max_chunk_size: (Optional, integer) Specifies the maximum size of a chunk in words. Defaults to 250. This value cannot be higher than 300 or lower than 20 (for sentence strategy) or 10 (for word strategy).
overlap: (Optional, integer) Only for word chunking strategy. Specifies the number of overlapping words for chunks. Defaults to 100. This value cannot be higher than the half of max_chunk_size.
sentence_overlap: (Optional, integer) Only for sentence chunking strategy. Specifies the numnber of overlapping sentences for chunks. It can be either 1 or 0. Defaults to 1.
strategy: (Optional, string) Specifies the chunking strategy. It could be either sentence or word.

service

(Required, string) The type of service supported for the specified task type. In this case, amazonbedrock.

service_settings

(Required, object) Settings used to install the inference model.

These settings are specific to the amazonbedrock service.

access_key: (Required, string) A valid AWS access key that has permissions to use Amazon Bedrock and access to models for inference requests.
secret_key: (Required, string) A valid AWS secret key that is paired with the access_key. To create or manage access and secret keys, see Managing access keys for IAM users in the AWS documentation.

You need to provide the access and secret keys only once, during the inference model creation. The Get inference API does not retrieve your access or secret keys. After creating the inference model, you cannot change the associated key pairs. If you want to use a different access and secret key pair, delete the inference model and recreate it with the same name and the updated keys.

provider

(Required, string) The model provider for your deployment. Note that some providers may support only certain task types. Supported providers include:

amazontitan - available for text_embedding and completion task types
anthropic - available for completion task type only
ai21labs - available for completion task type only
cohere - available for text_embedding and completion task types
meta - available for completion task type only
mistral - available for completion task type only

model

(Required, string) The base model ID or an ARN to a custom model based on a foundational model. The base model IDs can be found in the Amazon Bedrock model IDs documentation. Note that the model ID must be available for the provider chosen, and your IAM user must have access to the model.

region

(Required, string) The region that your model or ARN is deployed in. The list of available regions per model can be found in the Model support by AWS region documentation.

rate_limit

(Optional, object) By default, the amazonbedrock service sets the number of requests allowed per minute to 240. This helps to minimize the number of rate limit errors returned from Amazon Bedrock. To modify this, set the requests_per_minute setting of this object in your service settings:

"rate_limit": {
    "requests_per_minute": <<number_of_requests>>
}

task_settings: (Optional, object) Settings to configure the inference task. These settings are specific to the <task_type> you specified.

task_settings for the completion task type

max_new_tokens: (Optional, integer) Sets the maximum number for the output tokens to be generated. Defaults to 64.
temperature: (Optional, float) A number between 0.0 and 1.0 that controls the apparent creativity of the results. At temperature 0.0 the model is most deterministic, at temperature 1.0 most random. Should not be used if top_p or top_k is specified.
top_p: (Optional, float) Alternative to temperature. A number in the range of 0.0 to 1.0, to eliminate low-probability tokens. Top-p uses nucleus sampling to select top tokens whose sum of likelihoods does not exceed a certain value, ensuring both variety and coherence. Should not be used if temperature is specified.
top_k: (Optional, float) Only available for anthropic, cohere, and mistral providers. Alternative to temperature. Limits samples to the top-K most likely words, balancing coherence and variability. Should not be used if temperature is specified.

Amazon Bedrock service example

edit

The following example shows how to create an inference endpoint called amazon_bedrock_embeddings to perform a text_embedding task type.

Choose chat completion and embeddings models that you have access to from the Amazon Bedrock base models.

resp = client.inference.put(
    task_type="text_embedding",
    inference_id="amazon_bedrock_embeddings",
    inference_config={
        "service": "amazonbedrock",
        "service_settings": {
            "access_key": "<aws_access_key>",
            "secret_key": "<aws_secret_key>",
            "region": "us-east-1",
            "provider": "amazontitan",
            "model": "amazon.titan-embed-text-v2:0"
        }
    },
)
print(resp)

const response = await client.inference.put({
  task_type: "text_embedding",
  inference_id: "amazon_bedrock_embeddings",
  inference_config: {
    service: "amazonbedrock",
    service_settings: {
      access_key: "<aws_access_key>",
      secret_key: "<aws_secret_key>",
      region: "us-east-1",
      provider: "amazontitan",
      model: "amazon.titan-embed-text-v2:0",
    },
  },
});
console.log(response);

PUT _inference/text_embedding/amazon_bedrock_embeddings
{
    "service": "amazonbedrock",
    "service_settings": {
        "access_key": "<aws_access_key>",
        "secret_key": "<aws_secret_key>",
        "region": "us-east-1",
        "provider": "amazontitan",
        "model": "amazon.titan-embed-text-v2:0"
    }
}

The next example shows how to create an inference endpoint called amazon_bedrock_completion to perform a completion task type.

resp = client.inference.put(
    task_type="completion",
    inference_id="amazon_bedrock_completion",
    inference_config={
        "service": "amazonbedrock",
        "service_settings": {
            "access_key": "<aws_access_key>",
            "secret_key": "<aws_secret_key>",
            "region": "us-east-1",
            "provider": "amazontitan",
            "model": "amazon.titan-text-premier-v1:0"
        }
    },
)
print(resp)

const response = await client.inference.put({
  task_type: "completion",
  inference_id: "amazon_bedrock_completion",
  inference_config: {
    service: "amazonbedrock",
    service_settings: {
      access_key: "<aws_access_key>",
      secret_key: "<aws_secret_key>",
      region: "us-east-1",
      provider: "amazontitan",
      model: "amazon.titan-text-premier-v1:0",
    },
  },
});
console.log(response);

PUT _inference/completion/amazon_bedrock_completion
{
    "service": "amazonbedrock",
    "service_settings": {
        "access_key": "<aws_access_key>",
        "secret_key": "<aws_secret_key>",
        "region": "us-east-1",
        "provider": "amazontitan",
        "model": "amazon.titan-text-premier-v1:0"
    }
}

« AlibabaCloud AI Search inference integration Anthropic inference integration »