Converse API
Amazon Bedrock's native unified API interface. It supports multi-turn conversations, tool calling, and streaming responses.
Timeline (last 90 days)
The Converse API is a unified conversation interface provided by Amazon Bedrock that shields differences in the underlying request body formats of models from different vendors and supports multi-turn conversations, tool calls, and system prompts; InvokeModel is a lower-level general inference interface whose request body must be constructed according to each model's proprietary schema.
For applications using the Converse API, switching models mainly costs changes to modelId and a few capability switches; applications using InvokeModel that directly assemble Anthropic native request bodies need to check the new version's field requirements.
Bedrock 提供两种调用 OpenAI 模型的接口：Converse API（原生接口，支持多轮对话、工具调用和流式响应）和 OpenAI 兼容 API（减少从 OpenAI 平台迁移的代码改动）。
Converse API 支持系统提示、温度 temperature、最大 token 数 maxTokens 等推理参数配置
Converse API 采用「角色-内容」消息模型，将每条消息标记为 user、assistant 或 system 三种角色，通过有序数组描述完整对话历史
在 Converse API 推出之前，开发者需要使用较底层的 InvokeModel API 来调用 Bedrock 上的模型，需自行构造每个模型供应商特定格式的请求 body
Converse API 在 2024 年中期推出，提供统一的跨模型对话调用接口，统一了请求和响应结构，并内置对工具调用和多模态输入的标准化支持
流式对话通过 ConverseStream 能力允许模型逐块（chunk）返回内容，显著缩短首字延迟（TTFT）
The Converse API provides a unified, cross-model conversational invocation interface, allowing developers to initiate conversation requests using the same message structure and parameter configuration.
AWS IAM adopts a "deny by default" zero-trust principle and can control access at the granularity of allowing a specific role to call only the Converse API of a particular model.
1 more timeline events
All Facts (11)
The Converse API is a unified conversation interface provided by Amazon Bedrock that shields differences in the underlying request body formats of models from different vendors and supports multi-turn conversations, tool calls, and system prompts; InvokeModel is a lower-level general inference interface whose request body must be constructed according to each model's proprietary schema.
50%UnverifiedFor applications using the Converse API, switching models mainly costs changes to modelId and a few capability switches; applications using InvokeModel that directly assemble Anthropic native request bodies need to check the new version's field requirements.
50%UnverifiedBedrock 提供两种调用 OpenAI 模型的接口：Converse API（原生接口，支持多轮对话、工具调用和流式响应）和 OpenAI 兼容 API（减少从 OpenAI 平台迁移的代码改动）。
50%UnverifiedConverse API 支持系统提示、温度 temperature、最大 token 数 maxTokens 等推理参数配置
50%UnverifiedConverse API 采用「角色-内容」消息模型，将每条消息标记为 user、assistant 或 system 三种角色，通过有序数组描述完整对话历史
50%Unverified在 Converse API 推出之前，开发者需要使用较底层的 InvokeModel API 来调用 Bedrock 上的模型，需自行构造每个模型供应商特定格式的请求 body
50%UnverifiedConverse API 在 2024 年中期推出，提供统一的跨模型对话调用接口，统一了请求和响应结构，并内置对工具调用和多模态输入的标准化支持
50%Unverified流式对话通过 ConverseStream 能力允许模型逐块（chunk）返回内容，显著缩短首字延迟（TTFT）
50%UnverifiedThe Converse API provides a unified, cross-model conversational invocation interface, allowing developers to initiate conversation requests using the same message structure and parameter configuration.
50%UnverifiedAWS IAM adopts a "deny by default" zero-trust principle and can control access at the granularity of allowing a specific role to call only the Converse API of a particular model.
50%UnverifiedThe Converse API supports inference parameter configuration such as system prompts, temperature, and maximum token count (maxTokens).
50%