From 8a8fc6a4409f9593c86a559aba34618f0af67840 Mon Sep 17 00:00:00 2001 From: Arvin Xu Date: Thu, 12 Sep 2024 11:54:36 +0800 Subject: [PATCH 1/4] =?UTF-8?q?=F0=9F=92=84=20style:=20improve=20models=20?= =?UTF-8?q?and=20add=20more=20info=20for=20providers=20and=20models=20(#39?= =?UTF-8?q?11)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * 🔧 chore: add more info for providers and models * 🔧 chore: add more info for providers and models * ✅ test: fix test * ✅ test: fix test * ✅ test: fix test --- src/config/modelProviders/ai360.ts | 91 ++++------------ src/config/modelProviders/anthropic.ts | 41 ++++++- src/config/modelProviders/baichuan.ts | 41 +++++-- src/config/modelProviders/bedrock.ts | 99 +++++++++++++---- src/config/modelProviders/deepseek.ts | 13 ++- src/config/modelProviders/fireworksai.ts | 8 +- src/config/modelProviders/google.ts | 55 +++++++++- src/config/modelProviders/groq.ts | 53 +++++++-- src/config/modelProviders/minimax.ts | 2 + src/config/modelProviders/openai.ts | 103 +++++++++++++++++- src/const/discover.ts | 1 + .../openai/__snapshots__/index.test.ts.snap | 56 ++++++++++ .../config/__snapshots__/index.test.ts.snap | 32 ++++++ src/server/routers/edge/config/index.test.ts | 4 + src/types/llm.ts | 30 ++++- src/utils/parseModels.test.ts | 8 ++ 16 files changed, 522 insertions(+), 115 deletions(-) create mode 100644 src/const/discover.ts diff --git a/src/config/modelProviders/ai360.ts b/src/config/modelProviders/ai360.ts index 2957a4e7b930..1918b4a17fd8 100644 --- a/src/config/modelProviders/ai360.ts +++ b/src/config/modelProviders/ai360.ts @@ -1,6 +1,5 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://ai.360.cn/platform/docs/overview const Ai360: ModelProviderCard = { chatModels: [ { @@ -9,92 +8,46 @@ const Ai360: ModelProviderCard = { functionCall: false, id: '360gpt2-pro', maxOutput: 7000, + pricing: { + currency: 'CNY', + input: 5, + output: 5, + }, tokens: 8192, }, { displayName: '360GPT Pro', - functionCall: false, id: '360gpt-pro', maxOutput: 7000, + pricing: { + currency: 'CNY', + input: 5, + output: 5, + }, tokens: 8192, }, - { - displayName: '360GPT Pro Perf', - functionCall: false, - id: '360gpt-pro-perf', - maxOutput: 7000, - tokens: 8192, - }, - { - displayName: '360GPT Pro sc202401v3', - functionCall: false, - id: '360gpt-pro-sc202401v3', - maxOutput: 2048, - tokens: 4096, - }, - { - displayName: '360GPT Pro sc202401v2', - functionCall: false, - id: '360gpt-pro-sc202401v2', - maxOutput: 2048, - tokens: 4096, - }, - { - displayName: '360GPT Pro sc202401v1', - functionCall: false, - id: '360gpt-pro-sc202401v1', - maxOutput: 2048, - tokens: 4096, - }, - { - displayName: '360GPT Pro v2.0.3', - functionCall: false, - id: '360gpt-pro-v2.0.3', - maxOutput: 2048, - tokens: 4096, - }, { displayName: '360GPT Turbo', enabled: true, - functionCall: false, id: '360gpt-turbo', - maxOutput: 8192, + maxOutput: 7000, + pricing: { + currency: 'CNY', + input: 2, + output: 2, + }, tokens: 8192, }, { displayName: '360GPT Turbo Responsibility 8K', enabled: true, - functionCall: false, id: '360gpt-turbo-responsibility-8k', maxOutput: 2048, - tokens: 8192, - }, - { - displayName: '360GPT Turbo 32K Responsibility 240530', - functionCall: false, - id: '360gpt-turbo-32k-responsibility-240530', - maxOutput: 32_000, - tokens: 32_000, - }, - { - displayName: '360GPT Turbo 32K Responsibility 240516', - functionCall: false, - id: '360gpt-turbo-32k-responsibility-240516', - maxOutput: 32_000, - tokens: 32_000, - }, - { - displayName: '360GPT_S1_QIYUAN', - functionCall: false, - id: '360GPT_S1_QIYUAN', - maxOutput: 2048, - tokens: 4096, - }, - { - displayName: '360GPT_S2_V9', - functionCall: false, - id: '360GPT_S2_V9', - maxOutput: 7000, + pricing: { + currency: 'CNY', + input: 2, + output: 2, + }, tokens: 8192, }, ], @@ -102,7 +55,9 @@ const Ai360: ModelProviderCard = { disableBrowserRequest: true, id: 'ai360', modelList: { showModelFetcher: true }, + modelsUrl: 'https://ai.360.cn/platform/docs/overview', name: '360 AI', + url: 'https://ai.360.com', }; export default Ai360; diff --git a/src/config/modelProviders/anthropic.ts b/src/config/modelProviders/anthropic.ts index dd3505d8c334..c7aa2b1c349d 100644 --- a/src/config/modelProviders/anthropic.ts +++ b/src/config/modelProviders/anthropic.ts @@ -1,6 +1,5 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.anthropic.com/claude/docs/models-overview#model-comparison const Anthropic: ModelProviderCard = { chatModels: [ { @@ -11,6 +10,13 @@ const Anthropic: ModelProviderCard = { functionCall: true, id: 'claude-3-5-sonnet-20240620', maxOutput: 8192, + pricing: { + cachedInput: 0.3, + input: 3, + output: 15, + writeCacheInput: 3.75, + }, + releasedAt: '2024-06-20', tokens: 200_000, vision: true, }, @@ -18,10 +24,14 @@ const Anthropic: ModelProviderCard = { description: 'Ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments', displayName: 'Claude 3 Sonnet', - enabled: true, functionCall: true, id: 'claude-3-sonnet-20240229', maxOutput: 4096, + pricing: { + input: 3, + output: 15, + }, + releasedAt: '2024-02-29', tokens: 200_000, vision: true, }, @@ -33,6 +43,11 @@ const Anthropic: ModelProviderCard = { functionCall: true, id: 'claude-3-opus-20240229', maxOutput: 4096, + pricing: { + input: 15, + output: 75, + }, + releasedAt: '2024-02-29', tokens: 200_000, vision: true, }, @@ -44,6 +59,11 @@ const Anthropic: ModelProviderCard = { functionCall: true, id: 'claude-3-haiku-20240307', maxOutput: 4096, + pricing: { + input: 0.25, + output: 1.25, + }, + releasedAt: '2024-03-07', tokens: 200_000, vision: true, }, @@ -51,27 +71,44 @@ const Anthropic: ModelProviderCard = { displayName: 'Claude 2.1', id: 'claude-2.1', maxOutput: 4096, + pricing: { + input: 8, + output: 24, + }, + releasedAt: '2023-11-21', tokens: 200_000, }, { displayName: 'Claude 2.0', id: 'claude-2.0', maxOutput: 4096, + pricing: { + input: 8, + output: 24, + }, + releasedAt: '2023-07-11', tokens: 100_000, }, { displayName: 'Claude Instant 1.2', id: 'claude-instant-1.2', maxOutput: 4096, + pricing: { + input: 0.8, + output: 2.4, + }, + releasedAt: '2023-08-09', tokens: 100_000, }, ], checkModel: 'claude-3-haiku-20240307', id: 'anthropic', + modelsUrl: 'https://docs.anthropic.com/en/docs/about-claude/models#model-names', name: 'Anthropic', proxyUrl: { placeholder: 'https://api.anthropic.com', }, + url: 'https://anthropic.com', }; export default Anthropic; diff --git a/src/config/modelProviders/baichuan.ts b/src/config/modelProviders/baichuan.ts index bf15e7f97d37..a148e135661c 100644 --- a/src/config/modelProviders/baichuan.ts +++ b/src/config/modelProviders/baichuan.ts @@ -4,54 +4,73 @@ import { ModelProviderCard } from '@/types/llm'; const Baichuan: ModelProviderCard = { chatModels: [ { - description: '模型能力国内第一，在知识百科、长文本、生成创作等中文任务上超越国外主流模型。还具备行业领先的多模态能力，多项权威评测基准表现优异。', + description: + '模型能力国内第一，在知识百科、长文本、生成创作等中文任务上超越国外主流模型。还具备行业领先的多模态能力，多项权威评测基准表现优异。', displayName: 'Baichuan 4', enabled: true, functionCall: true, id: 'Baichuan4', maxOutput: 4096, + pricing: { + currency: 'CNY', + input: 100, + output: 100, + }, tokens: 32_768, }, { - description: '针对企业高频场景优化，效果大幅提升，高性价比。相对于Baichuan2模型，内容创作提升20%，知识问答提升17%， 角色扮演能力提升40%。整体效果比GPT3.5更优。', + description: + '针对企业高频场景优化，效果大幅提升，高性价比。相对于Baichuan2模型，内容创作提升20%，知识问答提升17%， 角色扮演能力提升40%。整体效果比GPT3.5更优。', displayName: 'Baichuan 3 Turbo', enabled: true, functionCall: true, id: 'Baichuan3-Turbo', maxOutput: 8192, + pricing: { + currency: 'CNY', + input: 12, + output: 12, + }, tokens: 32_768, }, { - description: '具备 128K 超长上下文窗口，针对企业高频场景优化，效果大幅提升，高性价比。相对于Baichuan2模型，内容创作提升20%，知识问答提升17%， 角色扮演能力提升40%。整体效果比GPT3.5更优。', + description: + '具备 128K 超长上下文窗口，针对企业高频场景优化，效果大幅提升，高性价比。相对于Baichuan2模型，内容创作提升20%，知识问答提升17%， 角色扮演能力提升40%。整体效果比GPT3.5更优。', displayName: 'Baichuan 3 Turbo 128k', enabled: true, id: 'Baichuan3-Turbo-128k', maxOutput: 4096, + pricing: { + currency: 'CNY', + input: 24, + output: 24, + }, tokens: 128_000, }, { - description: '采用搜索增强技术实现大模型与领域知识、全网知识的全面链接。支持PDF、Word等多种文档上传及网址输入，信息获取及时、全面，输出结果准确、专业。', + description: + '采用搜索增强技术实现大模型与领域知识、全网知识的全面链接。支持PDF、Word等多种文档上传及网址输入，信息获取及时、全面，输出结果准确、专业。', displayName: 'Baichuan 2 Turbo', id: 'Baichuan2-Turbo', maxOutput: 8192, + pricing: { + currency: 'CNY', + input: 8, + output: 8, + }, tokens: 32_768, }, - { - description: '具备 192K 超长上下文窗口，采用搜索增强技术实现大模型与领域知识、全网知识的全面链接。支持PDF、Word等多种文档上传及网址输入，信息获取及时、全面，输出结果准确、专业。', - displayName: 'Baichuan 2 Turbo 192k', - id: 'Baichuan2-Turbo-192k', - maxOutput: 2048, - tokens: 192_000, - }, ], checkModel: 'Baichuan3-Turbo', id: 'baichuan', modelList: { showModelFetcher: true }, + modelsUrl: 'https://platform.baichuan-ai.com/price', name: 'Baichuan', smoothing: { speed: 2, text: true, }, + url: 'https://platform.baichuan-ai.com', }; export default Baichuan; diff --git a/src/config/modelProviders/bedrock.ts b/src/config/modelProviders/bedrock.ts index 38e8a0a02fd4..ec7affe02041 100644 --- a/src/config/modelProviders/bedrock.ts +++ b/src/config/modelProviders/bedrock.ts @@ -1,12 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.aws.amazon.com/bedrock/latest/userguide/model-ids.html // ref https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html // ref https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/models // ref https://us-west-2.console.aws.amazon.com/bedrock/home?region=us-west-2#/models const Bedrock: ModelProviderCard = { chatModels: [ -/* + /* // TODO: Not support for now { description: 'Amazon Titan Text Lite is a light weight efficient model ideal for fine-tuning for English-language tasks, including like summarization and copywriting, where customers want a smaller, more cost-effective model that is also highly customizable.', @@ -28,96 +27,156 @@ const Bedrock: ModelProviderCard = { }, */ { - description: 'Claude 3.5 Sonnet raises the industry bar for intelligence, outperforming competitor models and Claude 3 Opus on a wide range of evaluations, with the speed and cost of our mid-tier model, Claude 3 Sonnet.', + description: + 'Claude 3.5 Sonnet raises the industry bar for intelligence, outperforming competitor models and Claude 3 Opus on a wide range of evaluations, with the speed and cost of our mid-tier model, Claude 3 Sonnet.', displayName: 'Claude 3.5 Sonnet', enabled: true, functionCall: true, id: 'anthropic.claude-3-5-sonnet-20240620-v1:0', + pricing: { + input: 3, + output: 15, + }, tokens: 200_000, vision: true, }, { - description: 'Claude 3 Sonnet by Anthropic strikes the ideal balance between intelligence and speed—particularly for enterprise workloads. It offers maximum utility at a lower price than competitors, and is engineered to be the dependable, high-endurance workhorse for scaled AI deployments. Claude 3 Sonnet can process images and return text outputs, and features a 200K context window.', + description: + 'Claude 3 Sonnet by Anthropic strikes the ideal balance between intelligence and speed—particularly for enterprise workloads. It offers maximum utility at a lower price than competitors, and is engineered to be the dependable, high-endurance workhorse for scaled AI deployments. Claude 3 Sonnet can process images and return text outputs, and features a 200K context window.', displayName: 'Claude 3 Sonnet', enabled: true, functionCall: true, id: 'anthropic.claude-3-sonnet-20240229-v1:0', + pricing: { + input: 3, + output: 15, + }, tokens: 200_000, vision: true, }, { - description: 'Claude 3 Opus is Anthropic most powerful AI model, with state-of-the-art performance on highly complex tasks. It can navigate open-ended prompts and sight-unseen scenarios with remarkable fluency and human-like understanding. Claude 3 Opus shows us the frontier of what’s possible with generative AI. Claude 3 Opus can process images and return text outputs, and features a 200K context window.', + description: + 'Claude 3 Opus is Anthropic most powerful AI model, with state-of-the-art performance on highly complex tasks. It can navigate open-ended prompts and sight-unseen scenarios with remarkable fluency and human-like understanding. Claude 3 Opus shows us the frontier of what’s possible with generative AI. Claude 3 Opus can process images and return text outputs, and features a 200K context window.', displayName: 'Claude 3 Opus', enabled: true, functionCall: true, id: 'anthropic.claude-3-opus-20240229-v1:0', + pricing: { + input: 15, + output: 75, + }, tokens: 200_000, vision: true, }, { - description: 'Claude 3 Haiku is Anthropic fastest, most compact model for near-instant responsiveness. It answers simple queries and requests with speed. Customers will be able to build seamless AI experiences that mimic human interactions. Claude 3 Haiku can process images and return text outputs, and features a 200K context window.', + description: + 'Claude 3 Haiku is Anthropic fastest, most compact model for near-instant responsiveness. It answers simple queries and requests with speed. Customers will be able to build seamless AI experiences that mimic human interactions. Claude 3 Haiku can process images and return text outputs, and features a 200K context window.', displayName: 'Claude 3 Haiku', enabled: true, functionCall: true, id: 'anthropic.claude-3-haiku-20240307-v1:0', + pricing: { + input: 0.25, + output: 1.25, + }, tokens: 200_000, vision: true, }, { - description: 'An update to Claude 2 that features double the context window, plus improvements across reliability, hallucination rates, and evidence-based accuracy in long document and RAG contexts.', + description: + 'An update to Claude 2 that features double the context window, plus improvements across reliability, hallucination rates, and evidence-based accuracy in long document and RAG contexts.', displayName: 'Claude 2.1', id: 'anthropic.claude-v2:1', + pricing: { + input: 8, + output: 24, + }, tokens: 200_000, }, { - description: 'Anthropic highly capable model across a wide range of tasks from sophisticated dialogue and creative content generation to detailed instruction following.', + description: + 'Anthropic highly capable model across a wide range of tasks from sophisticated dialogue and creative content generation to detailed instruction following.', displayName: 'Claude 2.0', id: 'anthropic.claude-v2', + pricing: { + input: 8, + output: 24, + }, tokens: 100_000, }, { - description: 'A fast, affordable yet still very capable model, which can handle a range of tasks including casual dialogue, text analysis, summarization, and document question-answering.', + description: + 'A fast, affordable yet still very capable model, which can handle a range of tasks including casual dialogue, text analysis, summarization, and document question-answering.', displayName: 'Claude Instant', id: 'anthropic.claude-instant-v1', + pricing: { + input: 0.8, + output: 2.4, + }, tokens: 100_000, }, { - description: 'An update to Meta Llama 3 8B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + description: + 'An update to Meta Llama 3 8B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', displayName: 'Llama 3.1 8B Instruct', enabled: true, functionCall: true, id: 'meta.llama3-1-8b-instruct-v1:0', + pricing: { + input: 0.22, + output: 0.22, + }, tokens: 128_000, }, { - description: 'An update to Meta Llama 3 70B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + description: + 'An update to Meta Llama 3 70B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', displayName: 'Llama 3.1 70B Instruct', enabled: true, functionCall: true, id: 'meta.llama3-1-70b-instruct-v1:0', + pricing: { + input: 0.99, + output: 0.99, + }, tokens: 128_000, }, { - description: 'Meta Llama 3.1 405B Instruct is the largest and most powerful of the Llama 3.1 Instruct models that is a highly advanced model for conversational inference and reasoning, synthetic data generation, and a base to do specialized continual pre-training or fine-tuning on a specific domain. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + description: + 'Meta Llama 3.1 405B Instruct is the largest and most powerful of the Llama 3.1 Instruct models that is a highly advanced model for conversational inference and reasoning, synthetic data generation, and a base to do specialized continual pre-training or fine-tuning on a specific domain. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', displayName: 'Llama 3.1 405B Instruct', enabled: true, functionCall: true, id: 'meta.llama3-1-405b-instruct-v1:0', + pricing: { + input: 5.32, + output: 16, + }, tokens: 128_000, }, { - description: 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for limited computational power and resources, edge devices, and faster training times.', + description: + 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for limited computational power and resources, edge devices, and faster training times.', displayName: 'Llama 3 8B Instruct', id: 'meta.llama3-8b-instruct-v1:0', + pricing: { + input: 0.3, + output: 0.6, + }, tokens: 8000, }, { - description: 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for content creation, conversational AI, language understanding, R&D, and Enterprise applications.', + description: + 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for content creation, conversational AI, language understanding, R&D, and Enterprise applications.', displayName: 'Llama 3 70B Instruct', id: 'meta.llama3-70b-instruct-v1:0', + pricing: { + input: 2.65, + output: 3.5, + }, tokens: 8000, }, -/* + /* // TODO: Not support for now { description: 'A 7B dense Transformer, fast-deployed and easily customisable. Small, yet powerful for a variety of use cases. Supports English and code, and a 32k context window.', @@ -157,7 +216,7 @@ const Bedrock: ModelProviderCard = { tokens: 32_000, }, */ -/* + /* // TODO: Not support for now { description: 'Command R+ is a highly performant generative language model optimized for large scale production workloads.', @@ -176,7 +235,7 @@ const Bedrock: ModelProviderCard = { tokens: 128_000, }, */ -/* + /* // Cohere Command (Text) and AI21 Labs Jurassic-2 (Text) don't support chat with the Converse API { description: 'Command is Cohere flagship text generation model. It is trained to follow user commands and to be instantly useful in practical business applications.', @@ -191,7 +250,7 @@ const Bedrock: ModelProviderCard = { tokens: 4000, }, */ -/* + /* // TODO: Not support for now { description: 'The latest Foundation Model from AI21 Labs, Jamba-Instruct offers an impressive 256K context window and delivers the best value per price on core text generation, summarization, and question answering tasks for the enterprise.', @@ -200,7 +259,7 @@ const Bedrock: ModelProviderCard = { tokens: 256_000, }, */ -/* + /* // Cohere Command (Text) and AI21 Labs Jurassic-2 (Text) don't support chat with the Converse API { description: 'Jurassic-2 Mid is less powerful than Ultra, yet carefully designed to strike the right balance between exceptional quality and affordability. Jurassic-2 Mid can be applied to any language comprehension or generation task including question answering, summarization, long-form copy generation, advanced information extraction and many others.', @@ -218,7 +277,9 @@ const Bedrock: ModelProviderCard = { ], checkModel: 'anthropic.claude-instant-v1', id: 'bedrock', + modelsUrl: 'https://docs.aws.amazon.com/bedrock/latest/userguide/model-ids.html', name: 'Bedrock', + url: 'https://docs.aws.amazon.com/bedrock/latest/userguide/what-is-bedrock.html', }; export default Bedrock; diff --git a/src/config/modelProviders/deepseek.ts b/src/config/modelProviders/deepseek.ts index 7ceca9dada7f..842b1ca58d03 100644 --- a/src/config/modelProviders/deepseek.ts +++ b/src/config/modelProviders/deepseek.ts @@ -4,18 +4,27 @@ import { ModelProviderCard } from '@/types/llm'; const DeepSeek: ModelProviderCard = { chatModels: [ { - description: '擅长通用对话任务', - displayName: 'DeepSeek V2.5', + description: + '融合通用与代码能力的全新开源模型, 不仅保留了原有 Chat 模型的通用对话能力和 Coder 模型的强大代码处理能力，还更好地对齐了人类偏好。