化学习微调)”策略精调的开源语言模型库。',
105
106
  displayName: 'OpenChat 7B',
106
107
  id: 'openchat/openchat-7b',
107
- tokens: 4096,
108
108
  },
109
109
  ],
110
110
  checkModel: 'meta-llama/llama-3.1-8b-instruct',