AWS bedrock documentation change
Summary
Restructured endpoint documentation to focus on inference endpoints, removed control plane details, added comprehensive comparison tables between bedrock-mantle and bedrock-runtime endpoints, and provided updated usage recommendations.
Security assessment
The changes focus on endpoint functionality, API support, and feature comparisons without addressing vulnerabilities or security mechanisms. The authentication references (SigV4 and API keys) are existing features mentioned in operational comparisons, not new security documentation.
Diff
diff --git a/bedrock/latest/userguide/endpoints.md b/bedrock/latest/userguide/endpoints.md index 33a9d10bc..aae14bb3b 100644 --- a//bedrock/latest/userguide/endpoints.md +++ b//bedrock/latest/userguide/endpoints.md @@ -9,5 +9 @@ -Amazon Bedrock supports various endpoints depending on whether you want to perform control plane operators or [inference](./inference.html) operations. - -**Control Plane operations** - -Amazon Bedrock control plane operations can be used with the endpoint: `bedrock.{region}.amazonaws.com.` It can be used for managing resources, such as listing available [models](./models.html), creating [custom model](./custom-models.html) jobs, and managing [provisioned throughput](./prov-throughput.html). Read more on our control plane API [here](https://docs.aws.amazon.com/general/latest/gr/bedrock.html). +Amazon Bedrock supports various endpoints for performing [inference](./inference.html) operations. @@ -21,2 +17,69 @@ Amazon Bedrock supports the following primary two end points for performing infe -`bedrock-mantle.{region}.api.aws` | [Responses API](./bedrock-mantle.html) / [Chat Completions API](./bedrock-mantle.html) / [Messages API](./model-parameters-anthropic-claude-messages.html) | Region-specific endpoints for making inference requests for models hosted in Amazon Bedrock using the OpenAI-compatible endpoints and the Anthropic Messages API. -`bedrock-runtime.{region}.amazonaws.com` | [InvokeModel](./inference-invoke.html) / [Converse](./conversation-inference.html) / [Chat Completions](./inference-chat-completions.html) / [Messages API](./model-parameters-anthropic-claude-messages.html) | Region-specific endpoints for making inference requests for models hosted in Amazon Bedrock using the InvokeModel/Converse/Chat Completions/Messages APIs. Read more on Amazon Bedrock Runtime APIs [here](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_Operations_Amazon_Bedrock_Runtime.html). +`bedrock-mantle.{region}.api.aws` | [Responses API](./bedrock-mantle.html) / [Chat Completions API](./inference-chat-completions-mantle.html) / [Messages API](./inference-messages-api.html) | Region-specific endpoints for making inference requests for models hosted in Amazon Bedrock using the OpenAI-compatible endpoints and the Anthropic Messages API. +`bedrock-runtime.{region}.amazonaws.com` | [InvokeModel](./inference-invoke.html) / [Converse](./conversation-inference.html) / [Chat Completions](./inference-chat-completions.html) / [Messages API](./inference-messages-api.html) | Region-specific endpoints for making inference requests for models hosted in Amazon Bedrock using the InvokeModel/Converse/Chat Completions/Messages APIs. Read more on Amazon Bedrock Runtime APIs [here](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_Operations_Amazon_Bedrock_Runtime.html). + +For new applications, we recommend the `bedrock-mantle` endpoint. It supports OpenAI-compatible APIs (Responses and Chat Completions) and the Anthropic Messages API, includes built-in tool use and stateful conversation management, and lets you bring an existing OpenAI SDK codebase to Amazon Bedrock by changing only the base URL and API key. The `bedrock-runtime` endpoint remains fully supported and is the right choice when you're using the Bedrock-native InvokeModel or Converse APIs, or when the model you want isn't yet available on `bedrock-mantle`. To see which endpoint each model supports, see [Endpoint availability](./models-endpoint-availability.html). + +The following tables compare what's available on each endpoint. + +API support API | `bedrock-runtime` | `bedrock-mantle` +---|---|--- +[InvokeModel](./inference-api.html) |  |  +[Converse / ConverseStream](./conversation-inference.html) |  |  +[Chat Completions (OpenAI-compatible)](./inference-chat-completions-mantle.html) |  |  +[Responses API (OpenAI-compatible)](./bedrock-mantle.html#bedrock-mantle-responses) |  |  +[Messages API (Anthropic-native)](./inference-messages-api.html) |  |  + +Inference capabilities Capability | `bedrock-runtime` | `bedrock-mantle` +---|---|--- +[Cross-region inference (geographic and global profiles)](./cross-region-inference.html) |  |  +[Stateful conversation management](./bedrock-mantle.html#bedrock-mantle-responses-state) |  |  +[Asynchronous (long-running) inference](./bedrock-mantle.html) |  |  +[Client-side tool use](./tool-use-client-side.html) |  |  +[Server-side tool use](./tool-use-server-side.html) |  |  +[Pre-configured ready-to-use tools](./tool-use.html) |  |  +[Projects](./projects.html) |  |  +[Workspaces](./workspaces.html) |  |  + +Operational Item | `bedrock-runtime` | `bedrock-mantle` +---|---|--- +AWS [SigV4](https://docs.aws.amazon.com/AmazonS3/latest/API/sig-v4-authenticating-requests.html) authentication |  |  +[Bedrock API key (also works with OpenAI SDK)](./api-keys.html) |  |  +[Usage attribution](./cost-management.html) | [IAM](./cost-mgmt-iam-principal-tracking.html), [per-request metadata tagging](./cost-mgmt-request-metadata.html) | [Projects](./projects.html), [Workspaces](./workspaces.html) + +**Throughput and quota approach** + +Each endpoint uses a different approach to managing throughput. + + * **`bedrock-runtime`** – In many traditional multi-tenant services, the architecture is designed around per-account quotas to manage fair-share access to shared resources. This is the approach used with `bedrock-runtime`. Each model has fixed throughput quotas (RPM and TPM) that you can request increases for. For details, see [Quotas for the bedrock-runtime endpoint](./quotas-runtime.html). + + * **`bedrock-mantle`** – This endpoint is architected with advanced scheduling and work-queueing mechanisms that deliver fair-share distribution while supporting higher initial throughput limits. This design also allows `bedrock-mantle` to host a broad set of models and deliver the full breadth of capabilities available across the model catalog. In most cases, requests are served immediately. In some cases, a request may be briefly queued while in-flight workloads complete and throughput becomes available. For details, see [Quotas for the bedrock-mantle endpoint](./quotas-mantle.html) and [Scaling and throughput best practices](./scaling-throughput-best-practices.html). + + + + +**Pricing** + +Per-token pricing for the same model is identical on `bedrock-runtime` and `bedrock-mantle`. Choose an endpoint based on the APIs and capabilities you need, not cost. For current pricing, see [Amazon Bedrock pricing](https://aws.amazon.com/bedrock/pricing/). + +**When to choose each endpoint** + +Start with `bedrock-mantle` when you want to: + + * Use the Responses API, Chat Completions API, or Messages API with stateful, multi-turn conversations. + + * Bring existing OpenAI SDK code to Amazon Bedrock by changing only the base URL and API key. + + * Run asynchronous or long-running inference workloads. + + * Build agentic workflows with server-side tool use or pre-configured tools. + + * Use [Projects (OpenAI-compatible)](./projects.html) or [Workspaces (Anthropic-compatible)](./workspaces.html) to isolate workloads and track cost and usage at the application level. + + + + +Use `bedrock-runtime` when you want to: + + * Continue using the Bedrock-native [InvokeModel](./inference-api.html) or [Converse](./conversation-inference.html) APIs. + + * Use a model that isn't yet available on `bedrock-mantle`. See [Endpoint availability](./models-endpoint-availability.html). @@ -24 +86,0 @@ Amazon Bedrock supports the following primary two end points for performing infe -**Which service endpoint should you use for your inference?** @@ -26 +87,0 @@ Amazon Bedrock supports the following primary two end points for performing infe -The endpoint you use depends on your use-case. @@ -28,4 +88,0 @@ The endpoint you use depends on your use-case. -**Endpoint** | **Features and Use Cases** ----|--- -`bedrock-mantle` | Supports OpenAI-compatible APIs ([Responses API](./bedrock-mantle.html), [Chat Completions API](./bedrock-mantle.html)) and the Anthropic-native [Messages API](./model-parameters-anthropic-claude-messages.html). **Use for:** Easily migrating from other models that use the OpenAI-compatible APIs. **Features:** Supports both client and server side tool use with Lambda functions and comes configured with ready-to-use tools in your applications. **Recommended for:** New users to Amazon Bedrock. -`bedrock-runtime` | Supports Amazon Bedrock native APIs ([Invoke API](./inference-invoke.html), [Converse API](./conversation-inference.html)), the Anthropic-native [Messages API](./model-parameters-anthropic-claude-messages.html), and the OpenAI-compatible [Chat Completions API](./inference-chat-completions.html). **Use for:** Running any model supported by Amazon Bedrock. The [Converse API](./conversation-inference.html) provides a unified interface for interacting with all models in Amazon Bedrock and the [Invoke API](./inference-invoke.html) provides direct access to models with more ability to control the request and response format. You can also use [Chat Completions API](./inference-chat-completions.html) to interact with your models. **Features:** Supports only client-side tool use with Lambda functions and does not come pre-configured with ready-to-use tools. Allows you to track usage and costs when invoking a model. @@ -33 +90 @@ The endpoint you use depends on your use-case. -If you are new to Amazon Bedrock, we recommend using open APIs such as the Messages API, Chat Completions API, or Responses API. These APIs are available on both endpoints, but we recommend the `bedrock-mantle` endpoint. +Both endpoints can be used together from the same application — choose per use case.