此外，DeepSeek-V2.5 在写作任务、指令跟随等多个方面也实现了大幅提升。', + displayName: 'DeepSeek-V2.5', enabled: true, functionCall: true, id: 'deepseek-chat', + pricing: { + cachedInput: 0.014, + input: 0.14, + output: 0.28, + }, + releasedAt: '2024-09-05', tokens: 128_000, }, ], checkModel: 'deepseek-chat', id: 'deepseek', modelList: { showModelFetcher: true }, + modelsUrl: 'https://platform.deepseek.com/api-docs/zh-cn/quick_start/pricing', name: 'DeepSeek', + url: 'https://deepseek.com', }; export default DeepSeek; diff --git a/src/config/modelProviders/fireworksai.ts b/src/config/modelProviders/fireworksai.ts index d34cdff7177c..5e544822525c 100644 --- a/src/config/modelProviders/fireworksai.ts +++ b/src/config/modelProviders/fireworksai.ts @@ -5,7 +5,8 @@ import { ModelProviderCard } from '@/types/llm'; const FireworksAI: ModelProviderCard = { chatModels: [ { - description: 'Fireworks latest and most performant function-calling model. Firefunction-v2 is based on Llama-3 and trained to excel at function-calling as well as chat and instruction-following. See blog post for more details https://fireworks.ai/blog/firefunction-v2-launch-post', + description: + 'Fireworks latest and most performant function-calling model. Firefunction-v2 is based on Llama-3 and trained to excel at function-calling as well as chat and instruction-following. See blog post for more details https://fireworks.ai/blog/firefunction-v2-launch-post', displayName: 'Firefunction V2', enabled: true, functionCall: true, @@ -20,7 +21,8 @@ const FireworksAI: ModelProviderCard = { tokens: 32_768, }, { - description: 'Vision-language model allowing both image and text as inputs (single image is recommended), trained on OSS model generated training data and open sourced on huggingface at fireworks-ai/FireLLaVA-13b', + description: + 'Vision-language model allowing both image and text as inputs (single image is recommended), trained on OSS model generated training data and open sourced on huggingface at fireworks-ai/FireLLaVA-13b', displayName: 'FireLLaVA-13B', enabled: true, functionCall: false, @@ -137,7 +139,9 @@ const FireworksAI: ModelProviderCard = { checkModel: 'accounts/fireworks/models/firefunction-v2', id: 'fireworksai', modelList: { showModelFetcher: true }, + modelsUrl: 'https://fireworks.ai/models?show=Serverless', name: 'Fireworks AI', + url: 'https://fireworks.ai', }; export default FireworksAI; diff --git a/src/config/modelProviders/google.ts b/src/config/modelProviders/google.ts index 7a657d100476..45cea49fc86a 100644 --- a/src/config/modelProviders/google.ts +++ b/src/config/modelProviders/google.ts @@ -1,7 +1,5 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://ai.google.dev/models/gemini -// api https://ai.google.dev/api/rest/v1beta/models/list const Google: ModelProviderCard = { chatModels: [ { @@ -11,6 +9,11 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-flash-latest', maxOutput: 8192, + pricing: { + cachedInput: 0.018_75, + input: 0.075, + output: 0.3, + }, tokens: 1_048_576 + 8192, vision: true, }, @@ -21,6 +24,12 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-flash-exp-0827', maxOutput: 8192, + pricing: { + cachedInput: 0.018_75, + input: 0.075, + output: 0.3, + }, + releasedAt: '2024-08-27', tokens: 1_048_576 + 8192, vision: true, }, @@ -30,6 +39,7 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-flash-8b-exp-0827', maxOutput: 8192, + releasedAt: '2024-08-27', tokens: 1_048_576 + 8192, vision: true, }, @@ -49,6 +59,12 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-pro-latest', maxOutput: 8192, + pricing: { + cachedInput: 0.875, + input: 3.5, + output: 10.5, + }, + releasedAt: '2024-02-15', tokens: 2_097_152 + 8192, vision: true, }, @@ -59,6 +75,12 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-pro-exp-0827', maxOutput: 8192, + pricing: { + cachedInput: 0.875, + input: 3.5, + output: 10.5, + }, + releasedAt: '2024-08-27', tokens: 2_097_152 + 8192, vision: true, }, @@ -68,6 +90,12 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-pro-exp-0801', maxOutput: 8192, + pricing: { + cachedInput: 0.875, + input: 3.5, + output: 10.5, + }, + releasedAt: '2024-08-01', tokens: 2_097_152 + 8192, vision: true, }, @@ -77,6 +105,12 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.5-pro-001', maxOutput: 8192, + pricing: { + cachedInput: 0.875, + input: 3.5, + output: 10.5, + }, + releasedAt: '2024-02-15', tokens: 2_097_152 + 8192, vision: true, }, @@ -86,6 +120,11 @@ const Google: ModelProviderCard = { displayName: 'Gemini 1.0 Pro', id: 'gemini-1.0-pro-latest', maxOutput: 2048, + pricing: { + input: 0.5, + output: 1.5, + }, + releasedAt: '2023-12-06', tokens: 30_720 + 2048, }, { @@ -95,6 +134,11 @@ const Google: ModelProviderCard = { functionCall: true, id: 'gemini-1.0-pro-001', maxOutput: 2048, + pricing: { + input: 0.5, + output: 1.5, + }, + releasedAt: '2023-12-06', tokens: 30_720 + 2048, }, { @@ -103,11 +147,17 @@ const Google: ModelProviderCard = { displayName: 'Gemini 1.0 Pro 002 (Tuning)', id: 'gemini-1.0-pro-002', maxOutput: 2048, + pricing: { + input: 0.5, + output: 1.5, + }, + releasedAt: '2023-12-06', tokens: 30_720 + 2048, }, ], checkModel: 'gemini-1.5-flash-latest', id: 'google', + modelsUrl: 'https://ai.google.dev/gemini-api/docs/models/gemini', name: 'Google', proxyUrl: { placeholder: 'https://generativelanguage.googleapis.com', @@ -116,6 +166,7 @@ const Google: ModelProviderCard = { speed: 2, text: true, }, + url: 'https://ai.google.dev', }; export default Google; diff --git a/src/config/modelProviders/groq.ts b/src/config/modelProviders/groq.ts index 5fa49f9c1a87..0aa8bf6e4a73 100644 --- a/src/config/modelProviders/groq.ts +++ b/src/config/modelProviders/groq.ts @@ -1,6 +1,5 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://console.groq.com/docs/models // ref https://console.groq.com/docs/tool-use const Groq: ModelProviderCard = { chatModels: [ @@ -10,6 +9,11 @@ const Groq: ModelProviderCard = { enabled: true, functionCall: true, id: 'llama-3.1-8b-instant', + maxOutput: 8192, + pricing: { + input: 0.05, + output: 0.08, + }, tokens: 131_072, }, { @@ -17,9 +21,14 @@ const Groq: ModelProviderCard = { enabled: true, functionCall: true, id: 'llama-3.1-70b-versatile', + maxOutput: 8192, + pricing: { + input: 0.59, + output: 0.79, + }, tokens: 131_072, }, -/* + /* // Offline due to overwhelming demand! Stay tuned for updates. { displayName: 'Llama 3.1 405B (Preview)', @@ -29,17 +38,25 @@ const Groq: ModelProviderCard = { }, */ { - displayName: 'Llama 3 Groq 8B Tool Use (Preview)', + displayName: 'Llama 3 Groq 8B Tool Use Preview 8K', enabled: true, functionCall: true, id: 'llama3-groq-8b-8192-tool-use-preview', + pricing: { + input: 0.19, + output: 0.19, + }, tokens: 8192, }, { - displayName: 'Llama 3 Groq 70B Tool Use (Preview)', + displayName: 'Llama 3 Groq 70B Tool Use Preview 8K', enabled: true, functionCall: true, id: 'llama3-groq-70b-8192-tool-use-preview', + pricing: { + input: 0.89, + output: 0.89, + }, tokens: 8192, }, { @@ -47,6 +64,10 @@ const Groq: ModelProviderCard = { enabled: true, functionCall: true, id: 'llama3-8b-8192', + pricing: { + input: 0.05, + output: 0.08, + }, tokens: 8192, }, { @@ -54,26 +75,42 @@ const Groq: ModelProviderCard = { enabled: true, functionCall: true, id: 'llama3-70b-8192', + pricing: { + input: 0.59, + output: 0.79, + }, tokens: 8192, }, { - displayName: 'Gemma 2 9B', + displayName: 'Gemma 2 9B 8k', enabled: true, functionCall: true, id: 'gemma2-9b-it', + pricing: { + input: 0.2, + output: 0.2, + }, tokens: 8192, }, { - displayName: 'Gemma 7B', + displayName: 'Gemma 7B 8k Instruct', functionCall: true, id: 'gemma-7b-it', + pricing: { + input: 0.07, + output: 0.07, + }, tokens: 8192, }, { - displayName: 'Mixtral 8x7B', + displayName: 'Mixtral 8x7B Instruct 32k', enabled: true, functionCall: true, id: 'mixtral-8x7b-32768', + pricing: { + input: 0.24, + output: 0.24, + }, tokens: 32_768, }, { @@ -86,10 +123,12 @@ const Groq: ModelProviderCard = { ], checkModel: 'gemma2-9b-it', id: 'groq', + modelsUrl: 'https://console.groq.com/docs/models', name: 'Groq', proxyUrl: { placeholder: 'https://api.groq.com/openai/v1', }, + url: 'https://groq.com', }; export default Groq; diff --git a/src/config/modelProviders/minimax.ts b/src/config/modelProviders/minimax.ts index 600f48c3a66b..6f114679bbc9 100644 --- a/src/config/modelProviders/minimax.ts +++ b/src/config/modelProviders/minimax.ts @@ -42,11 +42,13 @@ const Minimax: ModelProviderCard = { ], checkModel: 'abab6.5s-chat', id: 'minimax', + modelsUrl: 'https://platform.minimaxi.com/document/Models', name: 'Minimax', smoothing: { speed: 2, text: true, }, + url: 'https://www.minimaxi.com', }; export default Minimax; diff --git a/src/config/modelProviders/openai.ts b/src/config/modelProviders/openai.ts index d888942e893d..0fa8a70efb18 100644 --- a/src/config/modelProviders/openai.ts +++ b/src/config/modelProviders/openai.ts @@ -1,7 +1,6 @@ import { ModelProviderCard } from '@/types/llm'; // ref: -// https://platform.openai.com/docs/models // https://platform.openai.com/docs/deprecations const OpenAI: ModelProviderCard = { chatModels: [ @@ -12,6 +11,10 @@ const OpenAI: ModelProviderCard = { functionCall: true, id: 'gpt-4o-mini', maxOutput: 16_385, + pricing: { + input: 0.15, + output: 0.6, + }, tokens: 128_000, vision: true, }, @@ -21,6 +24,10 @@ const OpenAI: ModelProviderCard = { enabled: true, functionCall: true, id: 'gpt-4o', + pricing: { + input: 5, + output: 15, + }, tokens: 128_000, vision: true, }, @@ -30,6 +37,22 @@ const OpenAI: ModelProviderCard = { enabled: true, functionCall: true, id: 'gpt-4o-2024-08-06', + pricing: { + input: 2.5, + output: 10, + }, + tokens: 128_000, + vision: true, + }, + { + description: 'Latest GPT-4o snapshot that supports Structured Outputs', + displayName: 'GPT-4o (240513)', + functionCall: true, + id: 'gpt-4o-2024-05-13', + pricing: { + input: 5, + output: 15, + }, tokens: 128_000, vision: true, }, @@ -38,6 +61,10 @@ const OpenAI: ModelProviderCard = { displayName: 'ChatGPT-4o', enabled: true, id: 'chatgpt-4o-latest', + pricing: { + input: 5, + output: 15, + }, tokens: 128_000, vision: true, }, @@ -46,6 +73,10 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4 Turbo', functionCall: true, id: 'gpt-4-turbo', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, vision: true, }, @@ -54,6 +85,10 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4 Turbo Vision (240409)', functionCall: true, id: 'gpt-4-turbo-2024-04-09', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, vision: true, }, @@ -62,24 +97,40 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4 Turbo Preview', functionCall: true, id: 'gpt-4-turbo-preview', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, }, { displayName: 'GPT-4 Turbo Preview (0125)', functionCall: true, id: 'gpt-4-0125-preview', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, }, { description: 'Currently points to gpt-4-1106-vision-preview', // Will be discontinued on December 6, 2024 displayName: 'GPT-4 Turbo Vision Preview', id: 'gpt-4-vision-preview', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, vision: true, }, { displayName: 'GPT-4 Turbo Vision Preview (1106)', // Will be discontinued on December 6, 2024 id: 'gpt-4-1106-vision-preview', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, vision: true, }, @@ -87,6 +138,10 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4 Turbo Preview (1106)', functionCall: true, id: 'gpt-4-1106-preview', + pricing: { + input: 10, + output: 30, + }, tokens: 128_000, }, { @@ -94,12 +149,20 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4', functionCall: true, id: 'gpt-4', + pricing: { + input: 30, + output: 60, + }, tokens: 8192, }, { displayName: 'GPT-4 (0613)', functionCall: true, id: 'gpt-4-0613', + pricing: { + input: 30, + output: 60, + }, tokens: 8192, }, { @@ -107,12 +170,20 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-4 32K', functionCall: true, id: 'gpt-4-32k', + pricing: { + input: 60, + output: 120, + }, tokens: 32_768, }, { displayName: 'GPT-4 32K (0613)', // Will be discontinued on June 6, 2025 functionCall: true, id: 'gpt-4-32k-0613', + pricing: { + input: 60, + output: 120, + }, tokens: 32_768, }, { @@ -121,23 +192,39 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-3.5 Turbo', functionCall: true, id: 'gpt-3.5-turbo', + pricing: { + input: 0.5, + output: 1.5, + }, tokens: 16_385, }, { displayName: 'GPT-3.5 Turbo (0125)', functionCall: true, id: 'gpt-3.5-turbo-0125', + pricing: { + input: 0.5, + output: 1.5, + }, tokens: 16_385, }, { displayName: 'GPT-3.5 Turbo (1106)', functionCall: true, id: 'gpt-3.5-turbo-1106', + pricing: { + input: 1, + output: 2, + }, tokens: 16_385, }, { displayName: 'GPT-3.5 Turbo Instruct', id: 'gpt-3.5-turbo-instruct', + pricing: { + input: 1.5, + output: 2, + }, tokens: 4096, }, { @@ -145,18 +232,30 @@ const OpenAI: ModelProviderCard = { displayName: 'GPT-3.5 Turbo 16K', id: 'gpt-3.5-turbo-16k', legacy: true, + pricing: { + input: 3, + output: 4, + }, tokens: 16_385, }, { displayName: 'GPT-3.5 Turbo (0613)', // Will be discontinued on September 13, 2024 id: 'gpt-3.5-turbo-0613', legacy: true, + pricing: { + input: 1.5, + output: 2, + }, tokens: 4096, }, { description: 'Currently points to gpt-3.5-turbo-16k-0613', // Will be discontinued on September 13, 2024 id: 'gpt-3.5-turbo-16k-0613', legacy: true, + pricing: { + input: 3, + output: 4, + }, tokens: 16_385, }, ], @@ -164,7 +263,9 @@ const OpenAI: ModelProviderCard = { enabled: true, id: 'openai', modelList: { showModelFetcher: true }, + modelsUrl: 'https://platform.openai.com/docs/models', name: 'OpenAI', + url: 'https://openai.com', }; export default OpenAI; diff --git a/src/const/discover.ts b/src/const/discover.ts new file mode 100644 index 000000000000..3d202148e8a2 --- /dev/null +++ b/src/const/discover.ts @@ -0,0 +1 @@ +export const CNYtoUSD = 7.14; diff --git a/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap b/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap index 9bc50e9eef48..8f6a97d67eec 100644 --- a/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap +++ b/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap @@ -7,6 +7,10 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-3.5 Turbo", "functionCall": true, "id": "gpt-3.5-turbo", + "pricing": { + "input": 0.5, + "output": 1.5, + }, "tokens": 16385, }, { @@ -14,17 +18,29 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-3.5 Turbo 16K", "id": "gpt-3.5-turbo-16k", "legacy": true, + "pricing": { + "input": 3, + "output": 4, + }, "tokens": 16385, }, { "description": "Currently points to gpt-3.5-turbo-16k-0613", "id": "gpt-3.5-turbo-16k-0613", "legacy": true, + "pricing": { + "input": 3, + "output": 4, + }, "tokens": 16385, }, { "displayName": "GPT-4 Turbo Vision Preview (1106)", "id": "gpt-4-1106-vision-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, "vision": true, }, @@ -35,6 +51,10 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-4 Turbo Preview (0125)", "functionCall": true, "id": "gpt-4-0125-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, }, { @@ -42,11 +62,19 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-4 Turbo Preview", "functionCall": true, "id": "gpt-4-turbo-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, }, { "displayName": "GPT-3.5 Turbo Instruct", "id": "gpt-3.5-turbo-instruct", + "pricing": { + "input": 1.5, + "output": 2, + }, "tokens": 4096, }, { @@ -56,24 +84,40 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-3.5 Turbo (0613)", "id": "gpt-3.5-turbo-0613", "legacy": true, + "pricing": { + "input": 1.5, + "output": 2, + }, "tokens": 4096, }, { "displayName": "GPT-3.5 Turbo (1106)", "functionCall": true, "id": "gpt-3.5-turbo-1106", + "pricing": { + "input": 1, + "output": 2, + }, "tokens": 16385, }, { "displayName": "GPT-4 Turbo Preview (1106)", "functionCall": true, "id": "gpt-4-1106-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, }, { "description": "Currently points to gpt-4-1106-vision-preview", "displayName": "GPT-4 Turbo Vision Preview", "id": "gpt-4-vision-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, "vision": true, }, @@ -82,18 +126,30 @@ exports[`LobeOpenAI > models > should get models 1`] = ` "displayName": "GPT-4", "functionCall": true, "id": "gpt-4", + "pricing": { + "input": 30, + "output": 60, + }, "tokens": 8192, }, { "displayName": "GPT-3.5 Turbo (0125)", "functionCall": true, "id": "gpt-3.5-turbo-0125", + "pricing": { + "input": 0.5, + "output": 1.5, + }, "tokens": 16385, }, { "displayName": "GPT-4 (0613)", "functionCall": true, "id": "gpt-4-0613", + "pricing": { + "input": 30, + "output": 60, + }, "tokens": 8192, }, ] diff --git a/src/server/routers/edge/config/__snapshots__/index.test.ts.snap b/src/server/routers/edge/config/__snapshots__/index.test.ts.snap index 14926665383e..61a7f4eae096 100644 --- a/src/server/routers/edge/config/__snapshots__/index.test.ts.snap +++ b/src/server/routers/edge/config/__snapshots__/index.test.ts.snap @@ -24,6 +24,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-4-0125-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, }, ], @@ -37,6 +41,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-3.5-turbo-1106", + "pricing": { + "input": 1, + "output": 2, + }, "tokens": 16385, }, { @@ -45,6 +53,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-3.5-turbo", + "pricing": { + "input": 0.5, + "output": 1.5, + }, "tokens": 16385, }, { @@ -53,6 +65,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "id": "gpt-3.5-turbo-16k", "legacy": true, + "pricing": { + "input": 3, + "output": 4, + }, "tokens": 16385, }, { @@ -61,6 +77,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-4", + "pricing": { + "input": 30, + "output": 60, + }, "tokens": 8192, }, { @@ -69,6 +89,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-4-32k", + "pricing": { + "input": 60, + "output": 120, + }, "tokens": 32768, }, { @@ -76,6 +100,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "enabled": true, "functionCall": true, "id": "gpt-4-1106-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, }, { @@ -83,6 +111,10 @@ exports[`configRouter > getGlobalConfig > Model Provider env > OPENAI_MODEL_LIST "displayName": "GPT-4 Turbo Vision Preview", "enabled": true, "id": "gpt-4-vision-preview", + "pricing": { + "input": 10, + "output": 30, + }, "tokens": 128000, "vision": true, }, diff --git a/src/server/routers/edge/config/index.test.ts b/src/server/routers/edge/config/index.test.ts index 9a7407c58c6c..7b5bb041d9ca 100644 --- a/src/server/routers/edge/config/index.test.ts +++ b/src/server/routers/edge/config/index.test.ts @@ -96,6 +96,10 @@ describe('configRouter', () => { enabled: true, id: 'gpt-4-1106-preview', tokens: 128000, + pricing: { + input: 10, + output: 30, + }, }); process.env.OPENAI_MODEL_LIST = ''; diff --git a/src/types/llm.ts b/src/types/llm.ts index b812e853abae..98faf65872ee 100644 --- a/src/types/llm.ts +++ b/src/types/llm.ts @@ -33,10 +33,29 @@ export interface ChatModelCard { */ legacy?: boolean; maxOutput?: number; + pricing?: { + cachedInput?: number; + /** + * the currency of the pricing + * @default USD + */ + currency?: 'CNY' | 'USD'; + /** + * the input pricing, e.g. $1 / 1M tokens + */ + input?: number; + /** + * the output pricing, e.g. $2 / 1M tokens + */ + output?: number; + writeCacheInput?: number; + }; + releasedAt?: string; /** * the context window (or input + output tokens limit) */ tokens?: number; + /** * whether model supports vision */ @@ -79,6 +98,10 @@ export interface ModelProviderCard { placeholder?: string; showModelFetcher?: boolean; }; + /** + * the url show the all models in the provider + */ + modelsUrl?: string; /** * the name show for end user */ @@ -90,16 +113,21 @@ export interface ModelProviderCard { title?: string; } | false; + /** * whether show api key in the provider config * so provider like ollama don't need api key field */ showApiKey?: boolean; - /** * whether to smoothing the output */ smoothing?: SmoothingParams; + + /** + * provider's website url + */ + url?: string; } // 语言模型的设置参数 diff --git a/src/utils/parseModels.test.ts b/src/utils/parseModels.test.ts index 43b47703905b..9bc2c6058b60 100644 --- a/src/utils/parseModels.test.ts +++ b/src/utils/parseModels.test.ts @@ -274,6 +274,10 @@ describe('transformToChatModelCards', () => { enabled: true, id: 'gpt-4-0125-preview', tokens: 128000, + pricing: { + input: 10, + output: 30, + }, }, { description: 'GPT-4 Turbo 视觉版 (240409)', @@ -284,6 +288,10 @@ describe('transformToChatModelCards', () => { id: 'gpt-4-turbo-2024-04-09', tokens: 128000, vision: true, + pricing: { + input: 10, + output: 30, + }, }, ]); }); From 483f91671a1645327ee5f828918a9c6e9536f224 Mon Sep 17 00:00:00 2001 From: semantic-release-bot Date: Thu, 12 Sep 2024 04:01:21 +0000 Subject: [PATCH 2/4] :bookmark: chore(release): v1.16.8 [skip ci] MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit ### [Version 1.16.8](https://github.com/lobehub/lobe-chat/compare/v1.16.7...v1.16.8) Released on **2024-09-12** #### 💄 Styles - **misc**: Improve models and add more info for providers and models.
