diff --git a/README.md b/README.md index 8bfaa9377..1e65a0b1c 100644 --- a/README.md +++ b/README.md @@ -42,6 +42,8 @@ Start building LLM-empowered multi-agent applications in an easier way. ## News +- new**[2025-01-04]** AgentScope supports Anthropic API now. + - new**[2024-12-12]** We have updated the [roadmap of AgentScope](https://github.com/modelscope/agentscope/blob/main/docs/ROADMAP.md). - **[2024-09-06]** AgentScope version 0.1.0 is released now. @@ -95,6 +97,7 @@ services and third-party model APIs. | LiteLLM API | Chat | [`LiteLLMChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#litellm-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/litellm_chat_template.json) | [models supported by litellm](https://docs.litellm.ai/docs/)... | | Yi API | Chat | [`YiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/yi_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/yi_chat_template.json) | yi-large, yi-medium, ... | | Post Request based API | - | [`PostAPIModelWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#post-request-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/postapi_model_config_template.json) | - | +| Anthropic API | Chat | [`AnthropicChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/anthropic_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#anthropic-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/anthropic_chat_model_config_template.json) | claude-3-5-sonnet-20241022, ... | **Supported Local Model Deployment** diff --git a/README_JA.md b/README_JA.md index 71f8500d9..8f5c26575 100644 --- a/README_JA.md +++ b/README_JA.md @@ -42,6 +42,8 @@ LLMを活用したマルチエージェントアプリケーションをより ## ニュース +- new**[2025-01-04]** AgentScopeが現在Anthropic APIをサポートしています。 + - new**[2024-12-12]** AgentScopeの[ロードマップ](https://github.com/modelscope/agentscope/blob/main/docs/ROADMAP.md)を更新しました。 - **[2024-09-06]** AgentScopeバージョン0.1.0がリリースされました。 @@ -90,6 +92,7 @@ AgentScopeは、ローカルモデルサービスとサードパーティのモ | LiteLLM API | チャット | [`LiteLLMChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py) | [ガイダンス](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#litellm-api)
[テンプレート](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/litellm_chat_template.json) | [litellmがサポートするモデル](https://docs.litellm.ai/docs/)... | | Yi API | チャット | [`YiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/yi_model.py) | [ガイダンス](https://modelscope.github.io/agentscope/en/tutorial/203-model.html)
[テンプレート](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/yi_chat_template.json) | yi-large, yi-medium, ... | | Post Request based API | - | [`PostAPIModelWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | [ガイダンス](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#post-request-api)
[テンプレート](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/postapi_model_config_template.json) | - | +| Anthropic API | Chat | [`AnthropicChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/anthropic_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#anthropic-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/anthropic_chat_model_config_template.json) | claude-3-5-sonnet-20241022, ... | **サポートされているローカルモデルのデプロイ** diff --git a/README_ZH.md b/README_ZH.md index e1cf48158..54e57196f 100644 --- a/README_ZH.md +++ b/README_ZH.md @@ -42,6 +42,8 @@ ## 新闻 +- new**[2025-01-04]** AgentScope 已支持 Anthropic API + - new**[2024-12-12]** [AgentScope 开发路线图](https://github.com/modelscope/agentscope/blob/main/docs/ROADMAP.md) 已更新 - **[2024-09-06]** AgentScope v0.1.0 版本已上线 @@ -70,26 +72,26 @@ AgentScope是一个创新的多智能体开发平台，旨在赋予开发人员 AgentScope提供了一系列`ModelWrapper`来支持本地模型服务和第三方模型API。 -| API | Task | Model Wrapper | Configuration | Some Supported Models | -|------------------------|-----------------|---------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------| -| OpenAI API | Chat | [`OpenAIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_chat_template.json) | gpt-4o, gpt-4, gpt-3.5-turbo, ... | -| | Embedding | [`OpenAIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_embedding_template.json) | text-embedding-ada-002, ... | -| | DALL·E | [`OpenAIDALLEWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_dall_e_template.json) | dall-e-2, dall-e-3 | -| DashScope API | Chat | [`DashScopeChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_chat_template.json) | qwen-plus, qwen-max, ... | -| | Image Synthesis | [`DashScopeImageSynthesisWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_image_synthesis_template.json) | wanx-v1 | -| | Text Embedding | [`DashScopeTextEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_text_embedding_template.json) | text-embedding-v1, text-embedding-v2, ... | -| | Multimodal | [`DashScopeMultiModalWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_multimodal_template.json) | qwen-vl-max, qwen-vl-chat-v1, qwen-audio-chat | -| Gemini API | Chat | [`GeminiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/gemini_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#gemini-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/gemini_chat_template.json) | gemini-pro, ... | -| | Embedding | [`GeminiEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/gemini_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#gemini-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/gemini_embedding_template.json) | models/embedding-001, ... | -| ZhipuAI API | Chat | [`ZhipuAIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/zhipu_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#zhipu-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/zhipu_chat_template.json) | glm-4, ... | -| | Embedding | [`ZhipuAIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/zhipu_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#zhipu-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/zhipu_embedding_template.json) | embedding-2, ... | -| ollama | Chat | [`OllamaChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_chat_template.json) | llama3, llama2, Mistral, ... | -| | Embedding | [`OllamaEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_embedding_template.json) | llama2, Mistral, ... | -| | Generation | [`OllamaGenerationWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_generate_template.json) | llama2, Mistral, ... | -| LiteLLM API | Chat | [`LiteLLMChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#litellm-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/litellm_chat_template.json) | [models supported by litellm](https://docs.litellm.ai/docs/)... | -| Yi API | Chat | [`YiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/yi_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/yi_chat_template.json) | yi-large, yi-medium, ... | -| Post Request based API | - | [`PostAPIModelWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#post-request-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/postapi_model_config_template.json) | - | - +| API | Task | Model Wrapper | Configuration | Some Supported Models | +|------------------------|-----------------|---------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------| +| OpenAI API | Chat | [`OpenAIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_chat_template.json) | gpt-4o, gpt-4, gpt-3.5-turbo, ... | +| | Embedding | [`OpenAIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_embedding_template.json) | text-embedding-ada-002, ... | +| | DALL·E | [`OpenAIDALLEWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/openai_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#openai-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/openai_dall_e_template.json) | dall-e-2, dall-e-3 | +| DashScope API | Chat | [`DashScopeChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_chat_template.json) | qwen-plus, qwen-max, ... | +| | Image Synthesis | [`DashScopeImageSynthesisWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_image_synthesis_template.json) | wanx-v1 | +| | Text Embedding | [`DashScopeTextEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_text_embedding_template.json) | text-embedding-v1, text-embedding-v2, ... | +| | Multimodal | [`DashScopeMultiModalWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/dashscope_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#dashscope-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/dashscope_multimodal_template.json) | qwen-vl-max, qwen-vl-chat-v1, qwen-audio-chat | +| Gemini API | Chat | [`GeminiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/gemini_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#gemini-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/gemini_chat_template.json) | gemini-pro, ... | +| | Embedding | [`GeminiEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/gemini_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#gemini-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/gemini_embedding_template.json) | models/embedding-001, ... | +| ZhipuAI API | Chat | [`ZhipuAIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/zhipu_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#zhipu-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/zhipu_chat_template.json) | glm-4, ... | +| | Embedding | [`ZhipuAIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/zhipu_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#zhipu-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/zhipu_embedding_template.json) | embedding-2, ... | +| ollama | Chat | [`OllamaChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_chat_template.json) | llama3, llama2, Mistral, ... | +| | Embedding | [`OllamaEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_embedding_template.json) | llama2, Mistral, ... | +| | Generation | [`OllamaGenerationWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/ollama_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#ollama-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/ollama_generate_template.json) | llama2, Mistral, ... | +| LiteLLM API | Chat | [`LiteLLMChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#litellm-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/litellm_chat_template.json) | [models supported by litellm](https://docs.litellm.ai/docs/)... | +| Yi API | Chat | [`YiChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/yi_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/yi_chat_template.json) | yi-large, yi-medium, ... | +| Post Request based API | - | [`PostAPIModelWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#post-request-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/postapi_model_config_template.json) | - | +| Anthropic API | Chat | [`AnthropicChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/anthropic_model.py) | [guidance](https://modelscope.github.io/agentscope/en/tutorial/203-model.html#anthropic-api)
[template](https://github.com/modelscope/agentscope/blob/main/examples/model_configs_template/anthropic_chat_model_config_template.json) | claude-3-5-sonnet-20241022, ... | **支持的本地模型部署** AgentScope支持使用以下库快速部署本地模型服务。 diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md index 79e2f445f..bfe0716fc 100644 --- a/docs/ROADMAP.md +++ b/docs/ROADMAP.md @@ -23,6 +23,8 @@ Offering **agent-oriented programming (AOP)** as a new programming model to orga 2. Tools Calling + - ✅ Add Support for Anthropic API. + - 🚧 Support tools calling in user-assistant conversations. - OpenAI API - DashScope API diff --git a/docs/sphinx_doc/en/source/tutorial/203-model.md b/docs/sphinx_doc/en/source/tutorial/203-model.md index 1a69b0c49..06316b241 100644 --- a/docs/sphinx_doc/en/source/tutorial/203-model.md +++ b/docs/sphinx_doc/en/source/tutorial/203-model.md @@ -20,6 +20,7 @@ Currently, AgentScope supports the following model service APIs: - Post Request API, model inference services based on Post requests, including Huggingface/ModelScope Inference API and various post request based model APIs. +- Anthropic Chat API. ## Configuration @@ -94,6 +95,7 @@ In the current AgentScope, the supported `model_type` types, the corresponding | | Chat | [`PostAPIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `"post_api_chat"` | meta-llama/Meta-Llama-3-8B-Instruct, ... | | | Image Synthesis | [`PostAPIDALLEWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `post_api_dall_e` | - | | | | Embedding | [`PostAPIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `post_api_embedding` | - | +| Anthropic API | Chat | [`AnthropicChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/anthropic_model.py) | `"anthropic_chat"` | claude-3-5-sonnet-20241022, ... | #### Detailed Parameters @@ -538,6 +540,30 @@ com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py">agen +
