Chat with an AWS bedrock model

AWS Bedrock provides a number of language models, including those from Anthropic's Claude, using the Bedrock Converse API.

Authentication

Authentication is handled through {paws.common}, so if authentication does not work for you automatically, you'll need to follow the advice at https://www.paws-r-sdk.com/#credentials. In particular, if your org uses AWS SSO, you'll need to run aws sso login at the terminal.

Prompt caching

Bedrock supports prompt caching via cache checkpoints. When caching is enabled, ellmer places cache checkpoints on the system prompt and the last turn, so that the conversation history is cached across turns.

By default (cache = "auto"), caching is enabled for models known to support it (Anthropic Claude and Amazon Nova) and disabled for all other models. You can also set cache to "5m" or "1h" to force a specific TTL, or "none" to disable caching entirely. Note that individual models may have minimum input token thresholds before caching takes effect.

Note that token_usage() does not currently reflect the cost of writing to the cache, which is priced at a premium over regular input tokens. Cache read savings are reported correctly.

Usage

chat_aws_bedrock(
  system_prompt = NULL,
  base_url = NULL,
  model = NULL,
  profile = NULL,
  cache = c("auto", "5m", "1h", "none"),
  params = NULL,
  api_args = list(),
  api_headers = character(),
  echo = NULL
)

models_aws_bedrock(profile = NULL, base_url = NULL)

Arguments

system_prompt

A system prompt to set the behavior of the assistant.

base_url

The base URL to the API endpoint.

model

The model to use for the chat (defaults to "anthropic.claude-sonnet-4-5-20250929-v1:0"). We regularly update the default, so we strongly recommend explicitly specifying a model for anything other than casual use. Use models_models_aws_bedrock() to see all options. .

While ellmer provides a default model, there's no guarantee that you'll have access to it, so you'll need to specify a model that you can. If you're using cross-region inference, you'll need to use the inference profile ID, e.g. model="us.anthropic.claude-sonnet-4-5-20250929-v1:0".

profile

AWS profile to use.

cache

How long to cache inputs? The default, "auto", enables caching with a 5-minute TTL for models known to support it (Anthropic Claude and Amazon Nova) and disables caching for all other models. Set to "5m" or "1h" to force caching on, or "none" to disable it.

See details below.

params

Common model parameters, usually created by params().

api_args

Named list of arbitrary extra arguments appended to the body of every chat API call. Use params for common parameters. Model-specific inference parameters can be provided using the additionalModelRequestFields field, for example to enable thinking effort in Anthropic Claude models:

api_args = list(
  additionalModelRequestFields = list(
    thinking = list(type = "enabled", budget_tokens = 4000)
  )
)

See https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference-call.html for more details.

api_headers

Named character vector of arbitrary extra headers appended to every chat API call.

echo

One of the following options:

none: don't emit any output (default when running in a function).
output: echo text and tool-calling output as it streams in (default when running at the console).
all: echo all input and output.

Note this only affects the chat() method.

Value

A Chat object.

Examples

if (FALSE) { # \dontrun{
# Basic usage
chat <- chat_aws_bedrock()
chat$chat("Tell me three jokes about statisticians")
} # }