Improvements and Fixes #### Styles * **misc**: Improve models and add more info for providers and models, closes [#3911](https://github.com/lobehub/lobe-chat/issues/3911) ([8a8fc6a](https://github.com/lobehub/lobe-chat/commit/8a8fc6a))
[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
--- CHANGELOG.md | 25 +++++++++++++++++++++++++ package.json | 2 +- 2 files changed, 26 insertions(+), 1 deletion(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index cf4d6d1c8427..f966304f6e96 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -2,6 +2,31 @@ # Changelog +### [Version 1.16.8](https://github.com/lobehub/lobe-chat/compare/v1.16.7...v1.16.8) + +Released on **2024-09-12** + +#### 💄 Styles + +- **misc**: Improve models and add more info for providers and models. + +
+ +
+Improvements and Fixes + +#### Styles + +- **misc**: Improve models and add more info for providers and models, closes [#3911](https://github.com/lobehub/lobe-chat/issues/3911) ([8a8fc6a](https://github.com/lobehub/lobe-chat/commit/8a8fc6a)) + +
+ +
+ +[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top) + +
+ ### [Version 1.16.7](https://github.com/lobehub/lobe-chat/compare/v1.16.6...v1.16.7) Released on **2024-09-11** diff --git a/package.json b/package.json index a488db36dcf0..907a92a3097d 100644 --- a/package.json +++ b/package.json @@ -1,6 +1,6 @@ { "name": "@lobehub/chat", - "version": "1.16.7", + "version": "1.16.8", "description": "Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.", "keywords": [ "framework", From cfbf44f0ef173c14486976376389fae5a85e3d75 Mon Sep 17 00:00:00 2001 From: LovelyGuYiMeng <76251800+LovelyGuYiMeng@users.noreply.github.com> Date: Thu, 12 Sep 2024 13:34:59 +0800 Subject: [PATCH 3/4] =?UTF-8?q?=F0=9F=92=84style:=20update=20claude=20mode?= =?UTF-8?q?l=20list=20sort=20and=20update=20groq=20model=20info=20(#3918)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * Update anthropic.ts * Update groq.ts * Update groq.ts * Update anthropic.ts --- src/config/modelProviders/anthropic.ts | 33 +++++++++++++------------- src/config/modelProviders/groq.ts | 13 ++++------ 2 files changed, 22 insertions(+), 24 deletions(-) diff --git a/src/config/modelProviders/anthropic.ts b/src/config/modelProviders/anthropic.ts index c7aa2b1c349d..ec2721767a22 100644 --- a/src/config/modelProviders/anthropic.ts +++ b/src/config/modelProviders/anthropic.ts @@ -20,10 +20,27 @@ const Anthropic: ModelProviderCard = { tokens: 200_000, vision: true, }, + { + description: + 'Fastest and most compact model for near-instant responsiveness. Quick and accurate targeted performance', + displayName: 'Claude 3 Haiku', + enabled: true, + functionCall: true, + id: 'claude-3-haiku-20240307', + maxOutput: 4096, + pricing: { + input: 0.25, + output: 1.25, + }, + releasedAt: '2024-03-07', + tokens: 200_000, + vision: true, + }, { description: 'Ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments', displayName: 'Claude 3 Sonnet', + enabled: true, functionCall: true, id: 'claude-3-sonnet-20240229', maxOutput: 4096, @@ -51,22 +68,6 @@ const Anthropic: ModelProviderCard = { tokens: 200_000, vision: true, }, - { - description: - 'Fastest and most compact model for near-instant responsiveness. Quick and accurate targeted performance', - displayName: 'Claude 3 Haiku', - enabled: true, - functionCall: true, - id: 'claude-3-haiku-20240307', - maxOutput: 4096, - pricing: { - input: 0.25, - output: 1.25, - }, - releasedAt: '2024-03-07', - tokens: 200_000, - vision: true, - }, { displayName: 'Claude 2.1', id: 'claude-2.1', diff --git a/src/config/modelProviders/groq.ts b/src/config/modelProviders/groq.ts index 0aa8bf6e4a73..eee858f13663 100644 --- a/src/config/modelProviders/groq.ts +++ b/src/config/modelProviders/groq.ts @@ -38,7 +38,7 @@ const Groq: ModelProviderCard = { }, */ { - displayName: 'Llama 3 Groq 8B Tool Use Preview 8K', + displayName: 'Llama 3 Groq 8B Tool Use (Preview)', enabled: true, functionCall: true, id: 'llama3-groq-8b-8192-tool-use-preview', @@ -49,7 +49,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { - displayName: 'Llama 3 Groq 70B Tool Use Preview 8K', + displayName: 'Llama 3 Groq 70B Tool Use (Preview)', enabled: true, functionCall: true, id: 'llama3-groq-70b-8192-tool-use-preview', @@ -61,7 +61,6 @@ const Groq: ModelProviderCard = { }, { displayName: 'Meta Llama 3 8B', - enabled: true, functionCall: true, id: 'llama3-8b-8192', pricing: { @@ -72,7 +71,6 @@ const Groq: ModelProviderCard = { }, { displayName: 'Meta Llama 3 70B', - enabled: true, functionCall: true, id: 'llama3-70b-8192', pricing: { @@ -82,7 +80,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { - displayName: 'Gemma 2 9B 8k', + displayName: 'Gemma 2 9B', enabled: true, functionCall: true, id: 'gemma2-9b-it', @@ -93,7 +91,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { - displayName: 'Gemma 7B 8k Instruct', + displayName: 'Gemma 7B', functionCall: true, id: 'gemma-7b-it', pricing: { @@ -103,8 +101,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { - displayName: 'Mixtral 8x7B Instruct 32k', - enabled: true, + displayName: 'Mixtral 8x7B', functionCall: true, id: 'mixtral-8x7b-32768', pricing: { From ea9ff000a30e4c1cff41ec84e016f7ea265288d3 Mon Sep 17 00:00:00 2001 From: CanisMinor Date: Thu, 12 Sep 2024 14:06:41 +0800 Subject: [PATCH 4/4] =?UTF-8?q?=F0=9F=92=84=20style:=20Add=20model=20and?= =?UTF-8?q?=20provider=20desc=20and=20url=20(#3920)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/config/modelProviders/ai360.ts | 11 ++++ src/config/modelProviders/anthropic.ts | 16 ++++-- src/config/modelProviders/azure.ts | 15 ++++- src/config/modelProviders/baichuan.ts | 4 +- src/config/modelProviders/bedrock.ts | 38 ++++++------ src/config/modelProviders/deepseek.ts | 4 +- src/config/modelProviders/fireworksai.ts | 38 ++++++++++-- src/config/modelProviders/google.ts | 29 +++++----- src/config/modelProviders/groq.ts | 14 +++++ src/config/modelProviders/minimax.ts | 14 +++-- src/config/modelProviders/mistral.ts | 22 ++++++- src/config/modelProviders/moonshot.ts | 12 +++- src/config/modelProviders/novita.ts | 24 ++++++++ src/config/modelProviders/ollama.ts | 59 ++++++++++++++++++- src/config/modelProviders/openai.ts | 70 +++++++++++++++++------ src/config/modelProviders/openrouter.ts | 22 ++++++- src/config/modelProviders/perplexity.ts | 22 ++++++- src/config/modelProviders/qwen.ts | 19 +++--- src/config/modelProviders/siliconcloud.ts | 35 +++++++++++- src/config/modelProviders/spark.ts | 23 +++++--- src/config/modelProviders/stepfun.ts | 14 ++++- src/config/modelProviders/taichu.ts | 9 ++- src/config/modelProviders/togetherai.ts | 40 ++++++++++++- src/config/modelProviders/upstage.ts | 15 +++-- src/config/modelProviders/zeroone.ts | 6 +- src/config/modelProviders/zhipu.ts | 38 ++++++------ src/types/llm.ts | 1 + 27 files changed, 492 insertions(+), 122 deletions(-) diff --git a/src/config/modelProviders/ai360.ts b/src/config/modelProviders/ai360.ts index 1918b4a17fd8..7d40b45b1fa5 100644 --- a/src/config/modelProviders/ai360.ts +++ b/src/config/modelProviders/ai360.ts @@ -1,8 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; +// ref: https://ai.360.cn/platform/docs/overview const Ai360: ModelProviderCard = { chatModels: [ { + description: + '360GPT2 Pro 是 360 公司推出的高级自然语言处理模型，具备卓越的文本生成和理解能力，尤其在生成与创作领域表现出色，能够处理复杂的语言转换和角色演绎任务。', displayName: '360GPT2 Pro', enabled: true, functionCall: false, @@ -16,6 +19,8 @@ const Ai360: ModelProviderCard = { tokens: 8192, }, { + description: + '360GPT Pro 作为 360 AI 模型系列的重要成员，以高效的文本处理能力满足多样化的自然语言应用场景，支持长文本理解和多轮对话等功能。', displayName: '360GPT Pro', id: '360gpt-pro', maxOutput: 7000, @@ -27,6 +32,8 @@ const Ai360: ModelProviderCard = { tokens: 8192, }, { + description: + '360GPT Turbo 提供强大的计算和对话能力，具备出色的语义理解和生成效率，是企业和开发者理想的智能助理解决方案。', displayName: '360GPT Turbo', enabled: true, id: '360gpt-turbo', @@ -39,6 +46,8 @@ const Ai360: ModelProviderCard = { tokens: 8192, }, { + description: + '360GPT Turbo Responsibility 8K 强调语义安全和责任导向，专为对内容安全有高度要求的应用场景设计，确保用户体验的准确性与稳健性。', displayName: '360GPT Turbo Responsibility 8K', enabled: true, id: '360gpt-turbo-responsibility-8k', @@ -52,6 +61,8 @@ const Ai360: ModelProviderCard = { }, ], checkModel: '360gpt-turbo', + description: + '360 AI 是 360 公司推出的 AI 模型和服务平台，提供多种先进的自然语言处理模型，包括 360GPT2 Pro、360GPT Pro、360GPT Turbo 和 360GPT Turbo Responsibility 8K。这些模型结合了大规模参数和多模态能力，广泛应用于文本生成、语义理解、对话系统与代码生成等领域。通过灵活的定价策略，360 AI 满足多样化用户需求，支持开发者集成，推动智能化应用的革新和发展。', disableBrowserRequest: true, id: 'ai360', modelList: { showModelFetcher: true }, diff --git a/src/config/modelProviders/anthropic.ts b/src/config/modelProviders/anthropic.ts index ec2721767a22..53de7c26444d 100644 --- a/src/config/modelProviders/anthropic.ts +++ b/src/config/modelProviders/anthropic.ts @@ -1,10 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; +// ref: https://docs.anthropic.com/en/docs/about-claude/models#model-names const Anthropic: ModelProviderCard = { chatModels: [ { description: - 'Claude 3.5 Sonnet raises the industry bar for intelligence, outperforming competitor models and Claude 3 Opus on a wide range of evaluations, with the speed and cost of our mid-tier model, Claude 3 Sonnet.', + 'Claude 3.5 Sonnet 提供了超越 Opus 的能力和比 Sonnet 更快的速度，同时保持与 Sonnet 相同的价格。Sonnet 特别擅长编程、数据科学、视觉处理、代理任务。', displayName: 'Claude 3.5 Sonnet', enabled: true, functionCall: true, @@ -22,7 +23,7 @@ const Anthropic: ModelProviderCard = { }, { description: - 'Fastest and most compact model for near-instant responsiveness. Quick and accurate targeted performance', + 'Claude 3 Haiku 是 Anthropic 的最快且最紧凑的模型，旨在实现近乎即时的响应。它具有快速且准确的定向性能。', displayName: 'Claude 3 Haiku', enabled: true, functionCall: true, @@ -38,7 +39,7 @@ const Anthropic: ModelProviderCard = { }, { description: - 'Ideal balance of intelligence and speed for enterprise workloads. Maximum utility at a lower price, dependable, balanced for scaled deployments', + 'Claude 3 Sonnet 在智能和速度方面为企业工作负载提供了理想的平衡。它以更低的价格提供最大效用，可靠且适合大规模部署。', displayName: 'Claude 3 Sonnet', enabled: true, functionCall: true, @@ -54,7 +55,7 @@ const Anthropic: ModelProviderCard = { }, { description: - 'Most powerful model for highly complex tasks. Top-level performance, intelligence, fluency, and understanding', + 'Claude 3 Opus 是 Anthropic 用于处理高度复杂任务的最强大模型。它在性能、智能、流畅性和理解力方面表现卓越。', displayName: 'Claude 3 Opus', enabled: true, functionCall: true, @@ -69,6 +70,8 @@ const Anthropic: ModelProviderCard = { vision: true, }, { + description: + 'Claude 2 为企业提供了关键能力的进步，包括业界领先的 200K token 上下文、大幅降低模型幻觉的发生率、系统提示以及一个新的测试功能：工具调用。', displayName: 'Claude 2.1', id: 'claude-2.1', maxOutput: 4096, @@ -80,6 +83,8 @@ const Anthropic: ModelProviderCard = { tokens: 200_000, }, { + description: + 'Claude 2 为企业提供了关键能力的进步，包括业界领先的 200K token 上下文、大幅降低模型幻觉的发生率、系统提示以及一个新的测试功能：工具调用。', displayName: 'Claude 2.0', id: 'claude-2.0', maxOutput: 4096, @@ -91,6 +96,7 @@ const Anthropic: ModelProviderCard = { tokens: 100_000, }, { + description: 'Anthropic 的模型用于低延迟、高吞吐量的文本生成，支持生成数百页的文本。', displayName: 'Claude Instant 1.2', id: 'claude-instant-1.2', maxOutput: 4096, @@ -103,6 +109,8 @@ const Anthropic: ModelProviderCard = { }, ], checkModel: 'claude-3-haiku-20240307', + description: + 'Anthropic 是一家专注于人工智能研究和开发的公司，提供了一系列先进的语言模型，如 Claude 3.5 Sonnet、Claude 3 Sonnet、Claude 3 Opus 和 Claude 3 Haiku。这些模型在智能、速度和成本之间取得了理想的平衡，适用于从企业级工作负载到快速响应的各种应用场景。Claude 3.5 Sonnet 作为其最新模型，在多项评估中表现优异，同时保持了较高的性价比。', id: 'anthropic', modelsUrl: 'https://docs.anthropic.com/en/docs/about-claude/models#model-names', name: 'Anthropic', diff --git a/src/config/modelProviders/azure.ts b/src/config/modelProviders/azure.ts index 69489c0520be..7b28029198f2 100644 --- a/src/config/modelProviders/azure.ts +++ b/src/config/modelProviders/azure.ts @@ -1,11 +1,12 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models +// ref: https://learn.microsoft.com/azure/ai-services/openai/concepts/models const Azure: ModelProviderCard = { chatModels: [ { deploymentName: 'gpt-35-turbo', - description: 'GPT 3.5 Turbo，适用于各种文本生成和理解任务', + description: + 'GPT 3.5 Turbo，OpenAI提供的高效模型，适用于聊天和文本生成任务，支持并行函数调用。', displayName: 'GPT 3.5 Turbo', enabled: true, functionCall: true, @@ -15,6 +16,7 @@ const Azure: ModelProviderCard = { }, { deploymentName: 'gpt-35-turbo-16k', + description: 'GPT 3.5 Turbo 16k，高容量文本生成模型，适合复杂任务。', displayName: 'GPT 3.5 Turbo', functionCall: true, id: 'gpt-35-turbo-16k', @@ -22,6 +24,7 @@ const Azure: ModelProviderCard = { }, { deploymentName: 'gpt-4-turbo', + description: 'GPT 4 Turbo，多模态模型，提供杰出的语言理解和生成能力，同时支持图像输入。', displayName: 'GPT 4 Turbo', enabled: true, functionCall: true, @@ -31,7 +34,7 @@ const Azure: ModelProviderCard = { }, { deploymentName: 'gpt-4-vision', - description: 'GPT-4 视觉预览版，支持视觉任务', + description: 'GPT-4 视觉预览版，专为图像分析和处理任务设计。', displayName: 'GPT 4 Turbo with Vision Preview', id: 'gpt-4-vision-preview', tokens: 128_000, @@ -39,6 +42,7 @@ const Azure: ModelProviderCard = { }, { deploymentName: 'gpt-4o-mini', + description: 'GPT-4o Mini，小型高效模型，具备与GPT-4o相似的卓越性能。', displayName: 'GPT 4o Mini', enabled: true, functionCall: true, @@ -48,6 +52,7 @@ const Azure: ModelProviderCard = { }, { deploymentName: 'gpt-4o', + description: 'GPT-4o 是最新的多模态模型，结合高级文本和图像处理能力。', displayName: 'GPT 4o', enabled: true, functionCall: true, @@ -57,8 +62,12 @@ const Azure: ModelProviderCard = { }, ], defaultShowBrowserRequest: true, + description: + 'Azure 提供多种先进的AI模型，包括GPT-3.5和最新的GPT-4系列，支持多种数据类型和复杂任务，致力于安全、可靠和可持续的AI解决方案。', id: 'azure', + modelsUrl: 'https://learn.microsoft.com/azure/ai-services/openai/concepts/models', name: 'Azure', + url: 'https://azure.microsoft.com', }; export default Azure; diff --git a/src/config/modelProviders/baichuan.ts b/src/config/modelProviders/baichuan.ts index a148e135661c..c7fdc2cc912e 100644 --- a/src/config/modelProviders/baichuan.ts +++ b/src/config/modelProviders/baichuan.ts @@ -1,6 +1,6 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://platform.baichuan-ai.com/price +// ref: https://platform.baichuan-ai.com/price const Baichuan: ModelProviderCard = { chatModels: [ { @@ -62,6 +62,8 @@ const Baichuan: ModelProviderCard = { }, ], checkModel: 'Baichuan3-Turbo', + description: + '百川智能是一家专注于人工智能大模型研发的公司，其模型在国内知识百科、长文本处理和生成创作等中文任务上表现卓越，超越了国外主流模型。百川智能还具备行业领先的多模态能力，在多项权威评测中表现优异。其模型包括 Baichuan 4、Baichuan 3 Turbo 和 Baichuan 3 Turbo 128k 等，分别针对不同应用场景进行优化，提供高性价比的解决方案。', id: 'baichuan', modelList: { showModelFetcher: true }, modelsUrl: 'https://platform.baichuan-ai.com/price', diff --git a/src/config/modelProviders/bedrock.ts b/src/config/modelProviders/bedrock.ts index ec7affe02041..892df8d143b7 100644 --- a/src/config/modelProviders/bedrock.ts +++ b/src/config/modelProviders/bedrock.ts @@ -1,26 +1,26 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html -// ref https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/models -// ref https://us-west-2.console.aws.amazon.com/bedrock/home?region=us-west-2#/models +// ref :https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html +// ref :https://us-east-1.console.aws.amazon.com/bedrock/home?region=us-east-1#/models +// ref :https://us-west-2.console.aws.amazon.com/bedrock/home?region=us-west-2#/models const Bedrock: ModelProviderCard = { chatModels: [ /* // TODO: Not support for now { - description: 'Amazon Titan Text Lite is a light weight efficient model ideal for fine-tuning for English-language tasks, including like summarization and copywriting, where customers want a smaller, more cost-effective model that is also highly customizable.', + description: '亚马逊 Titan Text Lite 是一款轻量级高效模型，非常适合对英语任务进行微调，包括总结和文案编写等，客户希望有一个更小、更经济的模型，同时也非常可定制。', displayName: 'Titan Text G1 - Lite', id: 'amazon.titan-text-lite-v1', tokens: 4000, }, { - description: 'Amazon Titan Text Express has a context length of up to 8,000 tokens, making it well-suited for a wide range of advanced, general language tasks such as open-ended text generation and conversational chat, as well as support within Retrieval Augmented Generation (RAG). At launch, the model is optimized for English, with multilingual support for more than 100 additional languages available in preview.', + description: '亚马逊 Titan Text Express 的上下文长度可达 8,000 个标记，非常适合广泛的高级通用语言任务，如开放式文本生成和对话聊天，以及在检索增强生成 (RAG) 中的支持。