+ +#### Anthropic API + +
+ +Anthropic Chat API (agentscope.models.AnthropicChatWrapper) + + +```python +{ + "model_config": "my_anthropic_chat_config", + "model_type": "anthropic_chat", + "model_name": "claude-3-5-sonnet-20241022", + + # Required parameters + "api_key": "{your_api_key}", + + # Optional parameters + "temperature": 0.5 +} +``` +
+
diff --git a/docs/sphinx_doc/zh_CN/source/tutorial/203-model.md b/docs/sphinx_doc/zh_CN/source/tutorial/203-model.md index 43eecc7fe..8ce88f7e3 100644 --- a/docs/sphinx_doc/zh_CN/source/tutorial/203-model.md +++ b/docs/sphinx_doc/zh_CN/source/tutorial/203-model.md @@ -16,6 +16,7 @@ AgentScope中，模型的部署和调用是通过`ModelWrapper`来解耦开的 - LiteLLM API, 包括对话（Chat）, 支持各种模型的API. - Post请求API，基于Post请求实现的模型推理服务，包括Huggingface/ModelScope Inference API和各种符合Post请求格式的API。 +- Anthropic 对话 API。 ## 配置方式 @@ -114,6 +115,7 @@ API如下： | | Chat | [`PostAPIChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `"post_api_chat"` | meta-llama/Meta-Llama-3-8B-Instruct, ... | | | Image Synthesis | [`PostAPIDALLEWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `post_api_dall_e` | - | | | | Embedding | [`PostAPIEmbeddingWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/post_model.py) | `post_api_embedding` | - | +| Anthropic API | Chat | [`AnthropicChatWrapper`](https://github.com/modelscope/agentscope/blob/main/src/agentscope/models/anthropic_model.py) | `"anthropic_chat"` | claude-3-5-sonnet-20241022, ... | #### 详细参数 @@ -560,8 +562,29 @@ com/modelscope/agentscope/blob/main/src/agentscope/models/litellm_model.py">agen +
+ +#### Anthropic API + +
+ +Anthropic Chat API (agentscope.models.AnthropicChatWrapper) + + +```python +{ + "model_config": "my_anthropic_chat_config", + "model_type": "anthropic_chat", + "model_name": "claude-3-5-sonnet-20241022", + # 必要参数 + "api_key": "{your_api_key}", + # 可选参数 + "temperature": 0.5 +} +``` +

diff --git a/examples/model_configs_template/anthropic_chat_model_config_template.json b/examples/model_configs_template/anthropic_chat_model_config_template.json new file mode 100644 index 000000000..f697c0d83 --- /dev/null +++ b/examples/model_configs_template/anthropic_chat_model_config_template.json @@ -0,0 +1,9 @@ +[ + { + "config_name": "anthropic_chat-claude-3-5-sonnet-20241022", + "model_type": "anthropic_chat", + "model_name": "claude-3-5-sonnet-20241022", + + "api_key": "{your_api_key}" + } +] \ No newline at end of file diff --git a/examples/model_configs_template/postapi_model_config_template.json b/examples/model_configs_template/postapi_model_config_template.json index 3f4586621..a2bfbe0c1 100644 --- a/examples/model_configs_template/postapi_model_config_template.json +++ b/examples/model_configs_template/postapi_model_config_template.json @@ -1,6 +1,7 @@ [{ "config_name": "post_api_chat-flask_llama2-7b-chat-hf", "model_type": "post_api_chat", + "model_name": "llama2", "api_url": "http://127.0.0.1:8000/llm/", "json_args": { "max_length": 4096, @@ -10,6 +11,7 @@ { "config_name": "post_api_chat-flask_llama2-7b-chat-ms", "model_type": "post_api_chat", + "model_name": "llama2-7b", "api_url": "http://127.0.0.1:8000/llm/", "json_args": { "max_length": 4096, @@ -19,6 +21,7 @@ { "config_name": "post_api-fastchat_llama2-7b-chat-hf", "model_type": "openai_chat", + "model_name": "llama2-7b", "api_key": "EMPTY", "client_args": { "base_url": "http://127.0.0.1:8000/v1/" @@ -30,6 +33,7 @@ { "config_name": "post_api-vllm_llama2-7b-chat-hf", "model_type": "openai_chat", + "model_name": "llama2-7b", "api_key": "EMPTY", "client_args": { "base_url": "http://127.0.0.1:8000/v1/" @@ -41,6 +45,7 @@ { "config_name": "post_api_chat-model-inference-api-gpt2", "model_type": "post_api_chat", + "model_name": "gpt2", "headers": { "Authorization": "Bearer {YOUR_API_TOKEN}" }, @@ -49,6 +54,7 @@ { "config_name": "post_api_img_syn_config", "model_type": "post_api_dall_e", + "model_name": "{YOUR_MODEL}", "api_url": "http://xxx.xxx.xxx.xxx:xxxx/xxx", "headers": { "Content-Type": "application/json", @@ -67,6 +73,7 @@ { "config_name": "post_api_embedding_config", "model_type": "post_api_embedding", + "model_name": "{YOUR_MODEL}", "api_url": "http://xxx.xxx.xxx.xxx:xxxx/xxx", "headers": { "Content-Type": "application/json", diff --git a/setup.py b/setup.py index ecf690b6e..3b090c2f8 100644 --- a/setup.py +++ b/setup.py @@ -93,6 +93,7 @@ extra_litellm_requires = ["litellm"] extra_zhipuai_requires = ["zhipuai"] extra_ollama_requires = ["ollama>=0.1.7"] +extra_anthropic_requires = ["anthropic"] # Full requires extra_full_requires = ( @@ -105,6 +106,7 @@ + extra_litellm_requires + extra_zhipuai_requires + extra_ollama_requires + + extra_anthropic_requires ) # For online workstation @@ -143,6 +145,7 @@ "litellm": extra_litellm_requires, "zhipuai": extra_zhipuai_requires, "gemini": extra_gemini_requires, + "anthropic": extra_anthropic_requires, # For service functions "service": extra_service_requires, # For distribution mode diff --git a/src/agentscope/models/__init__.py b/src/agentscope/models/__init__.py index 60de46f44..9aad55d9a 100644 --- a/src/agentscope/models/__init__.py +++ b/src/agentscope/models/__init__.py @@ -38,6 +38,7 @@ from .yi_model import ( YiChatWrapper, ) +from .anthropic_model import AnthropicChatWrapper _BUILD_IN_MODEL_WRAPPERS = [ "PostAPIChatWrapper", @@ -57,6 +58,7 @@ "ZhipuAIEmbeddingWrapper", "LiteLLMChatWrapper", "YiChatWrapper", + "AnthropicChatWrapper", ] __all__ = [ @@ -81,4 +83,5 @@ "ZhipuAIEmbeddingWrapper", "LiteLLMChatWrapper", "YiChatWrapper", + "AnthropicChatWrapper", ] diff --git a/src/agentscope/models/anthropic_model.py b/src/agentscope/models/anthropic_model.py new file mode 100644 index 000000000..5c042be29 --- /dev/null +++ b/src/agentscope/models/anthropic_model.py @@ -0,0 +1,326 @@ +# -*- coding: utf-8 -*- +"""The Anthropic model wrapper for AgentScope.""" +from typing import Optional, Union, Generator, Any, Sequence + +from ..manager import FileManager +from ..message import Msg +from .model import ModelWrapperBase, ModelResponse +from ..utils.common import ( + _guess_type_by_extension, + _is_web_url, + _get_base64_from_image_path, +) + + +class AnthropicChatWrapper(ModelWrapperBase): + """The Anthropic model wrapper for AgentScope.""" + + model_type: str = "anthropic_chat" + + _supported_image_format: list[str] = ["jpeg", "png", "gif", "webp"] + + def __init__( + self, + model_name: str, + config_name: Optional[str] = None, + api_key: Optional[str] = None, + stream: bool = False, + client_kwargs: Optional[dict] = None, + ) -> None: + """Initialize the Anthropic model wrapper. + + Args: + model_name (`str`): + The name of the used model, e.g. `claude-3-5-sonnet-20241022`. + config_name (`Optional[str]`, defaults to `None`): + The name of the model configuration. + api_key (`Optional[str]`, defaults to `None`): + The API key for the Anthropic API. + stream (`bool`, defaults to `False`): + Enable streaming mode or not. + client_kwargs (`Optional[dict]`, defaults to `None`): + The additional keyword arguments for the anthropic client. + """ + super().__init__(config_name, model_name) + + try: + import anthropic + except ImportError as e: + raise ImportError( + "Please install the `anthropic` package by running " + "`pip install anthropic`.", + ) from e + + client_kwargs = client_kwargs or {} + + self.client = anthropic.Anthropic( + api_key=api_key, + **client_kwargs, + ) + self.stream = stream + + def format( + self, + *args: Union[Msg, Sequence[Msg]], + ) -> list[dict[str, object]]: + """Format the messages for anthropic model input. + + TODO: Add support for multimodal input. + + Args: + *args (`Union[Msg, list[Msg]]`): + The message(s) to be formatted. + + Returns: + `list[dict[str, object]]`: + A list of formatted messages. + """ + return ModelWrapperBase.format_for_common_chat_models(*args) + + @staticmethod + def _format_msg_with_url( + msg: Msg, + ) -> dict[str, Union[str, list[dict]]]: + """Format a message with image urls into the format that anthropic + LLM requires. + + Refer to https://docs.anthropic.com/en/api/messages-examples + + Args: + msg (`Msg`): + The message object to be formatted + + Returns: + `dict[str, Union[str, list[dict]]]`: + The message in the required format. + """ + urls = [msg.url] if isinstance(msg.url, str) else msg.url + + image_urls = [] + for url in urls: + if _guess_type_by_extension(url) == "image": + image_urls.append(url) + + content = [] + for image_url in image_urls: + extension = image_url.split(".")