Demo Accessing Bedrock With Command Line Part 3

This lesson demonstrates how to access Amazon Bedrock from the command line and run inference using the AWS CLI. Examples show a Windows Subsystem for Linux (WSL) environment on Windows 11, but the same commands work in macOS and Linux terminals.

Verify the AWS CLI is available

Run aws to confirm the CLI is installed. A minimal invocation without arguments returns usage help and an error about missing subcommands:

alistair@H0DEI-LEGION5: ~/bedrock-demo$ aws
aws: [ERROR]: An error occurred (ParamValidation): the following arguments are required: command

usage: aws [options] <command> <subcommand> [<subcommand> ...] [parameters]
To see help text, you can run:

    aws help
    aws <command> help
    aws <command> <subcommand> help

A quick check that the CLI can call AWS APIs: ec2 describe-instances returns JSON describing your EC2 instances (truncated here):

{
  "Reservations": [
    {
      "ReservationId": "r-098a420a10e3ec1b7",
      "Instances": [
        {
          "Architecture": "x86_64",
          "BlockDeviceMappings": [
            {
              "DeviceName": "/dev/xvda",
              "Ebs": {
                "AttachTime": "2021-07-31T10:40:35+00:00",
                "DeleteOnTermination": true,
                "Status": "attached",
                "VolumeId": "vol-0a6cbedbeafa99260"
              }
            }
          ],
          "IamInstanceProfile": {
            "Arn": "arn:aws:iam::485186561655:instance-profile/AmazonSSMRoleForInstancesQuickSetup"
          }
        }
      ]
    }
  ]
}

Bedrock runtime: invoke-model vs converse

Two primary Bedrock runtime subcommands are useful from the CLI:

invoke-model — low-level call where you format the payload to match the model’s expected input.
converse — higher-level, message-based interface that standardizes chat-style inputs and is better for multi-turn conversations.

Use invoke-model when you need fine-grained control over exact tokenization, markers, or model-specific input formats. Use converse for portable, standardized messages and when building chat flows or multi-turn interactions.

Here is a quick comparison to help choose:

Subcommand	Best for	Example inputs	Returned metadata
`invoke-model`	Model-specific control, special tokens	`fileb://invoke-payload.json` (raw model prompt JSON)	`generation`, token counts, stop_reason
`converse`	Chat-style multi-turn, standardized message format	`--messages file://converse-payload.json`	Model `MESSAGE`, `METRICS`, usage info

Files used in the demo

Listing the working directory shows two payload files and an output file:

alistair@H0DEI-LEGION5:~/bedrock-demo$ ls -l
-rw-r--r-- 1 alistair alistair 152 May 13 15:47 converse-payload.json
-rw-r--r-- 1 alistair alistair 240 May 13 15:40 invoke-payload.json
-rw-r--r-- 1 alistair alistair 655 May 14 13:17 output.json

invoke-model example (low-level JSON payload)

invoke-payload.json contains a model-specific prompt and generation settings:

# invoke-payload.json
{
  "prompt": "<|begin_of_text|><|start_header_id|>user<|end_header_id|>\nWhat is Amazon Bedrock? Explain in 2 short bullets.<|eot_id|><|start_header_id|>assistant<|end_header_id|>",
  "max_gen_len": 200,
  "temperature": 0.3,
  "top_p": 0.9
}

Call the Bedrock runtime with invoke-model. Use --model-id for the catalog name, --body fileb://... to supply the binary JSON payload, and --accept application/json to request JSON output written to output.json:

alistair@H0DEI-LEGION5:~/bedrock-demo$ aws bedrock-runtime invoke-model \
  --region us-east-1 \
  --model-id us.meta.llama3-1-8b-instruct-v1:0 \
  --body fileb://invoke-payload.json \
  --content-type application/json \
  --accept application/json output.json
{
  "contentType": "application/json"
}

Inspect the generated JSON result:

# output.json
{
  "generation": "\nAmazon Bedrock is a cloud-based platform that provides a suite of tools for developers to build, deploy, and manage applications. Here are 2 short bullets explaining it:\n\n• **Managed Foundation Models**: Amazon Bedrock provides managed access to multiple foundation models and the runtime infrastructure to deploy them at scale.\n• **Developer Tooling and Integration**: Bedrock offers tools and APIs to simplify building, testing, and integrating generative AI into applications.",
  "prompt_token_count": 49,
  "generation_token_count": 108,
  "stop_reason": "stop"
}

converse example (higher-level, message-based JSON)

converse-payload.json expresses the user message using the standardized message array:

# converse-payload.json
[
  {
    "role": "user",
    "content": [
      {
        "text": "Explain what Amazon Bedrock is in two short bullet points."
      }
    ]
  }
]