在推出时，该模型针对英语进行了优化，预览版还支持其他 100 多种语言。', displayName: 'Titan Text G1 - Express', id: 'amazon.titan-text-express-v1', tokens: 8000, }, { - description: 'Titan Text Premier is a powerful and advanced model within the Titan Text family, designed to deliver superior performance across a wide range of enterprise applications. With its cutting-edge capabilities, it offers enhanced accuracy and exceptional results, making it an excellent choice for organizations seeking top-notch text processing solutions.', + description: 'Titan Text Premier 是 Titan Text 系列中一款强大的先进模型，旨在为广泛的企业应用提供卓越的性能。凭借其尖端能力，它提供了更高的准确性和卓越的结果，是寻求一流文本处理解决方案的组织的绝佳选择。', displayName: 'Titan Text G1 - Premier', id: 'amazon.titan-text-premier-v1:0', tokens: 32_000, @@ -28,7 +28,7 @@ const Bedrock: ModelProviderCard = { */ { description: - 'Claude 3.5 Sonnet raises the industry bar for intelligence, outperforming competitor models and Claude 3 Opus on a wide range of evaluations, with the speed and cost of our mid-tier model, Claude 3 Sonnet.', + 'Claude 3.5 Sonnet 提升了行业标准，性能超过竞争对手模型和 Claude 3 Opus，在广泛的评估中表现出色，同时具有我们中等层级模型的速度和成本。', displayName: 'Claude 3.5 Sonnet', enabled: true, functionCall: true, @@ -42,7 +42,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Claude 3 Sonnet by Anthropic strikes the ideal balance between intelligence and speed—particularly for enterprise workloads. It offers maximum utility at a lower price than competitors, and is engineered to be the dependable, high-endurance workhorse for scaled AI deployments. Claude 3 Sonnet can process images and return text outputs, and features a 200K context window.', + 'Anthropic 的 Claude 3 Sonnet 在智能和速度之间达到了理想的平衡——特别适合企业工作负载。它以低于竞争对手的价格提供最大的效用，并被设计成为可靠的、高耐用的主力机，适用于规模化的 AI 部署。Claude 3 Sonnet 可以处理图像并返回文本输出，具有 200K 的上下文窗口。', displayName: 'Claude 3 Sonnet', enabled: true, functionCall: true, @@ -56,7 +56,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Claude 3 Opus is Anthropic most powerful AI model, with state-of-the-art performance on highly complex tasks. It can navigate open-ended prompts and sight-unseen scenarios with remarkable fluency and human-like understanding. Claude 3 Opus shows us the frontier of what’s possible with generative AI. Claude 3 Opus can process images and return text outputs, and features a 200K context window.', + 'Claude 3 Opus 是 Anthropic 最强大的 AI 模型，具有在高度复杂任务上的最先进性能。它可以处理开放式提示和未见过的场景，具有出色的流畅性和类人的理解能力。Claude 3 Opus 展示了生成 AI 可能性的前沿。Claude 3 Opus 可以处理图像并返回文本输出，具有 200K 的上下文窗口。', displayName: 'Claude 3 Opus', enabled: true, functionCall: true, @@ -70,7 +70,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Claude 3 Haiku is Anthropic fastest, most compact model for near-instant responsiveness. It answers simple queries and requests with speed. Customers will be able to build seamless AI experiences that mimic human interactions. Claude 3 Haiku can process images and return text outputs, and features a 200K context window.', + 'Claude 3 Haiku 是 Anthropic 最快、最紧凑的模型，提供近乎即时的响应速度。它可以快速回答简单的查询和请求。客户将能够构建模仿人类互动的无缝 AI 体验。Claude 3 Haiku 可以处理图像并返回文本输出，具有 200K 的上下文窗口。', displayName: 'Claude 3 Haiku', enabled: true, functionCall: true, @@ -84,7 +84,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'An update to Claude 2 that features double the context window, plus improvements across reliability, hallucination rates, and evidence-based accuracy in long document and RAG contexts.', + 'Claude 2 的更新版，具有双倍的上下文窗口，以及在长文档和 RAG 上下文中的可靠性、幻觉率和基于证据的准确性的改进。', displayName: 'Claude 2.1', id: 'anthropic.claude-v2:1', pricing: { @@ -95,7 +95,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Anthropic highly capable model across a wide range of tasks from sophisticated dialogue and creative content generation to detailed instruction following.', + 'Anthropic 在从复杂对话和创意内容生成到详细指令跟随的广泛任务中都表现出高度能力的模型。', displayName: 'Claude 2.0', id: 'anthropic.claude-v2', pricing: { @@ -106,7 +106,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'A fast, affordable yet still very capable model, which can handle a range of tasks including casual dialogue, text analysis, summarization, and document question-answering.', + '一款快速、经济且仍然非常有能力的模型，可以处理包括日常对话、文本分析、总结和文档问答在内的一系列任务。', displayName: 'Claude Instant', id: 'anthropic.claude-instant-v1', pricing: { @@ -117,7 +117,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'An update to Meta Llama 3 8B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + 'Meta Llama 3.1 8B Instruct 的更新版，包括扩展的 128K 上下文长度、多语言性和改进的推理能力。Llama 3.1 提供的多语言大型语言模型 (LLMs) 是一组预训练的、指令调整的生成模型，包括 8B、70B 和 405B 大小 (文本输入/输出)。Llama 3.1 指令调整的文本模型 (8B、70B、405B) 专为多语言对话用例进行了优化，并在常见的行业基准测试中超过了许多可用的开源聊天模型。Llama 3.1 旨在用于多种语言的商业和研究用途。指令调整的文本模型适用于类似助手的聊天，而预训练模型可以适应各种自然语言生成任务。Llama 3.1 模型还支持利用其模型的输出来改进其他模型，包括合成数据生成和精炼。Llama 3.1 是使用优化的变压器架构的自回归语言模型。调整版本使用监督微调 (SFT) 和带有人类反馈的强化学习 (RLHF) 来符合人类对帮助性和安全性的偏好。', displayName: 'Llama 3.1 8B Instruct', enabled: true, functionCall: true, @@ -130,7 +130,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'An update to Meta Llama 3 70B Instruct that includes an expanded 128K context length, multilinguality and improved reasoning capabilities. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + 'Meta Llama 3.1 70B Instruct 的更新版，包括扩展的 128K 上下文长度、多语言性和改进的推理能力。Llama 3.1 提供的多语言大型语言模型 (LLMs) 是一组预训练的、指令调整的生成模型，包括 8B、70B 和 405B 大小 (文本输入/输出)。Llama 3.1 指令调整的文本模型 (8B、70B、405B) 专为多语言对话用例进行了优化，并在常见的行业基准测试中超过了许多可用的开源聊天模型。Llama 3.1 旨在用于多种语言的商业和研究用途。指令调整的文本模型适用于类似助手的聊天，而预训练模型可以适应各种自然语言生成任务。Llama 3.1 模型还支持利用其模型的输出来改进其他模型，包括合成数据生成和精炼。Llama 3.1 是使用优化的变压器架构的自回归语言模型。调整版本使用监督微调 (SFT) 和带有人类反馈的强化学习 (RLHF) 来符合人类对帮助性和安全性的偏好。', displayName: 'Llama 3.1 70B Instruct', enabled: true, functionCall: true, @@ -143,7 +143,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Meta Llama 3.1 405B Instruct is the largest and most powerful of the Llama 3.1 Instruct models that is a highly advanced model for conversational inference and reasoning, synthetic data generation, and a base to do specialized continual pre-training or fine-tuning on a specific domain. The Llama 3.1 offering of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 8B, 70B and 405B sizes (text in/text out). The Llama 3.1 instruction-tuned text only models (8B, 70B, 405B) are optimized for multilingual dialogue use cases and outperform many of the available open source chat models on common industry benchmarks. Llama 3.1 is intended for commercial and research use in multiple languages. Instruction tuned text only models are intended for assistant-like chat, whereas pretrained models can be adapted for a variety of natural language generation tasks. The Llama 3.1 models also support the ability to leverage the outputs of its models to improve other models including synthetic data generation and distillation. Llama 3.1 is an auto-regressive language model that uses an optimized transformer architecture. The tuned versions use supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF) to align with human preferences for helpfulness and safety.', + 'Meta Llama 3.1 405B Instruct 是 Llama 3.1 Instruct 模型中最大、最强大的模型，是一款高度先进的对话推理和合成数据生成模型，也可以用作在特定领域进行专业持续预训练或微调的基础。Llama 3.1 提供的多语言大型语言模型 (LLMs) 是一组预训练的、指令调整的生成模型，包括 8B、70B 和 405B 大小 (文本输入/输出)。Llama 3.1 指令调整的文本模型 (8B、70B、405B) 专为多语言对话用例进行了优化，并在常见的行业基准测试中超过了许多可用的开源聊天模型。Llama 3.1 旨在用于多种语言的商业和研究用途。指令调整的文本模型适用于类似助手的聊天，而预训练模型可以适应各种自然语言生成任务。Llama 3.1 模型还支持利用其模型的输出来改进其他模型，包括合成数据生成和精炼。Llama 3.1 是使用优化的变压器架构的自回归语言模型。调整版本使用监督微调 (SFT) 和带有人类反馈的强化学习 (RLHF) 来符合人类对帮助性和安全性的偏好。', displayName: 'Llama 3.1 405B Instruct', enabled: true, functionCall: true, @@ -156,7 +156,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for limited computational power and resources, edge devices, and faster training times.', + 'Meta Llama 3 是一款面向开发者、研究人员和企业的开放大型语言模型 (LLM)，旨在帮助他们构建、实验并负责任地扩展他们的生成 AI 想法。作为全球社区创新的基础系统的一部分，它非常适合计算能力和资源有限、边缘设备和更快的训练时间。', displayName: 'Llama 3 8B Instruct', id: 'meta.llama3-8b-instruct-v1:0', pricing: { @@ -167,7 +167,7 @@ const Bedrock: ModelProviderCard = { }, { description: - 'Meta Llama 3 is an accessible, open large language model (LLM) designed for developers, researchers, and businesses to build, experiment, and responsibly scale their generative AI ideas. Part of a foundational system, it serves as a bedrock for innovation in the global community. Ideal for content creation, conversational AI, language understanding, R&D, and Enterprise applications.', + 'Meta Llama 3 是一款面向开发者、研究人员和企业的开放大型语言模型 (LLM)，旨在帮助他们构建、实验并负责任地扩展他们的生成 AI 想法。作为全球社区创新的基础系统的一部分，它非常适合内容创建、对话 AI、语言理解、研发和企业应用。', displayName: 'Llama 3 70B Instruct', id: 'meta.llama3-70b-instruct-v1:0', pricing: { @@ -276,6 +276,8 @@ const Bedrock: ModelProviderCard = { */ ], checkModel: 'anthropic.claude-instant-v1', + description: + 'Bedrock 是亚马逊 AWS 提供的一项服务，专注于为企业提供先进的 AI 语言模型和视觉模型。其模型家族包括 Anthropic 的 Claude 系列、Meta 的 Llama 3.1 系列等，涵盖从轻量级到高性能的多种选择，支持文本生成、对话、图像处理等多种任务，适用于不同规模和需求的企业应用。', id: 'bedrock', modelsUrl: 'https://docs.aws.amazon.com/bedrock/latest/userguide/model-ids.html', name: 'Bedrock', diff --git a/src/config/modelProviders/deepseek.ts b/src/config/modelProviders/deepseek.ts index 842b1ca58d03..ec66bc456a3b 100644 --- a/src/config/modelProviders/deepseek.ts +++ b/src/config/modelProviders/deepseek.ts @@ -1,6 +1,6 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://platform.deepseek.com/api-docs/pricing +// ref: https://platform.deepseek.com/api-docs/pricing const DeepSeek: ModelProviderCard = { chatModels: [ { @@ -20,6 +20,8 @@ const DeepSeek: ModelProviderCard = { }, ], checkModel: 'deepseek-chat', + description: + 'DeepSeek 是一家专注于人工智能技术研究和应用的公司，其最新模型 DeepSeek-V2.5 融合了通用对话和代码处理能力，并在人类偏好对齐、写作任务和指令跟随等方面实现了显著提升。', id: 'deepseek', modelList: { showModelFetcher: true }, modelsUrl: 'https://platform.deepseek.com/api-docs/zh-cn/quick_start/pricing', diff --git a/src/config/modelProviders/fireworksai.ts b/src/config/modelProviders/fireworksai.ts index 5e544822525c..0c5d94242189 100644 --- a/src/config/modelProviders/fireworksai.ts +++ b/src/config/modelProviders/fireworksai.ts @@ -1,12 +1,12 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://fireworks.ai/models?show=Serverless -// ref https://fireworks.ai/pricing +// ref: https://fireworks.ai/models?show=Serverless +// ref: https://fireworks.ai/pricing const FireworksAI: ModelProviderCard = { chatModels: [ { description: - 'Fireworks latest and most performant function-calling model. Firefunction-v2 is based on Llama-3 and trained to excel at function-calling as well as chat and instruction-following. See blog post for more details https://fireworks.ai/blog/firefunction-v2-launch-post', + 'Fireworks 公司最新推出的 Firefunction-v2 是一款性能卓越的函数调用模型，基于 Llama-3 开发，并通过大量优化，特别适用于函数调用、对话及指令跟随等场景。', displayName: 'Firefunction V2', enabled: true, functionCall: true, @@ -14,7 +14,7 @@ const FireworksAI: ModelProviderCard = { tokens: 8192, }, { - description: 'Fireworks open-source function calling model.', + description: 'Fireworks 开源函数调用模型，提供卓越的指令执行能力和开放可定制的特性。', displayName: 'Firefunction V1', functionCall: true, id: 'accounts/fireworks/models/firefunction-v1', @@ -22,7 +22,7 @@ const FireworksAI: ModelProviderCard = { }, { description: - 'Vision-language model allowing both image and text as inputs (single image is recommended), trained on OSS model generated training data and open sourced on huggingface at fireworks-ai/FireLLaVA-13b', + 'fireworks-ai/FireLLaVA-13b 是一款视觉语言模型，可以同时接收图像和文本输入，经过高质量数据训练，适合多模态任务。', displayName: 'FireLLaVA-13B', enabled: true, functionCall: false, @@ -31,6 +31,8 @@ const FireworksAI: ModelProviderCard = { vision: true, }, { + description: + 'Llama 3.1 8B 指令模型，专为多语言对话优化，能够在常见行业基准上超越多数开源及闭源模型。', displayName: 'Llama 3.1 8B Instruct', enabled: true, functionCall: false, @@ -38,6 +40,8 @@ const FireworksAI: ModelProviderCard = { tokens: 131_072, }, { + description: + 'Llama 3.1 70B 指令模型，提供卓越的自然语言理解和生成能力，是对话及分析任务的理想选择。', displayName: 'Llama 3.1 70B Instruct', enabled: true, functionCall: false, @@ -45,6 +49,8 @@ const FireworksAI: ModelProviderCard = { tokens: 131_072, }, { + description: + 'Llama 3.1 405B 指令模型，具备超大规模参数，适合复杂任务和高负载场景下的指令跟随。', displayName: 'Llama 3.1 405B Instruct', enabled: true, functionCall: false, @@ -52,30 +58,38 @@ const FireworksAI: ModelProviderCard = { tokens: 131_072, }, { + description: 'Llama 3 8B 指令模型，优化用于对话及多语言任务，表现卓越且高效。', displayName: 'Llama 3 8B Instruct', functionCall: false, id: 'accounts/fireworks/models/llama-v3-8b-instruct', tokens: 8192, }, { + description: 'Llama 3 70B 指令模型，专为多语言对话和自然语言理解优化，性能优于多数竞争模型。', displayName: 'Llama 3 70B Instruct', functionCall: false, id: 'accounts/fireworks/models/llama-v3-70b-instruct', tokens: 8192, }, { + description: + 'Llama 3 8B 指令模型（HF 版本），与官方实现结果一致，具备高度一致性和跨平台兼容性。', displayName: 'Llama 3 8B Instruct (HF version)', functionCall: false, id: 'accounts/fireworks/models/llama-v3-8b-instruct-hf', tokens: 8192, }, { + description: + 'Llama 3 70B 指令模型（HF 版本），与官方实现结果保持一致，适合高质量的指令跟随任务。', displayName: 'Llama 3 70B Instruct (HF version)', functionCall: false, id: 'accounts/fireworks/models/llama-v3-70b-instruct-hf', tokens: 8192, }, { + description: + 'Gemma 2 9B 指令模型，基于之前的Google技术，适合回答问题、总结和推理等多种文本生成任务。', displayName: 'Gemma 2 9B Instruct', enabled: true, functionCall: false, @@ -83,6 +97,7 @@ const FireworksAI: ModelProviderCard = { tokens: 8192, }, { + description: 'Mixtral MoE 8x7B 指令模型，多专家架构提供高效的指令跟随及执行。', displayName: 'Mixtral MoE 8x7B Instruct', enabled: true, functionCall: false, @@ -90,6 +105,8 @@ const FireworksAI: ModelProviderCard = { tokens: 32_768, }, { + description: + 'Mixtral MoE 8x22B 指令模型，大规模参数和多专家架构，全方位支持复杂任务的高效处理。', displayName: 'Mixtral MoE 8x22B Instruct', enabled: true, functionCall: false, @@ -97,12 +114,16 @@ const FireworksAI: ModelProviderCard = { tokens: 65_536, }, { + description: + 'Mixtral MoE 8x7B 指令模型（HF 版本），性能与官方实现一致，适合多种高效任务场景。', displayName: 'Mixtral MoE 8x7B Instruct (HF version)', functionCall: false, id: 'accounts/fireworks/models/mixtral-8x7b-instruct-hf', tokens: 32_768, }, { + description: + 'Phi 3 Vision 指令模型，轻量级多模态模型，能够处理复杂的视觉和文本信息，具备较强的推理能力。', displayName: 'Phi 3 Vision Instruct', enabled: true, functionCall: false, @@ -111,6 +132,7 @@ const FireworksAI: ModelProviderCard = { vision: true, }, { + description: 'Yi-Large 模型，具备卓越的多语言处理能力，可用于各类语言生成和理解任务。', displayName: 'Yi-Large', enabled: true, functionCall: false, @@ -118,18 +140,22 @@ const FireworksAI: ModelProviderCard = { tokens: 32_768, }, { + description: 'StarCoder 7B 模型，针对80多种编程语言训练，拥有出色的编程填充能力和语境理解。', displayName: 'StarCoder 7B', functionCall: false, id: 'accounts/fireworks/models/starcoder-7b', tokens: 8192, }, { + description: + 'StarCoder 15.5B 模型，支持高级编程任务，多语言能力增强，适合复杂代码生成和理解。', displayName: 'StarCoder 15.5B', functionCall: false, id: 'accounts/fireworks/models/starcoder-16b', tokens: 8192, }, { + description: 'MythoMax L2 13B 模型，结合新颖的合并技术，擅长叙事和角色扮演。', displayName: 'MythoMax L2 13b', functionCall: false, id: 'accounts/fireworks/models/mythomax-l2-13b', @@ -137,6 +163,8 @@ const FireworksAI: ModelProviderCard = { }, ], checkModel: 'accounts/fireworks/models/firefunction-v2', + description: + 'Fireworks AI 是一家领先的高级语言模型服务商，专注于功能调用和多模态处理。