[-1].lower() + extension = "jpeg" if extension == "jpg" else extension + if extension not in AnthropicChatWrapper._supported_image_format: + raise TypeError( + "Anthropic model only supports image formats " + f"{AnthropicChatWrapper._supported_image_format}, " + f"got {extension}", + ) + + if _is_web_url(image_url): + # Download the image locally + file_manager = FileManager.get_instance() + image_url = file_manager.save_image(image_url) + + data_base64 = _get_base64_from_image_path(image_url) + + content.append( + { + "type": "image", + "source": [ + { + "type": "base64", + "media_type": f"image/{extension}", + "data": data_base64, + } + for _ in image_urls + ], + }, + ) + + if msg.content is not None: + content.append( + { + "type": "text", + "text": msg.content, + }, + ) + return { + "role": msg.role, + "content": content, + } + + def __call__( # pylint: disable=too-many-branches + self, + messages: list[dict[str, Union[str, list[dict]]]], + stream: Optional[bool] = None, + max_tokens: int = 2048, + **kwargs: Any, + ) -> ModelResponse: + """Call the Anthropic model. + + .. note:: The official Anthropic API supports system prompt by a + separate argument "system". For the convenience of the users, we + allow the system prompt to be the first message in the input messages. + + Args: + messages (`list[dict[str, Union[str, list[dict]]]]`): + A list of message dictionaries. Each dictionary should have + 'role' and 'content' keys. + stream (`Optional[bool]`, defaults to `None`): + Enable streaming mode or not. + max_tokens (`int`, defaults to `2048`): + The max tokens in generation. + **kwargs (`Any`): + The additional keyword arguments for the model. + + Returns: + `ModelResponse`: + The model response. + """ + # Check the input messages + if isinstance(messages, list): + if len(messages) == 0: + raise ValueError("The input messages should not be empty.") + + for msg in messages: + if not isinstance(msg, dict): + raise ValueError( + "The input messages should be a list of dictionaries, " + f"got {type(msg)}", + ) + if "role" not in msg or "content" not in msg: + raise ValueError( + f"Each message should have 'role' and 'content' keys, " + f"got {msg}", + ) + if msg["role"] not in ["assistant", "user", "system"]: + raise ValueError( + f"Invalid role {msg['role']}. The role must be one of " + f"['assistant', 'user', 'system']", + ) + + else: + raise ValueError( + "The input messages should be a list of dictionaries, " + f"got {type(messages)}", + ) + + # Check the stream + if stream is None: + stream = stream or self.stream + + # Prepare the keyword arguments + kwargs.update( + { + "model": self.model_name, + "stream": stream, + "max_tokens": max_tokens, + }, + ) + + # Extract the system message + if messages[0]["role"] == "system": + if not isinstance(messages[0]["content"], str): + raise ValueError( + "The content of the system message should be a string, " + f"got {type(messages[0]['content'])}", + ) + + kwargs["system"] = messages[0]["content"] + messages = messages[1:] + + kwargs["messages"] = messages + + # Call the model + response = self.client.messages.create(**kwargs) + + # Get the response according to the stream + if stream: + + def generator() -> Generator[str, None, None]: + # Used in model invocation recording + gathered_response = {} + + text = "" + current_block = {} + for chunk in response: + chunk = chunk.model_dump() + chunk_type = chunk.get("type", None) + + if chunk_type == "message_start": + gathered_response.update(**chunk["message"]) + + if chunk_type == "message_delta": + for key, cost in chunk.get("usage", {}).items(): + gathered_response["usage"][key] = ( + gathered_response["usage"].get(key, 0) + cost + ) + + if chunk_type == "content_block_start": + # Refresh the current block + current_block = chunk["content_block"] + + if chunk_type == "content_block_delta": + delta = chunk.get("delta", {}) + if delta.get("type", None) == "text_delta": + # To recover the complete response with multiple + # blocks in its content field + current_block["text"] = current_block.get( + "text", + "", + ) + delta.get("text", "") + # Used for feedback + text += delta.get("text", "") + yield text + + # TODO: Support tool calls in streaming mode + + if chunk_type == "content_block_stop": + gathered_response["content"].append(current_block) + + self._save_model_invocation_and_update_monitor( + kwargs, + gathered_response, + ) + + return ModelResponse( + stream=generator(), + ) + + else: + response = response.model_dump() + + # Save the model invocation and update the monitor + self._save_model_invocation_and_update_monitor( + kwargs, + response, + ) + + texts = [] + # Gather text from content blocks + for block in response.get("content", []): + if ( + isinstance(block, dict) + and block.get("type", None) == "text" + ): + texts.append(block.get("text", "")) + + # Return the response + return ModelResponse( + text="\n".join(texts), + raw=response, + ) + + def _save_model_invocation_and_update_monitor( + self, + kwargs: dict, + response: dict, + ) -> None: + self._save_model_invocation( + arguments=kwargs, + response=response, + ) + + usage = response.get("usage", None) + if usage is not None: + self.monitor.update_text_and_embedding_tokens( + model_name=self.model_name, + prompt_tokens=usage.get("input_tokens", 0), + completion_tokens=usage.get("output_tokens", 0), + ) diff --git a/src/agentscope/models/model.py b/src/agentscope/models/model.py index 0f2cca403..afc484269 100644 --- a/src/agentscope/models/model.py +++ b/src/agentscope/models/model.py @@ -5,7 +5,7 @@ import inspect import time from functools import wraps -from typing import Sequence, Any, Callable, Union, List +from typing import Sequence, Any, Callable, Union, List, Optional from loguru import logger @@ -102,8 +102,8 @@ class in model configuration.""" def __init__( self, # pylint: disable=W0613 - config_name: str, - model_name: str, + config_name: Optional[str] = None, + model_name: Optional[str] = None, **kwargs: Any, ) -> None: """Base class for model wrapper. @@ -112,17 +112,25 @@ def __init__( `__call__` function. Args: - config_name (`str`): + config_name (`Optional[str]`, defaults to `None`): The id of the model, which is used to extract configuration from the config file. - model_name (`str`): + model_name (`Optional[str]`, defaults to `None`): The name of the model. """ self.monitor = MonitorManager.get_instance() self.config_name = config_name + + if model_name is None: + raise ValueError( + "Model name should be provided for model " + f"configuration [{config_name}].", + ) + self.model_name = model_name - logger.info(f"Initialize model by configuration [{config_name}]") + + logger.debug(f"Initialize model by configuration [{config_name}]") def __call__(self, *args: Any, **kwargs: Any) -> ModelResponse: """Processing input with the model.""" diff --git a/src/agentscope/models/post_model.py b/src/agentscope/models/post_model.py index 48797627c..63f8e11f9 100644 --- a/src/agentscope/models/post_model.py +++ b/src/agentscope/models/post_model.py @@ -3,7 +3,7 @@ import json import time from abc import ABC -from typing import Any, Union, Sequence, List +from typing import Any, Union, Sequence, List, Optional import requests from loguru import logger @@ -25,6 +25,7 @@ def __init__( self, config_name: str, api_url: str, + model_name: Optional[str] = None, headers: dict = None, max_length: int = 2048, timeout: int = 30, @@ -42,6 +43,9 @@ def __init__( The id of the model. api_url (`str`): The url of the post request api. + model_name (`str`): + The name of the model. If `None`, the model name will be + extracted from the `json_args`. headers (`dict`, defaults to `None`): The headers of the api. Defaults to None. max_length (`int`, defaults to `2048`): @@ -76,13 +80,14 @@ def __init__( **post_args ) """ - if json_args is not None: - model_name = json_args.get( - "model", - json_args.get("model_name", None), - ) - else: - model_name = None + if model_name is None: + if json_args is not None: + model_name = json_args.get( + "model", + json_args.get("model_name", None), + ) + else: + model_name = None super().