Invoke the converse subcommand and request plain text output. The CLI returns the assistant message plus additional metrics and usage lines:

alistair@H0DEI-LEGION5:~/bedrock-demo$ aws bedrock-runtime converse \
  --region us-east-1 \
  --model-id us.meta.llama3-1-8b-instruct-v1:0 \
  --messages file://converse-payload.json \
  --output text
end_turn
METRICS 656
MESSAGE assistant
CONTENT

Here are two short bullet points explaining what Amazon Bedrock is:

• Amazon Bedrock is a cloud-based platform that gives developers managed access to powerful foundation models and the runtime infrastructure to deploy them.
• Amazon Bedrock simplifies building and integrating generative AI by providing APIs, developer tools, and scalable, secure model hosting.
USAGE 28 98 126

Note that converse includes METRICS and USAGE lines that are useful for logging, debugging, and cost monitoring.

Authentication and AWS CLI profiles

This guide shows examples using locally configured long-lived credentials. For production or shared environments, prefer temporary credentials using AWS SSO, IAM roles, or other secure methods.

Avoid committing long-lived AWS access keys into source control or sharing them. Use temporary credentials (for example, aws sso login, sts assume-role) or properly scoped IAM roles wherever possible.

List existing CLI profiles:

alistair@H0DEI-LEGION5:~/bedrock-demo$ aws configure list-profiles
default
PowerUserAccessPermissionSet-851189578280
new-empty-profile
current-console-user-profile

Create (or update) a named profile. You will be prompted for keys and defaults:

alistair@H0DEI-LEGION5:~/bedrock-demo$ aws configure --profile my-new-profile
Tip: You can deliver temporary credentials to the AWS CLI using AWS SSO or your console session; for example, run 'aws sso login' for SSO-based access.

AWS Access Key ID [None]: XXXXXXXXXXXXXX
AWS Secret Access Key [None]: XXXXXXXXXXXXXX
Default region name [None]: eu-west-1
Default output format [None]: json

Verify the profile was added:

alistair@H0DEI-LEGION5:~/bedrock-demo$ aws configure list-profiles
default
PowerUserAccessPermissionSet-851189578280
new-empty-profile
current-console-user-profile
my-new-profile

To use a specific profile when invoking Bedrock, include --profile my-new-profile in the AWS CLI command. This helps manage multiple credential sets, roles, or environments.

Summary

You can access Amazon Bedrock runtime endpoints directly from any terminal with the AWS CLI.
Choose invoke-model for precise, model-specific inputs and converse for standardized chat-style message flows and multi-turn interactions.
Both subcommands return useful metadata (metrics, usage, token counts) that help with logging and cost monitoring.
Manage credentials carefully: prefer temporary credentials and profiles; pass --profile to isolate credentials per task.

A "Results" slide showing four numbered dark-blue panels. Each panel lists a developer/operator capability: authenticate securely with the AWS CLI, access Amazon Bedrock from the terminal, invoke foundation models without the console, and quickly test prompts and model responses.

That concludes this lesson. A related short but important topic is inference profiles, which can affect how you interact with foundation models.

Introduction to Gen AI and LLMs

Introduction to Amazon Bedrock

Getting Started With Amazon Bedrock

Foundation Models in Amazon Bedrock

Basic Concepts of Prompt Engineering

Introduction to Text Generation With Amazon Bedrock

How Foundation Models Process Input

Extract Insights Using Prompt Engineering

Working With Different Task Types

Integrating Amazon Bedrock With Other AWS Services

Build a Marketing Email Generation Application

Introduction to Retrieval Augmentation With Bedrock Knowledge Bases

Best Practices and Optimization

Governance, Safety in Responsible AI

Security

Monitoring and Logging

Taking action with Bedrock Agents

Advanced Topics (Optional)

Final Project

Wrap Up

Demo Accessing Bedrock With Command Line Part 3

Verify the AWS CLI is available

Bedrock runtime: invoke-model vs converse

Files used in the demo

invoke-model example (low-level JSON payload)

converse example (higher-level, message-based JSON)

Authentication and AWS CLI profiles

Summary

Links and references

Watch Video

​Verify the AWS CLI is available

​Bedrock runtime: invoke-model vs converse

​Files used in the demo

​invoke-model example (low-level JSON payload)

​converse example (higher-level, message-based JSON)

​Authentication and AWS CLI profiles

​Summary

​Links and references

Watch Video

Verify the AWS CLI is available

Bedrock runtime: invoke-model vs converse

Files used in the demo

invoke-model example (low-level JSON payload)

converse example (higher-level, message-based JSON)

Authentication and AWS CLI profiles

Summary

Links and references