其最新模型 Firefunction V2 基于 Llama-3，优化用于函数调用、对话及指令跟随。视觉语言模型 FireLLaVA-13B 支持图像和文本混合输入。其他 notable 模型包括 Llama 系列和 Mixtral 系列，提供高效的多语言指令跟随与生成支持。', id: 'fireworksai', modelList: { showModelFetcher: true }, modelsUrl: 'https://fireworks.ai/models?show=Serverless', diff --git a/src/config/modelProviders/google.ts b/src/config/modelProviders/google.ts index 45cea49fc86a..babb9e83cacc 100644 --- a/src/config/modelProviders/google.ts +++ b/src/config/modelProviders/google.ts @@ -1,9 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; +// ref: https://ai.google.dev/gemini-api/docs/models/gemini const Google: ModelProviderCard = { chatModels: [ { - description: 'Fast and versatile multimodal model for scaling across diverse tasks', + description: + 'Gemini 1.5 Flash 是Google最新的多模态AI模型，具备快速处理能力，支持文本、图像和视频输入，适用于多种任务的高效扩展。', displayName: 'Gemini 1.5 Flash', enabled: true, functionCall: true, @@ -18,7 +20,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Fast and versatile multimodal model for scaling across diverse tasks', + description: 'Gemini 1.5 Flash 0827 提供了优化后的多模态处理能力，适用多种复杂任务场景。', displayName: 'Gemini 1.5 Flash 0827', enabled: true, functionCall: true, @@ -34,7 +36,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Fast and versatile multimodal model for scaling across diverse tasks', + description: 'Gemini 1.5 Flash 8B 0827 专为处理大规模任务场景设计，提供无与伦比的处理速度。', displayName: 'Gemini 1.5 Flash 8B 0827', functionCall: true, id: 'gemini-1.5-flash-8b-exp-0827', @@ -44,7 +46,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Fast and versatile multimodal model for scaling across diverse tasks', + description: 'Gemini 1.5 Flash 001 是一款高效的多模态模型，支持广泛应用的扩展。', displayName: 'Gemini 1.5 Flash 001', functionCall: true, id: 'gemini-1.5-flash-001', @@ -53,7 +55,8 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Mid-size multimodal model that supports up to 2 million tokens', + description: + 'Gemini 1.5 Pro 支持高达200万个tokens，是中型多模态模型的理想选择，适用于复杂任务的多方面支持。', displayName: 'Gemini 1.5 Pro', enabled: true, functionCall: true, @@ -69,7 +72,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Mid-size multimodal model that supports up to 2 million tokens', + description: 'Gemini 1.5 Pro 0827 结合最新优化技术，带来更高效的多模态数据处理能力。', displayName: 'Gemini 1.5 Pro 0827', enabled: true, functionCall: true, @@ -85,7 +88,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Mid-size multimodal model that supports up to 2 million tokens', + description: 'Gemini 1.5 Pro 0801 提供出色的多模态处理能力，为应用开发带来更大灵活性。', displayName: 'Gemini 1.5 Pro 0801', functionCall: true, id: 'gemini-1.5-pro-exp-0801', @@ -100,7 +103,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: 'Mid-size multimodal model that supports up to 2 million tokens', + description: 'Gemini 1.5 Pro 001 是可扩展的多模态AI解决方案，支持广泛的复杂任务。', displayName: 'Gemini 1.5 Pro 001', functionCall: true, id: 'gemini-1.5-pro-001', @@ -115,8 +118,7 @@ const Google: ModelProviderCard = { vision: true, }, { - description: - 'The best model for scaling across a wide range of tasks. This is the latest model.', + description: 'Gemini 1.0 Pro 是Google的高性能AI模型，专为广泛任务扩展而设计。', displayName: 'Gemini 1.0 Pro', id: 'gemini-1.0-pro-latest', maxOutput: 2048, @@ -129,7 +131,7 @@ const Google: ModelProviderCard = { }, { description: - 'The best model for scaling across a wide range of tasks. This is a stable model that supports tuning.', + 'Gemini 1.0 Pro 001 (Tuning) 提供稳定并可调优的性能，是复杂任务解决方案的理想选择。', displayName: 'Gemini 1.0 Pro 001 (Tuning)', functionCall: true, id: 'gemini-1.0-pro-001', @@ -142,8 +144,7 @@ const Google: ModelProviderCard = { tokens: 30_720 + 2048, }, { - description: - 'The best model for scaling across a wide range of tasks. Released April 9, 2024.', + description: 'Gemini 1.0 Pro 002 (Tuning) 提供出色的多模态支持，专注于复杂任务的有效解决。', displayName: 'Gemini 1.0 Pro 002 (Tuning)', id: 'gemini-1.0-pro-002', maxOutput: 2048, @@ -156,6 +157,8 @@ const Google: ModelProviderCard = { }, ], checkModel: 'gemini-1.5-flash-latest', + description: + 'Google 的 Gemini 系列是其最先进、通用的 A I模型，由 Google DeepMind 打造，专为多模态设计，支持文本、代码、图像、音频和视频的无缝理解与处理。适用于从数据中心到移动设备的多种环境，极大提升了AI模型的效率与应用广泛性。', id: 'google', modelsUrl: 'https://ai.google.dev/gemini-api/docs/models/gemini', name: 'Google', diff --git a/src/config/modelProviders/groq.ts b/src/config/modelProviders/groq.ts index eee858f13663..ec4a11b7e40c 100644 --- a/src/config/modelProviders/groq.ts +++ b/src/config/modelProviders/groq.ts @@ -5,6 +5,8 @@ const Groq: ModelProviderCard = { chatModels: [ // TODO: During preview launch, Groq is limiting 3.1 models to max_tokens of 8k. { + description: + 'Llama 3.1 8B 是一款高效能模型，提供了快速的文本生成能力，非常适合需要大规模效率和成本效益的应用场景。', displayName: 'Llama 3.1 8B (Preview)', enabled: true, functionCall: true, @@ -17,6 +19,8 @@ const Groq: ModelProviderCard = { tokens: 131_072, }, { + description: + 'Llama 3.1 70B 提供更强大的AI推理能力，适合复杂应用，支持超多的计算处理并保证高效和准确率。', displayName: 'Llama 3.1 70B (Preview)', enabled: true, functionCall: true, @@ -38,6 +42,7 @@ const Groq: ModelProviderCard = { }, */ { + description: 'Llama 3 Groq 8B Tool Use 是针对高效工具使用优化的模型，支持快速并行计算。', displayName: 'Llama 3 Groq 8B Tool Use (Preview)', enabled: true, functionCall: true, @@ -49,6 +54,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Llama 3 Groq 70B Tool Use 提供强大的工具调用能力，支持复杂任务的高效处理。', displayName: 'Llama 3 Groq 70B Tool Use (Preview)', enabled: true, functionCall: true, @@ -60,6 +66,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Meta Llama 3 8B 带来优质的推理效能，适合多场景应用需求。', displayName: 'Meta Llama 3 8B', functionCall: true, id: 'llama3-8b-8192', @@ -70,6 +77,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Meta Llama 3 70B 提供无与伦比的复杂性处理能力，为高要求项目量身定制。', displayName: 'Meta Llama 3 70B', functionCall: true, id: 'llama3-70b-8192', @@ -80,6 +88,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Gemma 2 9B 是一款优化用于特定任务和工具整合的模型。', displayName: 'Gemma 2 9B', enabled: true, functionCall: true, @@ -91,6 +100,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Gemma 7B 适合中小规模任务处理，兼具成本效益。', displayName: 'Gemma 7B', functionCall: true, id: 'gemma-7b-it', @@ -101,6 +111,7 @@ const Groq: ModelProviderCard = { tokens: 8192, }, { + description: 'Mixtral 8x7B 提供高容错的并行计算能力，适合复杂任务。', displayName: 'Mixtral 8x7B', functionCall: true, id: 'mixtral-8x7b-32768', @@ -111,6 +122,7 @@ const Groq: ModelProviderCard = { tokens: 32_768, }, { + description: 'LLaVA 1.5 7B 提供视觉处理能力融合，通过视觉信息输入生成复杂输出。', displayName: 'LLaVA 1.5 7B', enabled: true, id: 'llava-v1.5-7b-4096-preview', @@ -119,6 +131,8 @@ const Groq: ModelProviderCard = { }, ], checkModel: 'gemma2-9b-it', + description: + 'Groq 的 LPU 推理引擎在最新的独立大语言模型（LLM）基准测试中表现卓越，以其惊人的速度和效率重新定义了 AI 解决方案的标准。Groq 是一种即时推理速度的代表，在基于云的部署中展现了良好的性能。', id: 'groq', modelsUrl: 'https://console.groq.com/docs/models', name: 'Groq', diff --git a/src/config/modelProviders/minimax.ts b/src/config/modelProviders/minimax.ts index 6f114679bbc9..4b8ac4da6694 100644 --- a/src/config/modelProviders/minimax.ts +++ b/src/config/modelProviders/minimax.ts @@ -1,10 +1,10 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://www.minimaxi.com/document/guides/chat-model/pro/api +// ref: https://platform.minimaxi.com/document/Models const Minimax: ModelProviderCard = { chatModels: [ { - description: '通用场景', + 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。', displayName: 'abab6.5s', enabled: true, functionCall: true, @@ -12,7 +12,7 @@ const Minimax: ModelProviderCard = { tokens: 245_760, }, { - description: '英文等多语种人设对话场景', + 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。', displayName: 'abab6.5g', enabled: true, functionCall: true, @@ -20,7 +20,7 @@ const Minimax: ModelProviderCard = { tokens: 8192, }, { - description: '中文人设对话场景', + 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。', displayName: 'abab6.5t', enabled: true, functionCall: true, @@ -28,19 +28,21 @@ const Minimax: ModelProviderCard = { tokens: 8192, }, { - description: '生产力场景', + 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。', displayName: 'abab5.5', id: 'abab5.5-chat', tokens: 16_384, }, { - description: '中文人设对话场景', + 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。', displayName: 'abab5.5s', id: 'abab5.5s-chat', tokens: 8192, }, ], checkModel: 'abab6.5s-chat', + description: + 'MiniMax 是 2021 年成立的通用人工智能科技公司，致力于与用户共创智能。MiniMax 自主研发了不同模态的通用大模型，其中包括万亿参数的 MoE 文本大模型、语音大模型以及图像大模型。并推出了海螺 AI 等应用。', id: 'minimax', modelsUrl: 'https://platform.minimaxi.com/document/Models', name: 'Minimax', diff --git a/src/config/modelProviders/mistral.ts b/src/config/modelProviders/mistral.ts index 184e2a1ee562..2f772ff90796 100644 --- a/src/config/modelProviders/mistral.ts +++ b/src/config/modelProviders/mistral.ts @@ -1,29 +1,36 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.mistral.ai/getting-started/models/ -// ref https://docs.mistral.ai/capabilities/function_calling/ +// ref: https://docs.mistral.ai/getting-started/models/ +// ref: https://docs.mistral.ai/capabilities/function_calling/ const Mistral: ModelProviderCard = { chatModels: [ { + description: + '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。', displayName: 'Mistral 7B', enabled: true, id: 'open-mistral-7b', tokens: 32_768, }, { + description: + '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。', displayName: 'Mixtral 8x7B', enabled: true, id: 'open-mixtral-8x7b', tokens: 32_768, }, { + description: + '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。', displayName: 'Mixtral 8x22B', - enabled: true, functionCall: true, id: 'open-mixtral-8x22b', tokens: 65_536, }, { + description: + '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。', displayName: 'Mistral Nemo', enabled: true, functionCall: true, @@ -31,6 +38,8 @@ const Mistral: ModelProviderCard = { tokens: 128_000, }, { + description: + '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。', displayName: 'Mistral Large', enabled: true, functionCall: true, @@ -38,12 +47,15 @@ const Mistral: ModelProviderCard = { tokens: 128_000, }, { + 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。', displayName: 'Codestral', enabled: true, id: 'codestral-latest', tokens: 32_768, }, { + description: + '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。', displayName: 'Codestral Mamba', enabled: true, id: 'open-codestral-mamba', @@ -51,8 +63,12 @@ const Mistral: ModelProviderCard = { }, ], checkModel: 'open-mistral-7b', + description: + 'Mistral 提供先进的通用、专业和研究型模型，广泛应用于复杂推理、多语言任务、代码生成等领域，通过功能调用接口，用户可以集成自定义功能，实现特定应用。', id: 'mistral', + modelsUrl: 'https://docs.mistral.ai/getting-started/models', name: 'Mistral', + url: 'https://mistral.ai', }; export default Mistral; diff --git a/src/config/modelProviders/moonshot.ts b/src/config/modelProviders/moonshot.ts index e1ed87ff926f..e72cb554d69c 100644 --- a/src/config/modelProviders/moonshot.ts +++ b/src/config/modelProviders/moonshot.ts @@ -1,9 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://platform.moonshot.cn/docs/intro#模型列表 +// ref: https://platform.moonshot.cn/docs/intro#模型列表 const Moonshot: ModelProviderCard = { chatModels: [ { + description: + '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。', displayName: 'Moonshot V1 128K', enabled: true, functionCall: true, @@ -11,6 +13,8 @@ const Moonshot: ModelProviderCard = { tokens: 128_000, }, { + description: + '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。', displayName: 'Moonshot V1 32K', enabled: true, functionCall: true, @@ -18,6 +22,8 @@ const Moonshot: ModelProviderCard = { tokens: 32_768, }, { + description: + '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。', displayName: 'Moonshot V1 8K', enabled: true, functionCall: true, @@ -26,12 +32,16 @@ const Moonshot: ModelProviderCard = { }, ], checkModel: 'moonshot-v1-8k', + description: + 'Moonshot 是由北京月之暗面科技有限公司推出的开源平台，提供多种自然语言处理模型，应用领域广泛，包括但不限于内容创作、学术研究、智能推荐、医疗诊断等，支持长文本处理和复杂生成任务。', id: 'moonshot', + modelsUrl: 'https://platform.moonshot.cn/docs/intro', name: 'Moonshot', smoothing: { speed: 2, text: true, }, + url: 'https://www.moonshot.cn', }; export default Moonshot; diff --git a/src/config/modelProviders/novita.ts b/src/config/modelProviders/novita.ts index 568bc8ad153e..461a4dfd9257 100644 --- a/src/config/modelProviders/novita.ts +++ b/src/config/modelProviders/novita.ts @@ -4,94 +4,118 @@ import { ModelProviderCard } from '@/types/llm'; const Novita: ModelProviderCard = { chatModels: [ { + description: + '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。', displayName: 'Llama 3.1 8B Instruct', enabled: true, id: 'meta-llama/llama-3.1-8b-instruct', tokens: 8192, }, { + description: + '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。', displayName: 'Llama 3.1 70B Instruct', enabled: true, id: 'meta-llama/llama-3.1-70b-instruct', tokens: 131_072, }, { + description: + '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。', displayName: 'Llama 3.1 405B Instruct', enabled: true, id: 'meta-llama/llama-3.1-405b-instruct', tokens: 32_768, }, { + 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。', displayName: 'Llama 3 8B Instruct', id: 'meta-llama/llama-3-8b-instruct', tokens: 8192, }, { + 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。', displayName: 'Llama 3 70B Instruct', id: 'meta-llama/llama-3-70b-instruct', tokens: 8192, }, { + 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。', displayName: 'Gemma 2 9B', enabled: true, id: 'google/gemma-2-9b-it', tokens: 8192, }, { + description: 'Mistral Nemo 是多语言支持和高性能编程的7.3B参数模型。', displayName: 'Mistral Nemo', enabled: true, id: 'mistralai/mistral-nemo', tokens: 32_768, }, { + 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。', displayName: 'Mistral 7B Instruct', enabled: true, id: 'mistralai/mistral-7b-instruct', tokens: 32_768, }, { + 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。', displayName: 'WizardLM 2 7B', enabled: true, id: 'microsoft/wizardlm 2-7b', tokens: 32_768, }, { + 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。', displayName: 'WizardLM-2 8x22B', enabled: true, id: 'microsoft/wizardlm-2-8x22b', tokens: 65_535, }, { + description: 'Dolphin Mixtral 8x22B 是一款为指令遵循、对话和编程设计的模型。', displayName: 'Dolphin Mixtral 8x22B', id: 'cognitivecomputations/dolphin-mixtral-8x22b', tokens: 16_000, }, { + description: + '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。', displayName: 'Hermes 2 Pro Llama 3 8B', id: 'nousresearch/hermes-2-pro-llama-3-8b', tokens: 8192, }, { + description: + '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。', displayName: 'Hermes 2 Mixtral 8x7B DPO', id: 'Nous-Hermes-2-Mixtral-8x7B-DPO', tokens: 32_768, }, { + 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。', displayName: 'MythoMax l2 13B', id: 'gryphe/mythomax-l2-13b', tokens: 4096, }, { + 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。', displayName: 'OpenChat 7B', id: 'openchat/openchat-7b', tokens: 4096, }, ], checkModel: 'meta-llama/llama-3.1-8b-instruct', + description: + 'Novita AI 是一个提供多种大语言模型与 AI 图像生成的 API 服务的平台，灵活、可靠且具有成本效益。