__init__(config_name=config_name, model_name=model_name) diff --git a/src/agentscope/utils/common.py b/src/agentscope/utils/common.py index bfdadb4e8..9ff974de1 100644 --- a/src/agentscope/utils/common.py +++ b/src/agentscope/utils/common.py @@ -390,6 +390,22 @@ def _is_web_url(url: str) -> bool: return parsed_url.scheme in ["http", "https", "ftp", "oss"] +def _get_base64_from_image_path(image_path: str) -> str: + """Get the base64 string from the image url. + + Args: + image_path (`str`): + The local path of the image. + """ + with open(image_path, "rb") as image_file: + base64_image = base64.b64encode(image_file.read()).decode( + "utf-8", + ) + extension = image_path.lower().split(".")[-1] + mime_type = f"image/{extension}" + return f"data:{mime_type};base64,{base64_image}" + + def _is_json_serializable(obj: Any) -> bool: """Check if the given object is json serializable.""" try: diff --git a/tests/minimal.py b/tests/minimal.py index affad7cb7..cfdcb9981 100644 --- a/tests/minimal.py +++ b/tests/minimal.py @@ -26,6 +26,7 @@ { "model_type": "post_api_chat", "config_name": "my_post_api", + "model_name": "llama", "api_url": "https://xxx", "headers": {}, "json_args": {}, diff --git a/tests/model_anthropic_test.py b/tests/model_anthropic_test.py new file mode 100644 index 000000000..9c9ec44b8 --- /dev/null +++ b/tests/model_anthropic_test.py @@ -0,0 +1,91 @@ +# -*- coding: utf-8 -*- +"""Unittests for anthropic model.""" +import unittest +from unittest.mock import patch, MagicMock + +import agentscope +from agentscope.manager import ModelManager +from agentscope.models import ModelResponse + + +class AnthropicModelWrapperTest(unittest.TestCase): + """Anthropic model wrapper unittests.""" + + def setUp(self) -> None: + """Init for ExampleTest.""" + agentscope.init( + model_configs={ + "config_name": "claude-3-5", + "model_type": "anthropic_chat", + "model_name": "claude-3-5-sonnet-20241022", + "api_key": "xxx", + "stream": False, + }, + save_api_invoke=False, + ) + self.mock_response = { + "id": "msg_018zRB2VgEx2hS5TGxMhLz6Y", + "content": [ + { + "text": "你好！我是 Bob。今天天气不错，你觉得呢?", + "type": "text", + }, + ], + "model": "claude-3-5-sonnet-20241022", + "role": "assistant", + "stop_reason": "end_turn", + "stop_sequence": None, + "type": "message", + "usage": { + "cache_creation_input_tokens": 0, + "cache_read_input_tokens": 0, + "input_tokens": 21, + "output_tokens": 26, + }, + } + + self.model_response_gt = ModelResponse( + text="你好！我是 Bob。今天天气不错，你觉得呢?", + raw=self.mock_response, + ) + + @patch("anthropic.Anthropic") + def test_model_calling(self, mock_anthropic: MagicMock) -> None: + """Test if the model is called successfully.""" + mock_response = MagicMock() + mock_response.model_dump.return_value = { + "id": "msg_018zRB2VgEx2hS5TGxMhLz6Y", + "content": [ + { + "text": "你好！我是 Bob。今天天气不错，你觉得呢?", + "type": "text", + }, + ], + "model": "claude-3-5-sonnet-20241022", + "role": "assistant", + "stop_reason": "end_turn", + "stop_sequence": None, + "type": "message", + "usage": { + "cache_creation_input_tokens": 0, + "cache_read_input_tokens": 0, + "input_tokens": 21, + "output_tokens": 26, + }, + } + + mock_client = mock_anthropic.return_value + + mock_client.messages.create.return_value = mock_response + + model = ModelManager.get_instance().get_model_by_config_name( + "claude-3-5", + ) + response = model([{"role": "user", "content": "你好"}]) + + self.assertEqual(response.raw, self.model_response_gt.raw) + self.assertEqual(response.text, self.model_response_gt.text) + + +if __name__ == "__main__": + unittest.main() diff --git a/tests/model_test.py b/tests/model_test.py index 2ec1d8e14..dafc02c79 100644 --- a/tests/model_test.py +++ b/tests/model_test.py @@ -27,6 +27,7 @@ OpenAIDALLEWrapper, OpenAIChatWrapper, PostAPIChatWrapper, + AnthropicChatWrapper, ) @@ -42,6 +43,8 @@ def format( self, *args: Union[Msg, Sequence[Msg]], ) -> Union[List[dict], str]: + """Format the input for the model""" + print(*args) return "" @@ -75,6 +78,7 @@ def test_build_in_model_wrapper_classes(self) -> None: "zhipuai_embedding": ZhipuAIEmbeddingWrapper, "litellm_chat": LiteLLMChatWrapper, "yi_chat": YiChatWrapper, + "anthropic_chat": AnthropicChatWrapper, }, ) @@ -93,6 +97,7 @@ def test_load_model_configs(self, mock_logging: MagicMock) -> None: { "model_type": "post_api_chat", "config_name": "my_post_api", + "model_name": "llama", "api_url": "https://xxx", "headers": {}, "json_args": {},