Bedrock

Overview

CloudMock emulates Amazon Bedrock, supporting foundation model listing, provisioned throughput management, model customization jobs, guardrails, model invocation, and tagging.

Supported Operations

Operation	Status	Notes
GetFoundationModel	Supported	Returns foundation model details
ListFoundationModels	Supported	Lists available foundation models
CreateProvisionedModelThroughput	Supported	Creates provisioned throughput
GetProvisionedModelThroughput	Supported	Returns provisioned throughput details
ListProvisionedModelThroughputs	Supported	Lists provisioned throughputs
UpdateProvisionedModelThroughput	Supported	Updates provisioned throughput
DeleteProvisionedModelThroughput	Supported	Deletes provisioned throughput
CreateModelCustomizationJob	Supported	Creates a model customization job
GetModelCustomizationJob	Supported	Returns customization job details
ListModelCustomizationJobs	Supported	Lists customization jobs
StopModelCustomizationJob	Supported	Stops a customization job
InvokeModel	Supported	Returns a stub model response
CreateGuardrail	Supported	Creates a guardrail
ApplyGuardrail	Supported	Applies a guardrail to content
TagResource	Supported	Adds tags to a resource
UntagResource	Supported	Removes tags from a resource
ListTagsForResource	Supported	Lists tags for a resource

Quick Start

Node.js

import { BedrockClient, ListFoundationModelsCommand } from '@aws-sdk/client-bedrock';

const client = new BedrockClient({
  endpoint: 'http://localhost:4566',
  region: 'us-east-1',
  credentials: { accessKeyId: 'test', secretAccessKey: 'test' },
});

const { modelSummaries } = await client.send(new ListFoundationModelsCommand({}));
console.log(modelSummaries);

Python

import boto3

client = boto3.client('bedrock',
    endpoint_url='http://localhost:4566',
    region_name='us-east-1',
    aws_access_key_id='test',
    aws_secret_access_key='test')

response = client.list_foundation_models()
print(response['modelSummaries'])

Configuration

cloudmock.yml

services:
  bedrock:
    enabled: true

Known Differences from AWS

• InvokeModel returns stub responses, not actual model inference results
• Foundation model list contains a predefined set of models
• Model customization jobs do not perform actual training
• Guardrail evaluation is simplified