它支持 Llama3、Mistral 等最新的开源模型，并为生成式 AI 应用开发提供了全面、用户友好且自动扩展的 API 解决方案，适合 AI 初创公司的快速发展。', disableBrowserRequest: true, id: 'novita', modelList: { showModelFetcher: true }, + modelsUrl: 'https://novita.ai/model-api/product/llm-api', name: 'Novita', + url: 'https://novita.ai', }; export default Novita; diff --git a/src/config/modelProviders/ollama.ts b/src/config/modelProviders/ollama.ts index c77769c3fe48..b6b1bde02242 100644 --- a/src/config/modelProviders/ollama.ts +++ b/src/config/modelProviders/ollama.ts @@ -1,208 +1,259 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://ollama.com/library +// ref: https://ollama.com/library const Ollama: ModelProviderCard = { chatModels: [ { + description: + 'Llama 3.1 是 Meta 推出的领先模型，支持高达 405B 参数，可应用于复杂对话、多语言翻译和数据分析领域。', displayName: 'Llama 3.1 8B', enabled: true, id: 'llama3.1', tokens: 128_000, }, { + description: + 'Llama 3.1 是 Meta 推出的领先模型，支持高达 405B 参数，可应用于复杂对话、多语言翻译和数据分析领域。', displayName: 'Llama 3.1 70B', id: 'llama3.1:70b', tokens: 128_000, }, { + description: + 'Llama 3.1 是 Meta 推出的领先模型，支持高达 405B 参数，可应用于复杂对话、多语言翻译和数据分析领域。', displayName: 'Llama 3.1 405B', id: 'llama3.1:405b', tokens: 128_000, }, { + description: + 'Code Llama 是一款专注于代码生成和讨论的 LLM，结合广泛的编程语言支持，适用于开发者环境。', displayName: 'Code Llama 7B', enabled: true, id: 'codellama', tokens: 16_384, }, { + description: + 'Code Llama 是一款专注于代码生成和讨论的 LLM，结合广泛的编程语言支持，适用于开发者环境。', displayName: 'Code Llama 13B', id: 'codellama:13b', tokens: 16_384, }, { + description: + 'Code Llama 是一款专注于代码生成和讨论的 LLM，结合广泛的编程语言支持，适用于开发者环境。', displayName: 'Code Llama 34B', id: 'codellama:34b', tokens: 16_384, }, { + description: + 'Code Llama 是一款专注于代码生成和讨论的 LLM，结合广泛的编程语言支持，适用于开发者环境。', displayName: 'Code Llama 70B', id: 'codellama:70b', tokens: 16_384, }, { + description: 'Gemma 2 是 Google 推出的高效模型，涵盖从小型应用到复杂数据处理的多种应用场景。', displayName: 'Gemma 2 2B', id: 'gemma2:2b', tokens: 8192, }, { + description: 'Gemma 2 是 Google 推出的高效模型，涵盖从小型应用到复杂数据处理的多种应用场景。', displayName: 'Gemma 2 9B', enabled: true, id: 'gemma2', tokens: 8192, }, { + description: 'Gemma 2 是 Google 推出的高效模型，涵盖从小型应用到复杂数据处理的多种应用场景。', displayName: 'Gemma 2 27B', id: 'gemma2:27b', tokens: 8192, }, { + description: 'CodeGemma 专用于不同编程任务的轻量级语言模型，支持快速迭代和集成。', displayName: 'CodeGemma 2B', id: 'codegemma:2b', tokens: 8192, }, { + description: 'CodeGemma 专用于不同编程任务的轻量级语言模型，支持快速迭代和集成。', displayName: 'CodeGemma 7B', enabled: true, id: 'codegemma', tokens: 8192, }, { + description: 'Phi-3 是微软推出的轻量级开放模型，适用于高效集成和大规模知识推理。', displayName: 'Phi-3 3.8B', enabled: true, id: 'phi3', tokens: 128_000, }, { + description: 'Phi-3 是微软推出的轻量级开放模型，适用于高效集成和大规模知识推理。', displayName: 'Phi-3 14B', id: 'phi3:14b', tokens: 128_000, }, { + description: + 'WizardLM 2 是微软AI提供的语言模型，在复杂对话、多语言、推理和智能助手领域表现尤为出色。', displayName: 'WizardLM 2 7B', enabled: true, id: 'wizardlm2', tokens: 32_768, }, { + description: + 'WizardLM 2 是微软AI提供的语言模型，在复杂对话、多语言、推理和智能助手领域表现尤为出色。', displayName: 'WizardLM 2 8x22B', id: 'wizardlm2:8x22b', tokens: 65_536, }, { + description: 'MathΣtral 专为科学研究和数学推理设计，提供有效的计算能力和结果解释。', displayName: 'MathΣtral 7B', enabled: true, id: 'mathstral', tokens: 32_768, }, { + description: 'Mistral 是 Mistral AI 发布的 7B 模型，适合多变的语言处理需求。', displayName: 'Mistral 7B', enabled: true, id: 'mistral', tokens: 32_768, }, { + description: + 'Mixtral 是 Mistral AI 的专家模型，具有开源权重，并在代码生成和语言理解方面提供支持。', displayName: 'Mixtral 8x7B', enabled: true, id: 'mixtral', tokens: 32_768, }, { + description: + 'Mixtral 是 Mistral AI 的专家模型，具有开源权重，并在代码生成和语言理解方面提供支持。', displayName: 'Mixtral 8x22B', id: 'mixtral:8x22b', tokens: 65_536, }, { + description: + 'Mixtral Large 是 Mistral 的旗舰模型，结合代码生成、数学和推理的能力，支持 128k 上下文窗口。', displayName: 'Mixtral Large 123B', enabled: true, id: 'mistral-large', tokens: 128_000, }, { + description: 'Mistral Nemo 由 Mistral AI 和 NVIDIA 合作推出，是高效性能的 12B 模型。', displayName: 'Mixtral Nemo 12B', enabled: true, id: 'mistral-nemo', tokens: 128_000, }, { + description: 'Codestral 是 Mistral AI 的首款代码模型，为代码生成任务提供优异支持。', displayName: 'Codestral 22B', enabled: true, id: 'codestral', tokens: 32_768, }, { + description: 'Aya 23 是 Cohere 推出的多语言模型，支持 23 种语言，为多元化语言应用提供便利。', displayName: 'Aya 23 8B', enabled: true, id: 'aya', tokens: 8192, }, { + description: 'Aya 23 是 Cohere 推出的多语言模型，支持 23 种语言，为多元化语言应用提供便利。', displayName: 'Aya 23 35B', id: 'aya:35b', tokens: 8192, }, { + description: 'Command R 是优化用于对话和长上下文任务的LLM，特别适合动态交互与知识管理。', displayName: 'Command R 35B', enabled: true, id: 'command-r', tokens: 131_072, }, { + description: 'Command R+ 是一款高性能的大型语言模型，专为真实企业场景和复杂应用而设计。', displayName: 'Command R+ 104B', enabled: true, id: 'command-r-plus', tokens: 131_072, }, { + description: 'DeepSeek V2 是高效的 Mixture-of-Experts 语言模型，适用于经济高效的处理需求。', displayName: 'DeepSeek V2 16B', enabled: true, id: 'deepseek-v2', tokens: 32_768, }, { + description: 'DeepSeek V2 236B 是 DeepSeek 的设计代码模型，提供强大的代码生成能力。', displayName: 'DeepSeek V2 236B', id: 'deepseek-v2:236b', tokens: 128_000, }, { + description: + 'DeepSeek Coder V2 是开源的混合专家代码模型，在代码任务方面表现优异，与 GPT4-Turbo 相媲美。', displayName: 'DeepSeek Coder V2 16B', enabled: true, id: 'deepseek-coder-v2', tokens: 128_000, }, { + description: + 'DeepSeek Coder V2 是开源的混合专家代码模型，在代码任务方面表现优异，与 GPT4-Turbo 相媲美。', displayName: 'DeepSeek Coder V2 236B', id: 'deepseek-coder-v2:236b', tokens: 128_000, }, { + description: 'Qwen2 是阿里巴巴的新一代大规模语言模型，以优异的性能支持多元化的应用需求。', displayName: 'Qwen2 0.5B', id: 'qwen2:0.5b', tokens: 128_000, }, { + description: 'Qwen2 是阿里巴巴的新一代大规模语言模型，以优异的性能支持多元化的应用需求。', displayName: 'Qwen2 1.5B', id: 'qwen2:1.5b', tokens: 128_000, }, { + description: 'Qwen2 是阿里巴巴的新一代大规模语言模型，以优异的性能支持多元化的应用需求。', displayName: 'Qwen2 7B', enabled: true, id: 'qwen2', tokens: 128_000, }, { + description: 'Qwen2 是阿里巴巴的新一代大规模语言模型，以优异的性能支持多元化的应用需求。', displayName: 'Qwen2 72B', id: 'qwen2:72b', tokens: 128_000, }, { + description: 'CodeQwen1.5 是基于大量代码数据训练的大型语言模型，专为解决复杂编程任务。', displayName: 'CodeQwen1.5 7B', enabled: true, id: 'codeqwen', tokens: 65_536, }, { + description: 'LLaVA 是结合视觉编码器和 Vicuna 的多模态模型，用于强大的视觉和语言理解。', displayName: 'LLaVA 7B', enabled: true, id: 'llava', @@ -210,12 +261,14 @@ const Ollama: ModelProviderCard = { vision: true, }, { + description: 'LLaVA 是结合视觉编码器和 Vicuna 的多模态模型，用于强大的视觉和语言理解。', displayName: 'LLaVA 13B', id: 'llava:13b', tokens: 4096, vision: true, }, { + description: 'LLaVA 是结合视觉编码器和 Vicuna 的多模态模型，用于强大的视觉和语言理解。', displayName: 'LLaVA 34B', id: 'llava:34b', tokens: 4096, @@ -223,10 +276,14 @@ const Ollama: ModelProviderCard = { }, ], defaultShowBrowserRequest: true, + description: + 'Ollama 提供的模型广泛涵盖代码生成、数学运算、多语种处理和对话互动等领域，支持企业级和本地化部署的多样化需求。', id: 'ollama', modelList: { showModelFetcher: true }, + modelsUrl: 'https://ollama.com/library', name: 'Ollama', showApiKey: false, + url: 'https://ollama.com', }; export default Ollama; diff --git a/src/config/modelProviders/openai.ts b/src/config/modelProviders/openai.ts index 0fa8a70efb18..3021fce7b4c9 100644 --- a/src/config/modelProviders/openai.ts +++ b/src/config/modelProviders/openai.ts @@ -1,11 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref: -// https://platform.openai.com/docs/deprecations +// ref: https://platform.openai.com/docs/deprecations const OpenAI: ModelProviderCard = { chatModels: [ { - description: 'Currently points to gpt-4o-mini-2024-07-18', + description: + 'GPT-4o mini是OpenAI在GPT-4 Omni之后推出的最新模型，支持图文输入并输出文本。作为他们最先进的小型模型，它比其他近期的前沿模型便宜很多，并且比GPT-3.5 Turbo便宜超过60%。它保持了最先进的智能，同时具有显著的性价比。GPT-4o mini在MMLU测试中获得了 82% 的得分，目前在聊天偏好上排名高于 GPT-4。', displayName: 'GPT-4o mini', enabled: true, functionCall: true, @@ -19,7 +19,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'Currently points to gpt-4o-2024-05-13', + description: + 'ChatGPT-4o 是一款动态模型，实时更新以保持当前最新版本。它结合了强大的语言理解与生成能力，适合于大规模应用场景，包括客户服务、教育和技术支持。', displayName: 'GPT-4o', enabled: true, functionCall: true, @@ -32,7 +33,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'Latest GPT-4o snapshot that supports Structured Outputs', + description: + 'ChatGPT-4o 是一款动态模型，实时更新以保持当前最新版本。它结合了强大的语言理解与生成能力，适合于大规模应用场景，包括客户服务、教育和技术支持。', displayName: 'GPT-4o (240806)', enabled: true, functionCall: true, @@ -45,7 +47,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'Latest GPT-4o snapshot that supports Structured Outputs', + description: + 'ChatGPT-4o 是一款动态模型，实时更新以保持当前最新版本。它结合了强大的语言理解与生成能力，适合于大规模应用场景，包括客户服务、教育和技术支持。', displayName: 'GPT-4o (240513)', functionCall: true, id: 'gpt-4o-2024-05-13', @@ -57,7 +60,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'Dynamic model continuously updated to the current version of GPT-4o in ChatGPT', + description: + 'ChatGPT-4o 是一款动态模型，实时更新以保持当前最新版本。它结合了强大的语言理解与生成能力，适合于大规模应用场景，包括客户服务、教育和技术支持。', displayName: 'ChatGPT-4o', enabled: true, id: 'chatgpt-4o-latest', @@ -69,7 +73,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'GPT-4 Turbo with Vision', + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', displayName: 'GPT-4 Turbo', functionCall: true, id: 'gpt-4-turbo', @@ -81,7 +86,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'GPT-4 Turbo 视觉版 (240409)', + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', displayName: 'GPT-4 Turbo Vision (240409)', functionCall: true, id: 'gpt-4-turbo-2024-04-09', @@ -93,7 +99,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - description: 'Currently points to gpt-4-0125-preview', + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', displayName: 'GPT-4 Turbo Preview', functionCall: true, id: 'gpt-4-turbo-preview', @@ -104,6 +111,8 @@ const OpenAI: ModelProviderCard = { tokens: 128_000, }, { + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', displayName: 'GPT-4 Turbo Preview (0125)', functionCall: true, id: 'gpt-4-0125-preview', @@ -114,7 +123,8 @@ const OpenAI: ModelProviderCard = { tokens: 128_000, }, { - description: 'Currently points to gpt-4-1106-vision-preview', // Will be discontinued on December 6, 2024 + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', // Will be discontinued on December 6, 2024 displayName: 'GPT-4 Turbo Vision Preview', id: 'gpt-4-vision-preview', pricing: { @@ -125,7 +135,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { - displayName: 'GPT-4 Turbo Vision Preview (1106)', // Will be discontinued on December 6, 2024 + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', // Will be discontinued on December 6, 2024 id: 'gpt-4-1106-vision-preview', pricing: { input: 10, @@ -135,6 +146,8 @@ const OpenAI: ModelProviderCard = { vision: true, }, { + description: + '最新的 GPT-4 Turbo 模型具备视觉功能。现在，视觉请求可以使用 JSON 模式和函数调用。 GPT-4 Turbo 是一个增强版本，为多模态任务提供成本效益高的支持。它在准确性和效率之间找到平衡，适合需要进行实时交互的应用程序场景。', displayName: 'GPT-4 Turbo Preview (1106)', functionCall: true, id: 'gpt-4-1106-preview', @@ -145,7 +158,8 @@ const OpenAI: ModelProviderCard = { tokens: 128_000, }, { - description: 'Currently points to gpt-4-0613', + description: + 'GPT-4 提供了一个更大的上下文窗口，能够处理更长的文本输入，适用于需要广泛信息整合和数据分析的场景。', displayName: 'GPT-4', functionCall: true, id: 'gpt-4', @@ -156,6 +170,8 @@ const OpenAI: ModelProviderCard = { tokens: 8192, }, { + description: + 'GPT-4 提供了一个更大的上下文窗口，能够处理更长的文本输入，适用于需要广泛信息整合和数据分析的场景。', displayName: 'GPT-4 (0613)', functionCall: true, id: 'gpt-4-0613', @@ -166,7 +182,8 @@ const OpenAI: ModelProviderCard = { tokens: 8192, }, { - description: 'Currently points to gpt-4-32k-0613', // Will be discontinued on June 6, 2025 + description: + 'GPT-4 提供了一个更大的上下文窗口，能够处理更长的文本输入，适用于需要广泛信息整合和数据分析的场景。', // Will be discontinued on June 6, 2025 displayName: 'GPT-4 32K', functionCall: true, id: 'gpt-4-32k', @@ -177,7 +194,10 @@ const OpenAI: ModelProviderCard = { tokens: 32_768, }, { - displayName: 'GPT-4 32K (0613)', // Will be discontinued on June 6, 2025 + // Will be discontinued on June 6, 2025 + description: + 'GPT-4 提供了一个更大的上下文窗口，能够处理更长的文本输入，适用于需要广泛信息整合和数据分析的场景。', + displayName: 'GPT-4 32K (0613)', functionCall: true, id: 'gpt-4-32k-0613', pricing: { @@ -199,6 +219,8 @@ const OpenAI: ModelProviderCard = { tokens: 16_385, }, { + description: + 'GPT 3.5 Turbo，适用于各种文本生成和理解任务，Currently points to gpt-3.5-turbo-0125', displayName: 'GPT-3.5 Turbo (0125)', functionCall: true, id: 'gpt-3.5-turbo-0125', @@ -209,6 +231,8 @@ const OpenAI: ModelProviderCard = { tokens: 16_385, }, { + description: + 'GPT 3.5 Turbo，适用于各种文本生成和理解任务，Currently points to gpt-3.5-turbo-0125', displayName: 'GPT-3.5 Turbo (1106)', functionCall: true, id: 'gpt-3.5-turbo-1106', @@ -219,6 +243,8 @@ const OpenAI: ModelProviderCard = { tokens: 16_385, }, { + description: + 'GPT 3.5 Turbo，适用于各种文本生成和理解任务，Currently points to gpt-3.5-turbo-0125', displayName: 'GPT-3.5 Turbo Instruct', id: 'gpt-3.5-turbo-instruct', pricing: { @@ -228,7 +254,9 @@ const OpenAI: ModelProviderCard = { tokens: 4096, }, { - description: 'Currently points to gpt-3.5-turbo-16k-0613', // Will be discontinued on September 13, 2024 + description: + 'GPT 3.5 Turbo，适用于各种文本生成和理解任务，Currently points to gpt-3.5-turbo-0125', // Will be discontinued on September 13, 2024 + displayName: 'GPT-3.5 Turbo 16K', id: 'gpt-3.5-turbo-16k', legacy: true, @@ -239,7 +267,10 @@ const OpenAI: ModelProviderCard = { tokens: 16_385, }, { - displayName: 'GPT-3.5 Turbo (0613)', // Will be discontinued on September 13, 2024 + description: + 'GPT-3.5 Turbo 是 OpenAI 的一款基础模型，结合了高效性和经济性，广泛用于文本生成、理解和分析，专为指导性提示进行调整，去除了与聊天相关的优化。', + displayName: 'GPT-3.5 Turbo (0613)', + // Will be discontinued on September 13, 2024 id: 'gpt-3.5-turbo-0613', legacy: true, pricing: { @@ -249,7 +280,8 @@ const OpenAI: ModelProviderCard = { tokens: 4096, }, { - description: 'Currently points to gpt-3.5-turbo-16k-0613', // Will be discontinued on September 13, 2024 + description: + 'GPT-3.5 Turbo 是 OpenAI 的一款基础模型，结合了高效性和经济性，广泛用于文本生成、理解和分析，专为指导性提示进行调整，去除了与聊天相关的优化。', // Will be discontinued on September 13, 2024 id: 'gpt-3.5-turbo-16k-0613', legacy: true, pricing: { @@ -260,6 +292,8 @@ const OpenAI: ModelProviderCard = { }, ], checkModel: 'gpt-4o-mini', + description: + 'OpenAI 是全球领先的人工智能研究机构，其开发的模型如GPT系列推动了自然语言处理的前沿。OpenAI 致力于通过创新和高效的AI解决方案改变多个行业。他们的产品具有显著的性能和经济性，广泛用于研究、商业和创新应用。', enabled: true, id: 'openai', modelList: { showModelFetcher: true }, diff --git a/src/config/modelProviders/openrouter.ts b/src/config/modelProviders/openrouter.ts index b55d518d28de..814fedced616 100644 --- a/src/config/modelProviders/openrouter.ts +++ b/src/config/modelProviders/openrouter.ts @@ -1,9 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://openrouter.ai/docs#models +// ref :https://openrouter.ai/docs#models const OpenRouter: ModelProviderCard = { chatModels: [ { + description: + '根据上下文长度、主题和复杂性，你的请求将发送到 Llama 3 70B Instruct、Claude 3.5 Sonnet（自我调节）或 GPT-4o。', displayName: 'Auto (best for prompt)', enabled: true, functionCall: false, @@ -12,6 +14,8 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: + 'Mistral 7B Instruct 是一款高效的多语言模型，优化用于对话和问答，能在资源受限的环境中表现出色。', displayName: 'Mistral 7B Instruct (free)', enabled: true, functionCall: false, @@ -20,6 +24,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Google 的 Gemma 7B 具有出色的计算效率，适适用于多种硬件架构，如GPU和TPU。', displayName: 'Google: Gemma 7B (free)', enabled: true, functionCall: false, @@ -28,6 +33,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'OpenChat 3.5 是一个创新的开源语言模型库，使用C-RLFT策略优化，支持多种任务。', displayName: 'OpenChat 3.5 (free)', enabled: true, functionCall: false, @@ -36,6 +42,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Nous Capybara 7B 专为基础任务设计，提供高效的语言处理能力。', displayName: 'Nous: Capybara 7B (free)', enabled: true, functionCall: false, @@ -44,6 +51,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Hugging Face的Zephyr 7B，以小巧身形实现了超越许多大型模型的性能。', displayName: 'Hugging Face: Zephyr 7B (free)', enabled: true, functionCall: false, @@ -52,6 +60,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'MythoMist 7B 为实验性最强的Mistral合并模型，结合了最新算法。', displayName: 'MythoMist 7B (free)', enabled: true, functionCall: false, @@ -60,6 +69,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Toppy M 7B 是一个开源模型，旨在通过开源科学推进人工智能的普及。', displayName: 'Toppy M 7B (free)', enabled: true, functionCall: false, @@ -68,6 +78,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Cinematika 7B 模型将电影剧本转换为小说风格，适用于创意内容生成。', displayName: 'Cinematika 7B (alpha) (free)', enabled: true, functionCall: false, @@ -76,6 +87,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'RWKV v5 World 3B 结合了 RNN 和 Transformer，实现了卓越的多语言处理能力。', displayName: 'RWKV v5 World 3B (free)', enabled: true, functionCall: false, @@ -84,6 +96,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'RWKV v5 3B AI Town 是为 AI 小镇项目前置量身定制的模型。', displayName: 'RWKV v5 3B AI Town (free)', enabled: true, functionCall: false, @@ -92,6 +105,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'RWKV v5 Eagle 7B 以超小成本实现高效的多语言翻译和处理，适用于边缘计算。', displayName: 'RWKV v5: Eagle 7B (free)', enabled: true, functionCall: false, @@ -100,6 +114,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: 'Mistral Mixtral 8x22B 提供高性能的基础模型。', displayName: 'Mistral: Mixtral 8x22B (base)', enabled: true, functionCall: false, @@ -108,6 +123,7 @@ const OpenRouter: ModelProviderCard = { vision: false, }, { + description: '微软的 WizardLM-2 8x22B 具有高容量和高效能特性。', displayName: 'Microsoft: WizardLM-2 8x22B', enabled: true, functionCall: false, @@ -117,9 +133,13 @@ const OpenRouter: ModelProviderCard = { }, ], checkModel: 'mistralai/mistral-7b-instruct:free', + description: + 'OpenRouter 是一个提供多种前沿大模型接口的服务平台，支持 OpenAI、Anthropic、LLaMA 及更多，适合多样化的开发和应用需求。用户可根据自身需求灵活选择最优的模型和价格，助力AI体验的提升。', id: 'openrouter', modelList: { showModelFetcher: true }, + modelsUrl: 'https://openrouter.ai/models', name: 'OpenRouter', + url: 'https://openrouter.ai', }; export default OpenRouter; diff --git a/src/config/modelProviders/perplexity.ts b/src/config/modelProviders/perplexity.ts index 3ea4d95b88f5..1fd3591bbdcf 100644 --- a/src/config/modelProviders/perplexity.ts +++ b/src/config/modelProviders/perplexity.ts @@ -1,55 +1,71 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.perplexity.ai/docs/model-cards +// ref :https://docs.perplexity.ai/docs/model-cards const Perplexity: ModelProviderCard = { chatModels: [ { + description: + 'Llama 3.1 Sonar Small Online 模型，具备8B参数，支持约127,000个标记的上下文长度，专为在线聊天设计，能高效处理各种文本交互。', displayName: 'Llama 3.1 Sonar Small Online', enabled: true, id: 'llama-3.1-sonar-small-128k-online', tokens: 128_000, }, { + description: + 'Llama 3.1 Sonar Large Online 模型，具备70B参数，支持约127,000个标记的上下文长度，适用于高容量和多样化聊天任务。', displayName: 'Llama 3.1 Sonar Large Online', enabled: true, id: 'llama-3.1-sonar-large-128k-online', tokens: 128_000, }, { + description: + 'Llama 3.1 Sonar Huge Online 模型，具备405B参数，支持约127,000个标记的上下文长度，设计用于复杂的在线聊天应用。', displayName: 'Llama 3.1 Sonar Huge Online', enabled: true, id: 'llama-3.1-sonar-huge-128k-online', tokens: 128_000, }, { + description: + 'Llama 3.1 Sonar Small Chat 模型，具备8B参数，专为离线聊天设计，支持约127,000个标记的上下文长度。', displayName: 'Llama 3.1 Sonar Small Chat', enabled: true, id: 'llama-3.1-sonar-small-128k-chat', tokens: 128_000, }, { + description: + 'Llama 3.1 Sonar Large Chat 模型，具备70B参数，支持约127,000个标记的上下文长度，适合于复杂的离线聊天任务。', displayName: 'Llama 3.1 Sonar Large Chat', enabled: true, id: 'llama-3.1-sonar-large-128k-chat', tokens: 128_000, }, { - displayName: 'Llama 3.1 8B Instruct', + description: + 'Llama 3.1 8B Instruct 模型，具备8B参数，支持画面指示任务的高效执行，提供优质的文本生成能力。', id: 'llama-3.1-8b-instruct', tokens: 128_000, }, { - displayName: 'Llama 3.1 70B Instruct', + description: + 'Llama 3.1 70B Instruct 模型，具备70B参数，能在大型文本生成和指示任务中提供卓越性能。', id: 'llama-3.1-70b-instruct', tokens: 128_000, }, ], checkModel: 'llama-3.1-8b-instruct', + description: + 'Perplexity 是一家领先的对话生成模型提供商，提供多种先进的Llama 3.1模型，支持在线和离线应用，特别适用于复杂的自然语言处理任务。', id: 'perplexity', + modelsUrl: 'https://docs.perplexity.ai/guides/model-cards', name: 'Perplexity', proxyUrl: { placeholder: 'https://api.perplexity.ai', }, + url: 'https://www.perplexity.ai', }; export default Perplexity; diff --git a/src/config/modelProviders/qwen.ts b/src/config/modelProviders/qwen.ts index 30f63b6338a7..b8c2a4b1cdc8 100644 --- a/src/config/modelProviders/qwen.ts +++ b/src/config/modelProviders/qwen.ts @@ -1,10 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://help.aliyun.com/zh/dashscope/developer-reference/api-details +// ref :https://help.aliyun.com/zh/dashscope/developer-reference/api-details const Qwen: ModelProviderCard = { chatModels: [ { - description: '通义千问超大规模语言模型，支持长文本上下文，以及基于长文档、多文档等多个场景的对话功能。', + description: + '通义千问超大规模语言模型，支持长文本上下文，以及基于长文档、多文档等多个场景的对话功能。', displayName: 'Qwen Long', enabled: true, id: 'qwen-long', @@ -61,7 +62,7 @@ const Qwen: ModelProviderCard = { tokens: 32_768, vision: true, }, - // ref https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-qianwen-7b-14b-72b-api-detailes + // ref :https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-qianwen-7b-14b-72b-api-detailes { description: '通义千问2对外开源的7B规模的模型', displayName: 'Qwen2 7B', @@ -87,16 +88,14 @@ const Qwen: ModelProviderCard = { tokens: 4096, // https://help.aliyun.com/zh/dashscope/developer-reference/use-qwen2-math-by-calling-api }, { - description: - '以 Qwen-7B 语言模型初始化，添加图像模型，图像输入分辨率为448的预训练模型。', + description: '以 Qwen-7B 语言模型初始化，添加图像模型，图像输入分辨率为448的预训练模型。', displayName: 'Qwen VL', id: 'qwen-vl-v1', tokens: 8192, // https://huggingface.co/Qwen/Qwen-VL/blob/main/config.json vision: true, }, { - description: - '通义千问VL支持灵活的交互方式，包括多图、多轮问答、创作等能力的模型。', + description: '通义千问VL支持灵活的交互方式，包括多图、多轮问答、创作等能力的模型。', displayName: 'Qwen VL Chat', id: 'qwen-vl-chat-v1', tokens: 8192, // https://huggingface.co/Qwen/Qwen-VL-Chat/blob/main/config.json @@ -104,14 +103,18 @@ const Qwen: ModelProviderCard = { }, ], checkModel: 'qwen-turbo', - disableBrowserRequest: true, // CORS issue + description: + '通义千问是阿里云自主研发的超大规模语言模型，具有强大的自然语言理解和生成能力。它可以回答各种问题、创作文字内容、表达观点看法、撰写代码等，在多个领域发挥作用。', + disableBrowserRequest: true, id: 'qwen', modelList: { showModelFetcher: true }, + modelsUrl: 'https://help.aliyun.com/zh/dashscope/developer-reference/api-details', name: 'Qwen', smoothing: { speed: 2, text: true, }, + url: 'https://tongyi.aliyun.com', }; export default Qwen; diff --git a/src/config/modelProviders/siliconcloud.ts b/src/config/modelProviders/siliconcloud.ts index bad0d84d0e4f..e4660d62ebf7 100644 --- a/src/config/modelProviders/siliconcloud.ts +++ b/src/config/modelProviders/siliconcloud.ts @@ -1,175 +1,208 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://siliconflow.cn/zh-cn/models +// ref :https://siliconflow.cn/zh-cn/models const SiliconCloud: ModelProviderCard = { chatModels: [ { + description: 'Qwen2 是全新的大型语言模型系列，旨在优化指令式任务的处理。', displayName: 'Qwen2 1.5B', id: 'Qwen/Qwen2-1.5B-Instruct', tokens: 32_768, }, { + description: 'Qwen2 是全新的大型语言模型系列，具有更强的理解和生成能力。', displayName: 'Qwen2 7B', enabled: true, id: 'Qwen/Qwen2-7B-Instruct', tokens: 32_768, }, { + description: 'Qwen2 是全新的系列，57B A14B 型号在指令任务中表现卓越。', displayName: 'Qwen2 57B A14B', id: 'Qwen/Qwen2-57B-A14B-Instruct', tokens: 32_768, }, { + description: 'Qwen2 是先进的通用语言模型，支持多种指令类型。', displayName: 'Qwen2 72B', enabled: true, id: 'Qwen/Qwen2-72B-Instruct', tokens: 32_768, }, { + description: 'Qwen2-Math 专注于数学领域的问题求解，为高难度题提供专业解答。', displayName: 'Qwen2 Math 72B', enabled: true, id: 'Qwen/Qwen2-Math-72B-Instruct', tokens: 32_768, }, { + description: 'Qwen1.5 通过结合高级预训练和微调提升对话表达能力。', displayName: 'Qwen1.5 7B', id: 'Qwen/Qwen1.5-7B-Chat', tokens: 32_768, }, { + description: 'Qwen1.5 通过大规模数据集训练，擅长复杂的语言任务。', displayName: 'Qwen1.5 14B', id: 'Qwen/Qwen1.5-14B-Chat', tokens: 32_768, }, { + description: 'Qwen1.5 具备多领域问答和文本生成的能力。', displayName: 'Qwen1.5 32B', id: 'Qwen/Qwen1.5-32B-Chat', tokens: 32_768, }, { + description: '作为Qwen2 的测试版，Qwen1.5 使用大规模数据实现了更精确的对话功能。', displayName: 'Qwen1.5 110B', id: 'Qwen/Qwen1.5-110B-Chat', tokens: 32_768, }, { + description: 'Yi-1.5 是Yi系列的进化版本，拥有高质量的预训练和丰富的微调数据。', displayName: 'Yi-1.5 6B', id: '01-ai/Yi-1.5-6B-Chat', tokens: 4096, }, { + description: 'Yi-1.5 9B 支持16K Tokens, 提供高效、流畅的语言生成能力。', displayName: 'Yi-1.5 9B', enabled: true, id: '01-ai/Yi-1.5-9B-Chat-16K', tokens: 16_384, }, { + description: 'Yi-1.5 34B, 以丰富的训练样本在行业应用中提供优越表现。', displayName: 'Yi-1.5 34B', enabled: true, id: '01-ai/Yi-1.5-34B-Chat-16K', tokens: 16_384, }, { + description: '作为双语会话语言模型, ChatGLM3能处理中英文转换任务。', displayName: 'GLM-3 6B', id: 'THUDM/chatglm3-6b', tokens: 32_768, }, { + description: 'GLM-4 9B 开放源码版本，为会话应用提供优化后的对话体验。', displayName: 'GLM-4 9B', enabled: true, id: 'THUDM/glm-4-9b-chat', tokens: 32_768, }, { + description: 'InternLM2.5 提供多场景下的智能对话解决方案。', displayName: 'Internlm 2.5 7B', enabled: true, id: 'internlm/internlm2_5-7b-chat', tokens: 32_768, }, { + description: '创新的开源模型InternLM2.5，通过大规模的参数提高了对话智能。', displayName: 'Internlm 2.5 20B', enabled: true, id: 'internlm/internlm2_5-20b-chat', tokens: 32_768, }, { + description: 'DeepSeek V2.5 集合了先前版本的优秀特征，增强了通用和编码能力。', displayName: 'DeepSeek V2.5', enabled: true, id: 'deepseek-ai/DeepSeek-V2.5', tokens: 32_768, }, { + description: 'DeepSeek V2 具备67亿参数，支持英中文本处理。', displayName: 'DeepSeek V2 Chat', id: 'deepseek-ai/DeepSeek-V2-Chat', tokens: 32_768, }, { + description: 'DeepSeek Coder V2 为代码任务设计, 专注于高效的代码生成。', displayName: 'DeepSeek V2 Coder', id: 'deepseek-ai/DeepSeek-Coder-V2-Instruct', tokens: 32_768, }, { + description: 'DeepSeek 67B 是为高复杂性对话训练的先进模型。', displayName: 'DeepSeek LLM 67B', id: 'deepseek-ai/deepseek-llm-67b-chat', tokens: 32_768, }, { + description: 'Gemma 2 是Google轻量化的开源文本模型系列。', displayName: 'Gemma 2 9B', enabled: true, id: 'google/gemma-2-9b-it', tokens: 8192, }, { + description: 'Gemma 2 延续了轻量化与高效的设计理念。', displayName: 'Gemma 2 27B', enabled: true, id: 'google/gemma-2-27b-it', tokens: 8192, }, { + description: 'LLaMA 3.1 提供多语言支持，是业界领先的生成模型之一。', displayName: 'Llama 3.1 8B', enabled: true, id: 'meta-llama/Meta-Llama-3.1-8B-Instruct', tokens: 32_768, }, { + description: 'LLaMA 3.1 70B 提供多语言的高效对话支持。', displayName: 'Llama 3.1 70B', enabled: true, id: 'meta-llama/Meta-Llama-3.1-70B-Instruct', tokens: 32_768, }, { + description: 'LLaMA 3.1 405B 是预训练和指令调整的强大机型。', displayName: 'Llama 3.1 405B', enabled: true, id: 'meta-llama/Meta-Llama-3.1-405B-Instruct', tokens: 32_768, }, { + description: 'Reflection Llama 3.1 通过Reflection-Tuning技术提升推理能力。', displayName: 'Reflection Llama 3.1 70B', id: 'mattshumer/Reflection-Llama-3.1-70B', tokens: 32_768, }, { + description: 'LLaMA 3 支持大容量文本生成和指令解析。', displayName: 'Llama 3 70B', id: 'meta-llama/Meta-Llama-3-70B-Instruct', tokens: 8192, }, { + description: 'Mistral 7B 是按需 fine-tuning的模型，为任务提供优化解答。', displayName: 'Mistral 7B', id: 'mistralai/Mistral-7B-Instruct-v0.2', tokens: 32_768, }, { + description: 'Mixtral 8x7B 是预训练的稀疏混合专家模型，用于通用性文本任务。', displayName: 'Mistral 8x7B', id: 'mistralai/Mixtral-8x7B-Instruct-v0.1', tokens: 32_768, }, ], checkModel: 'Qwen/Qwen2-1.5B-Instruct', + description: + 'SiliconFlow 致力于加速 AGI，以惠及人类，通过易用与成本低的 GenAI 堆栈提升大规模 AI 效率。', id: 'siliconcloud', modelList: { showModelFetcher: true }, + modelsUrl: 'https://siliconflow.cn/zh-cn/models', name: 'SiliconCloud', proxyUrl: { placeholder: 'https://api.siliconflow.cn/v1', }, + url: 'https://siliconflow.cn', }; export default SiliconCloud; diff --git a/src/config/modelProviders/spark.ts b/src/config/modelProviders/spark.ts index 3e030b94d299..4eaba9d8d658 100644 --- a/src/config/modelProviders/spark.ts +++ b/src/config/modelProviders/spark.ts @@ -1,11 +1,12 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://www.xfyun.cn/doc/spark/HTTP%E8%B0%83%E7%94%A8%E6%96%87%E6%A1%A3.html#_3-%E8%AF%B7%E6%B1%82%E8%AF%B4%E6%98%8E -// ref https://www.xfyun.cn/doc/spark/Web.html#_1-%E6%8E%A5%E5%8F%A3%E8%AF%B4%E6%98%8E +// ref: https://www.xfyun.cn/doc/spark/HTTP%E8%B0%83%E7%94%A8%E6%96%87%E6%A1%A3.html#_3-%E8%AF%B7%E6%B1%82%E8%AF%B4%E6%98%8E +// ref: https://www.xfyun.cn/doc/spark/Web.html#_1-%E6%8E%A5%E5%8F%A3%E8%AF%B4%E6%98%8E const Spark: ModelProviderCard = { chatModels: [ { - description: '轻量级大语言模型，低延迟，全免费 支持在线联网搜索功能 响应快速、便捷，全面免费开放 适用于低算力推理与模型精调等定制化场景', + description: + 'Spark Lite 是一款轻量级大语言模型，具备极低的延迟与高效的处理能力，完全免费开放，支持实时在线搜索功能。其快速响应的特性使其在低算力设备上的推理应用和模型微调中表现出色，为用户带来出色的成本效益和智能体验，尤其在知识问答、内容生成及搜索场景下表现不俗。', displayName: 'Spark Lite', enabled: true, functionCall: false, @@ -14,7 +15,8 @@ const Spark: ModelProviderCard = { tokens: 8192, }, { - description: '专业级大语言模型，兼顾模型效果与性能 数学、代码、医疗、教育等场景专项优化 支持联网搜索、天气、日期等多个内置插件 覆盖大部分知识问答、语言理解、文本创作等多个场景', + description: + 'Spark Pro 是一款为专业领域优化的高性能大语言模型，专注数学、编程、医疗、教育等多个领域，并支持联网搜索及内置天气、日期等插件。其优化后模型在复杂知识问答、语言理解及高层次文本创作中展现出色表现和高效性能，是适合专业应用场景的理想选择。', displayName: 'Spark Pro', enabled: true, functionCall: false, @@ -23,7 +25,8 @@ const Spark: ModelProviderCard = { tokens: 8192, }, { - description: '支持最长上下文的星火大模型，长文无忧 128K星火大模型强势来袭 通读全文，旁征博引 沟通无界，逻辑连贯', + description: + 'Spark Pro-128K 配置了特大上下文处理能力，能够处理多达128K的上下文信息，特别适合需通篇分析和长期逻辑关联处理的长文内容，可在复杂文本沟通中提供流畅一致的逻辑与多样的引用支持。', displayName: 'Spark Pro-128K', enabled: true, functionCall: false, @@ -32,7 +35,8 @@ const Spark: ModelProviderCard = { tokens: 128_000, }, { - description: '最全面的星火大模型版本，功能丰富 支持联网搜索、天气、日期等多个内置插件 核心能力全面升级，各场景应用效果普遍提升 支持System角色人设与FunctionCall函数调用', + description: + 'Spark3.5 Max 为功能最为全面的版本，支持联网搜索及众多内置插件。其全面优化的核心能力以及系统角色设定和函数调用功能，使其在各种复杂应用场景中的表现极为优异和出色。', displayName: 'Spark3.5 Max', enabled: true, functionCall: false, @@ -41,7 +45,8 @@ const Spark: ModelProviderCard = { tokens: 8192, }, { - description: '最强大的星火大模型版本，效果极佳 全方位提升效果，引领智能巅峰 优化联网搜索链路，提供精准回答 强化文本总结能力，提升办公生产力', + description: + 'Spark4.0 Ultra 是星火大模型系列中最为强大的版本，在升级联网搜索链路同时，提升对文本内容的理解和总结能力。它是用于提升办公生产力和准确响应需求的全方位解决方案，是引领行业的智能产品。', displayName: 'Spark4.0 Ultra', enabled: true, functionCall: false, @@ -51,13 +56,17 @@ const Spark: ModelProviderCard = { }, ], checkModel: 'general', + description: + '科大讯飞星火大模型提供多领域、多语言的强大 AI 能力，利用先进的自然语言处理技术，构建适用于智能硬件、智慧医疗、智慧金融等多种垂直场景的创新应用。', id: 'spark', modelList: { showModelFetcher: true }, + modelsUrl: 'https://xinghuo.xfyun.cn/spark', name: 'Spark', smoothing: { speed: 2, text: true, }, + url: 'https://www.xfyun.cn', }; export default Spark; diff --git a/src/config/modelProviders/stepfun.ts b/src/config/modelProviders/stepfun.ts index b6c03690c231..b37da210f792 100644 --- a/src/config/modelProviders/stepfun.ts +++ b/src/config/modelProviders/stepfun.ts @@ -1,45 +1,52 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://platform.stepfun.com/docs/llm/text +// ref: https://platform.stepfun.com/docs/llm/text // 根据文档，阶级星辰大模型的上下文长度，其 k 的含义均为 1000 const Stepfun: ModelProviderCard = { chatModels: [ { + description: '支持大规模上下文交互，适合复杂对话场景。', displayName: 'Step 2 16K', enabled: true, id: 'step-2-16k', tokens: 16_000, }, { + description: '具备超长上下文处理能力，尤其适合长文档分析。', displayName: 'Step 1 256K', id: 'step-1-256k', tokens: 256_000, }, { + description: '平衡性能与成本，适合一般场景。', displayName: 'Step 1 128K', enabled: true, id: 'step-1-128k', tokens: 128_000, }, { + description: '支持中等长度的对话，适用于多种应用场景。', displayName: 'Step 1 32K', enabled: true, id: 'step-1-32k', tokens: 32_000, }, { + description: '小型模型，适合轻量级任务。', displayName: 'Step 1 8K', enabled: true, id: 'step-1-8k', tokens: 8000, }, { + description: '高速模型，适合实时对话。', displayName: 'Step 1 Flash', enabled: true, id: 'step-1-flash', tokens: 8000, }, { + description: '支持视觉输入，增强多模态交互体验。', displayName: 'Step 1V 32K', enabled: true, id: 'step-1v-32k', @@ -47,6 +54,7 @@ const Stepfun: ModelProviderCard = { vision: true, }, { + description: '小型视觉模型，适合基本的图文任务。', displayName: 'Step 1V 8K', enabled: true, id: 'step-1v-8k', @@ -55,16 +63,20 @@ const Stepfun: ModelProviderCard = { }, ], checkModel: 'step-1-flash', + description: + '阶级星辰大模型具备行业领先的多模态及复杂推理能力，支持超长文本理解和强大的自主调度搜索引擎功能。', // after test, currently https://api.stepfun.com/v1/chat/completions has the CORS issue // So we should close the browser request mode disableBrowserRequest: true, id: 'stepfun', modelList: { showModelFetcher: true }, + modelsUrl: 'https://platform.stepfun.com/docs/llm/text', name: 'Stepfun', smoothing: { speed: 2, text: true, }, + url: 'https://stepfun.com', }; export default Stepfun; diff --git a/src/config/modelProviders/taichu.ts b/src/config/modelProviders/taichu.ts index da532927fe4b..b5bde15c0a06 100644 --- a/src/config/modelProviders/taichu.ts +++ b/src/config/modelProviders/taichu.ts @@ -1,10 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://ai-maas.wair.ac.cn/#/doc +// ref :https://ai-maas.wair.ac.cn/#/doc const Taichu: ModelProviderCard = { chatModels: [ { - description: '语言大模型', + description: + '紫东太初语言大模型具备超强语言理解能力以及文本创作、知识问答、代码编程、数学计算、逻辑推理、情感分析、文本摘要等能力。创新性地将大数据预训练与多源丰富知识相结合，通过持续打磨算法技术，并不断吸收海量文本数据中词汇、结构、语法、语义等方面的新知识，实现模型效果不断进化。为用户提供更加便捷的信息和服务以及更为智能化的体验。', displayName: 'Taichu-2.0', enabled: true, functionCall: false, @@ -13,9 +14,13 @@ const Taichu: ModelProviderCard = { }, ], checkModel: 'taichu_llm', + description: + '中科院自动化研究所和武汉人工智能研究院推出新一代多模态大模型，支持多轮问答、文本创作、图像生成、3D理解、信号分析等全面问答任务，拥有更强的认知、理解、创作能力，带来全新互动体验。', id: 'taichu', modelList: { showModelFetcher: true }, + modelsUrl: 'https://ai-maas.wair.ac.cn/#/doc', name: 'Taichu', + url: 'https://ai-maas.wair.ac.cn', }; export default Taichu; diff --git a/src/config/modelProviders/togetherai.ts b/src/config/modelProviders/togetherai.ts index ccb07b002062..1446f2c1a3e8 100644 --- a/src/config/modelProviders/togetherai.ts +++ b/src/config/modelProviders/togetherai.ts @@ -1,164 +1,200 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://docs.together.ai/docs/chat-models -// ref https://www.together.ai/pricing +// ref :https://docs.together.ai/docs/chat-models +// ref :https://www.together.ai/pricing const TogetherAI: ModelProviderCard = { chatModels: [ { + description: + 'Llama 3.1 8B 模型采用FP8量化，支持高达131,072个上下文标记，是开源模型中的佼佼者，适合复杂任务，表现优异于许多行业基准。', displayName: 'Llama 3.1 8B Instruct Turbo', enabled: true, id: 'meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo', tokens: 131_072, }, { + description: + 'Llama 3.1 70B 模型经过精细调整，适用于高负载应用，量化至FP8提供更高效的计算能力和准确性，确保在复杂场景中的卓越表现。', displayName: 'Llama 3.1 70B Instruct Turbo', enabled: true, id: 'meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo', tokens: 131_072, }, { + description: + '405B 的 Llama 3.1 Turbo 模型，为大数据处理提供超大容量的上下文支持，在超大规模的人工智能应用中表现突出。', displayName: 'Llama 3.1 405B Instruct Turbo', enabled: true, id: 'meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo', tokens: 8192, }, { + description: 'Llama 3 8B Instruct Turbo 是一款高效能的大语言模型，支持广泛的应用场景。', displayName: 'Llama 3 8B Instruct Turbo', id: 'meta-llama/Meta-Llama-3-8B-Instruct-Turbo', tokens: 8192, }, { + description: + 'Llama 3 70B Instruct Turbo 提供卓越的语言理解和生成能力，适合最苛刻的计算任务。', displayName: 'Llama 3 70B Instruct Turbo', id: 'meta-llama/Meta-Llama-3-70B-Instruct-Turbo', tokens: 8192, }, { + description: 'Llama 3 8B Instruct Lite 适合资源受限的环境，提供出色的平衡性能。', displayName: 'Llama 3 8B Instruct Lite', id: 'meta-llama/Meta-Llama-3-8B-Instruct-Lite', tokens: 8192, }, { + description: 'Llama 3 70B Instruct Lite 适合需要高效能和低延迟的环境。', displayName: 'Llama 3 70B Instruct Lite', id: 'meta-llama/Meta-Llama-3-70B-Instruct-Lite', tokens: 8192, }, { + description: 'LLaMA-3 Chat (8B) 提供多语言支持，涵盖丰富的领域知识。', displayName: 'LLaMA-3 Chat (8B)', id: 'meta-llama/Llama-3-8b-chat-hf', tokens: 8192, }, { + description: 'LLaMA-3 Chat (70B) 是功能强大的聊天模型，支持复杂的对话需求。', displayName: 'LLaMA-3 Chat (70B)', id: 'meta-llama/Llama-3-70b-chat-hf', tokens: 8192, }, { + description: 'LLaMA-2 Chat (13B) 提供优秀的语言处理能力和出色的交互体验。', displayName: 'LLaMA-2 Chat (13B)', id: 'meta-llama/Llama-2-13b-chat-hf', tokens: 4096, }, { + description: 'Gemma 2 9B 由Google开发，提供高效的指令响应和综合能力。', displayName: 'Gemma 2 9B', enabled: true, id: 'google/gemma-2-9b-it', tokens: 8192, }, { + description: 'Gemma 2 27B 是一款通用大语言模型，具有优异的性能和广泛的应用场景。', displayName: 'Gemma 2 27B', enabled: true, id: 'google/gemma-2-27b-it', tokens: 8192, }, { + description: 'Gemma Instruct (2B) 提供基本的指令处理能力，适合轻量级应用。', displayName: 'Gemma Instruct (2B)', id: 'google/gemma-2b-it', tokens: 8192, }, { + description: 'Mistral (7B) Instruct v0.3 提供高效的计算能力和自然语言理解，适合广泛的应用。', displayName: 'Mistral (7B) Instruct v0.3', enabled: true, id: 'mistralai/Mistral-7B-Instruct-v0.3', tokens: 32_768, }, { + description: 'Mistral (7B) Instruct v0.2 提供改进的指令处理能力和更精确的结果。', displayName: 'Mistral (7B) Instruct v0.2', id: 'mistralai/Mistral-7B-Instruct-v0.2', tokens: 32_768, }, { + description: 'Mistral (7B) Instruct 以高性能著称，适用于多种语言任务。', displayName: 'Mistral (7B) Instruct', id: 'mistralai/Mistral-7B-Instruct-v0.1', tokens: 8192, }, { + description: 'Mixtral-8x7B Instruct (46.7B) 提供高容量的计算框架，适合大规模数据处理。', displayName: 'Mixtral-8x7B Instruct (46.7B)', enabled: true, id: 'mistralai/Mixtral-8x7B-Instruct-v0.1', tokens: 32_768, }, { + description: 'Mixtral-8x22B Instruct (141B) 是一款超级大语言模型，支持极高的处理需求。', displayName: 'Mixtral-8x22B Instruct (141B)', enabled: true, id: 'mistralai/Mixtral-8x22B-Instruct-v0.1', tokens: 65_536, }, { + description: 'DeepSeek LLM Chat (67B) 是创新的 AI 模型 提供深度语言理解和互动能力。', displayName: 'DeepSeek LLM Chat (67B)', enabled: true, id: 'deepseek-ai/deepseek-llm-67b-chat', tokens: 4096, }, { + description: 'Qwen 2 Instruct (72B) 为企业级应用提供精准的指令理解和响应。', displayName: 'Qwen 2 Instruct (72B)', enabled: true, id: 'Qwen/Qwen2-72B-Instruct', tokens: 32_768, }, { + description: 'Qwen 1.5 Chat (72B) 提供快速响应和自然对话能力，适合多语言环境。', displayName: 'Qwen 1.5 Chat (72B)', id: 'Qwen/Qwen1.5-72B-Chat', tokens: 32_768, }, { + description: 'Qwen 1.5 Chat (110B) 是一款高效能的对话模型，支持复杂对话场景。', displayName: 'Qwen 1.5 Chat (110B)', id: 'Qwen/Qwen1.5-110B-Chat', tokens: 32_768, }, { + description: 'DBRX Instruct 提供高可靠性的指令处理能力，支持多行业应用。', displayName: 'DBRX Instruct', id: 'databricks/dbrx-instruct', tokens: 32_768, }, { + description: 'Upstage SOLAR Instruct v1 (11B) 适用于精细化指令任务，提供出色的语言处理能力。', displayName: 'Upstage SOLAR Instruct v1 (11B)', id: 'upstage/SOLAR-10.7B-Instruct-v1.0', tokens: 4096, }, { + description: 'Nous Hermes 2 - Mixtral 8x7B-DPO (46.7B) 是高精度的指令模型，适用于复杂计算。', displayName: 'Nous Hermes 2 - Mixtral 8x7B-DPO (46.7B)', id: 'NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO', tokens: 32_768, }, { + description: 'Nous Hermes-2 Yi (34B) 提供优化的语言输出和多样化的应用可能。', displayName: 'Nous Hermes-2 Yi (34B)', id: 'NousResearch/Nous-Hermes-2-Yi-34B', tokens: 4096, }, { + description: 'MythoMax-L2 (13B) 是一种创新模型，适合多领域应用和复杂任务。', displayName: 'MythoMax-L2 (13B)', id: 'Gryphe/MythoMax-L2-13b', tokens: 4096, }, { + description: 'StripedHyena Nous (7B) 通过高效的策略和模型架构，提供增强的计算能力。', displayName: 'StripedHyena Nous (7B)', id: 'togethercomputer/StripedHyena-Nous-7B', tokens: 32_768, }, ], checkModel: 'meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo', + description: + 'Together AI 致力于通过创新的 AI 模型实现领先的性能，提供广泛的自定义能力，包括快速扩展支持和直观的部署流程，满足企业的各种需求。', id: 'togetherai', modelList: { showModelFetcher: true }, + modelsUrl: 'https://docs.together.ai/docs/chat-models', name: 'Together AI', + url: 'https://www.together.ai', }; export default TogetherAI; diff --git a/src/config/modelProviders/upstage.ts b/src/config/modelProviders/upstage.ts index 382605dcd8fd..9f8d1d5299f3 100644 --- a/src/config/modelProviders/upstage.ts +++ b/src/config/modelProviders/upstage.ts @@ -1,10 +1,11 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://developers.upstage.ai/docs/getting-started/models +// ref :https://developers.upstage.ai/docs/getting-started/models const Upstage: ModelProviderCard = { chatModels: [ { - description: 'A more intelligent, instruction-following Solar LLM with IFEval 80+. The official version with expanded language support and longer context length will be released in November 2024. solar-pro supports English only at this time. solar-pro is an alias for our latest Solar Pro model. (Currently solar-pro-preview-240910)', + description: + 'Solar Pro 是 Upstage 推出的一款高智能LLM，专注于单GPU的指令跟随能力，IFEval得分80以上。目前支持英语，正式版本计划于2024年11月推出，将扩展语言支持和上下文长度。', displayName: 'Solar Pro', enabled: true, functionCall: false, @@ -12,7 +13,8 @@ const Upstage: ModelProviderCard = { tokens: 4096, }, { - description: 'A compact LLM offering superior performance to GPT-3.5, with robust multilingual capabilities for both English and Korean, delivering high efficiency in a smaller package. solar-1-mini-chat is alias for our latest solar-1-mini-chat model. (Currently solar-1-mini-chat-240612)', + description: + 'Solar Mini 是一种紧凑型 LLM，性能优于 GPT-3.5，具备强大的多语言能力，支持英语和韩语，提供高效小巧的解决方案。', displayName: 'Solar Mini', enabled: true, functionCall: true, @@ -20,7 +22,8 @@ const Upstage: ModelProviderCard = { tokens: 32_768, }, { - description: 'A compact LLM that extends the capabilities of solar-mini-chat with specialization in Japanese, while maintaining high efficiency and performance in English and Korean. solar-1-mini-chat-ja is alias for our latest solar-1-mini-chat-ja model.(Currently solar-1-mini-chat-ja-240612)', + description: + 'Solar Mini (Ja) 扩展了 Solar Mini 的能力，专注于日语，同时在英语和韩语的使用中保持高效和卓越性能。', displayName: 'Solar Mini (Ja)', functionCall: false, id: 'solar-1-mini-chat-ja', @@ -28,9 +31,13 @@ const Upstage: ModelProviderCard = { }, ], checkModel: 'solar-1-mini-chat', + description: + 'Upstage 专注于为各种商业需求开发AI模型，包括 Solar LLM 和文档 AI，旨在实现工作的人造通用智能（AGI）。通过 Chat API 创建简单的对话代理，并支持功能调用、翻译、嵌入以及特定领域应用。', id: 'upstage', modelList: { showModelFetcher: true }, + modelsUrl: 'https://developers.upstage.ai/docs/getting-started/models', name: 'Upstage', + url: 'https://upstage.ai', }; export default Upstage; diff --git a/src/config/modelProviders/zeroone.ts b/src/config/modelProviders/zeroone.ts index cd622de23337..0473829b9a90 100644 --- a/src/config/modelProviders/zeroone.ts +++ b/src/config/modelProviders/zeroone.ts @@ -1,6 +1,6 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://platform.lingyiwanwu.com/docs#%E6%A8%A1%E5%9E%8B +// ref :https://platform.lingyiwanwu.com/docs#%E6%A8%A1%E5%9E%8B const ZeroOne: ModelProviderCard = { chatModels: [ { @@ -71,8 +71,12 @@ const ZeroOne: ModelProviderCard = { }, ], checkModel: 'yi-spark', + description: + '01.AI 专注于AI 2.0时代的人工智能技术，大力推动“人+人工智能”的创新和应用，采用超强大模型和先进AI技术以提升人类生产力，实现技术赋能。', id: 'zeroone', + modelsUrl: 'https://platform.lingyiwanwu.com/docs', name: '01.AI', + url: 'https://01.ai', }; export default ZeroOne; diff --git a/src/config/modelProviders/zhipu.ts b/src/config/modelProviders/zhipu.ts index ca99e05b222e..2b83630b5f4d 100644 --- a/src/config/modelProviders/zhipu.ts +++ b/src/config/modelProviders/zhipu.ts @@ -1,13 +1,13 @@ import { ModelProviderCard } from '@/types/llm'; -// ref https://open.bigmodel.cn/dev/howuse/model +// ref :https://open.bigmodel.cn/dev/howuse/model // api https://open.bigmodel.cn/dev/api#language -// ref https://open.bigmodel.cn/modelcenter/square +// ref :https://open.bigmodel.cn/modelcenter/square const ZhiPu: ModelProviderCard = { chatModels: [ { description: - 'GLM-4-AllTools 是专门为支持智能体和相关任务而进一步优化的模型版本。它能够自主理解用户的意图，规划复杂的指令，并能够调用一个或多个工具（例如网络浏览器、代码解释器和文本生图像）以完成复杂的任务。', + 'GLM-4-AllTools 是一个多功能智能体模型，优化以支持复杂指令规划与工具调用，如网络浏览、代码解释和文本生成，适用于多任务执行。', displayName: 'GLM-4-AllTools', enabled: true, functionCall: true, @@ -15,7 +15,8 @@ const ZhiPu: ModelProviderCard = { tokens: 128_000, }, { - description: '高智能旗舰：性能全面提升，长文本和复杂任务能力显著增强', + description: + 'GLM-4-Plus 作为高智能旗舰，具备强大的处理长文本和复杂任务的能力，性能全面提升。', displayName: 'GLM-4-Plus', enabled: true, functionCall: true, @@ -23,7 +24,7 @@ const ZhiPu: ModelProviderCard = { tokens: 128_000, }, { - description: '高智能模型：适用于处理高度复杂和多样化的任务', + description: 'GLM-4-0520 是最新模型版本，专为高度复杂和多样化任务设计，表现卓越。', displayName: 'GLM-4-0520', enabled: true, functionCall: true, @@ -31,14 +32,14 @@ const ZhiPu: ModelProviderCard = { tokens: 128_000, }, { - description: '旧版旗舰：发布于2024年1月16日，目前已被 GLM-4-0520 取代', // deprecated on 2025-06 + description: 'GLM-4 是发布于2024年1月的旧旗舰版本，目前已被更强的 GLM-4-0520 取代。', displayName: 'GLM-4', functionCall: true, id: 'glm-4', tokens: 128_000, }, { - description: '性价比最高的版本，综合性能接近GLM-4，速度快，价格实惠', + description: 'GLM-4-Air 是性价比高的版本，性能接近GLM-4，提供快速度和实惠的价格。', displayName: 'GLM-4-Air', enabled: true, functionCall: true, @@ -46,14 +47,14 @@ const ZhiPu: ModelProviderCard = { tokens: 128_000, }, { - description: 'GLM-4-Air 的高性能版本，效果不变，推理速度达到其2.6倍', + description: 'GLM-4-AirX 提供 GLM-4-Air 的高效版本，推理速度可达其2.6倍。', displayName: 'GLM-4-AirX', functionCall: true, id: 'glm-4-airx', tokens: 8192, }, { - description: '超长输入：专为处理超长文本和记忆型任务设计', + description: 'GLM-4-Long 支持超长文本输入，适合记忆型任务与大规模文档处理。', displayName: 'GLM-4-Long', enabled: true, functionCall: true, @@ -61,7 +62,7 @@ const ZhiPu: ModelProviderCard = { tokens: 1_024_000, }, { - description: '适用简单任务，速度最快，价格最实惠的版本', + description: 'GLM-4-Flash 是处理简单任务的理想选择，速度最快且价格最优惠。', displayName: 'GLM-4-Flash', enabled: true, functionCall: true, @@ -69,8 +70,7 @@ const ZhiPu: ModelProviderCard = { tokens: 128_000, }, { - description: - '视频和图像理解：具备视频内容和多图片的理解能力', + description: 'GLM-4V-Plus 具备对视频内容及多图片的理解能力，适合多模态任务。', displayName: 'GLM-4V-Plus', enabled: true, id: 'glm-4v-plus', @@ -78,8 +78,7 @@ const ZhiPu: ModelProviderCard = { vision: true, }, { - description: - '图像理解：具备图像理解能力和推理能力', + description: 'GLM-4V 提供强大的图像理解与推理能力，支持多种视觉任务。', displayName: 'GLM-4V', enabled: true, id: 'glm-4v', @@ -88,28 +87,31 @@ const ZhiPu: ModelProviderCard = { }, { description: - 'CodeGeeX是一款强大的AI编程助手，提供智能问答和代码补全功能，支持多种编程语言，帮助开发者提高编程效率。', + 'CodeGeeX-4 是强大的AI编程助手，支持多种编程语言的智能问答与代码补全，提升开发效率。', displayName: 'CodeGeeX-4', id: 'codegeex-4', tokens: 128_000, }, { - description: - '支持基于人设的角色扮演、超长多轮的记忆、千人千面的角色对话，广泛应用于情感陪伴、游戏智能NPC、网红/明星/影视剧IP分身、数字人/虚拟主播、文字冒险游戏等拟人对话或游戏场景。', + description: 'CharGLM-3 专为角色扮演与情感陪伴设计，支持超长多轮记忆与个性化对话，应用广泛。', displayName: 'CharGLM-3', id: 'charglm-3', tokens: 4096, }, { - description: '心理模型：具备专业咨询能力，帮助用户理解情感并应对情绪问题', + description: 'Emohaa 是心理模型，具备专业咨询能力，帮助用户理解情感问题。', displayName: 'Emohaa', id: 'emohaa', tokens: 8192, }, ], checkModel: 'glm-4-flash', + description: + '智谱 AI 提供多模态与语言模型的开放平台，支持广泛的AI应用场景，包括文本处理、图像理解与编程辅助等。', id: 'zhipu', + modelsUrl: 'https://open.bigmodel.cn/dev/howuse/model', name: 'ZhiPu', + url: 'https://zhipuai.cn', }; export default ZhiPu; diff --git a/src/types/llm.ts b/src/types/llm.ts index 98faf65872ee..f1dd24efee52 100644 --- a/src/types/llm.ts +++ b/src/types/llm.ts @@ -80,6 +80,7 @@ export interface ModelProviderCard { * @default false */ defaultShowBrowserRequest?: boolean; + description?: string; /** * some provider server like stepfun and aliyun don't support browser request, * So we should disable it