Generate Profile Images (Amazon Bedrock)
Introduction
This workflow generates profile images from a text prompt using Amazon Bedrock foundational models.
Installation
Requirements
You must have an Amazon Web Services (AWS) account to use this workflow.
If you don’t have a connected account in Cerb for Amazon Web Services yet, you can follow these instructions to create one.
Amazon Bedrock
Log in to the AWS Console.
At the top of the page, search for Bedrock or navigate to Services » Machine Learning » Amazon Bedrock.
In the left sidebar, navigate to Bedrock configurations » Model access.
In a region of your choice, request access to at least one image generation model:
- Stability AI > Stable Image Core 1.0 (recommended)
- Amazon > Titan Image Generator G1 v2
Test image generation
In the Amazon Bedrock left sidebar, navigate to Playgrounds » Image.
Select one of your enabled image generation models.
At the bottom of the image playground, add an image generation prompt like:
A close-up profile picture of a female cyberpunk hacker in anime graphic novel style. She wears sleek, futuristic tech gear with digital implants and has a determined expression. Binary code and data streams flow visually into her head, symbolizing knowledge being downloaded directly into her brain. She is set against a high-tech background with holographic screens and computer interfaces. The color scheme features bright neon tones in blue, pink, and purple, capturing the cyberpunk aesthetic.
Click the orange Run button to the right of the text prompt at the bottom of the page.
In the top right of the playground, click the triple-dot menu and select View API Request.
This is the API request we’ll be using from Cerb automations.
Update your IAM policy
Navigate to IAM from the Services menu at the top of the page.
We need to update the policy in your Amazon Web Services (AWS) account to describe the services that your Cerb automations are allowed to use. This is accomplished with the Identity Access Management (IAM) service.
We’re going to add access to invoke AWS Bedrock models.
Select Policies in the navigation on the left.
Find your existing policy in the list or create a new one. In the earlier instructions we created a policy named CerbAutomationsPolicy.
Click on the policy and click the Edit button in the top right.
Click the Edit Policy button.
Select the JSON tab.
Add the following block to the Statement list:
{
"Sid": "CerbBedrockModels",
"Effect": "Allow",
"Action": "bedrock:InvokeModel",
"Resource": "*"
}
You can tighten up the Resource policy based on your needs once you confirm the integration is working.
Click the orange Next button in the bottom right.
Click the orange Save changes button in the bottom right.
Cerb
Install the workflow
This workflow is built into Cerb 11.0+. It will automatically update.
You can enable it from Search » Workflows » (+) » Generate Profile Images (Amazon Bedrock).
Link the workflow to the Amazon Web Services connected account you created above.
Usage
Use the Search menu in the top right to open the editor for any record with profile images: worker, group, organization, contact, bot, etc.
Click the Edit button to the right of the profile image.
Click the new AI Image Generation button.
Enter an image description like our earlier example:
A close-up profile picture of a female cyberpunk hacker in anime graphic novel style. She wears sleek, futuristic tech gear with digital implants and has a determined expression. Binary code and data streams flow visually into her head, symbolizing knowledge being downloaded directly into her brain. She is set against a high-tech background with holographic screens and computer interfaces. The color scheme features bright neon tones in blue, pink, and purple, capturing the cyberpunk aesthetic.
Click the blue Continue button.
You’ll be shown a preview of the generated image.
Click the blue Use this image button to keep it, or click Generate to generate a new image.
Reference
You can build your own Generate Profile Images workflow using this template as a reference.
Change occurrences of cerb.integrations.aws_bedrock.profile_images to your own workflow identifier. Use a prefix based on a domain you own (e.g. com.example.workflow).
workflow:
name: cerb.integrations.aws_bedrock.profile_images
version: 2024-10-25T00:00:00Z
description: Generate profile images from a text prompt using Amazon Bedrock foundational models
website: https://cerb.ai/workflows/cerb.integrations.aws_bedrock.profile_images/
requirements:
cerb_version: >=11.0 <11.1
cerb_plugins: cerberusweb.core
config:
chooser/aws_account_id:
label: AWS Account
record_type: connected_account
record_query: service:(type:aws)
multiple@bool: no
text/aws_region:
label: AWS Region
default: us-east-1
picklist/aws_model_id:
label: Bedrock Model
default: stability.stable-image-core-v1:0
options@list:
stability.stable-image-core-v1:0
amazon.titan-image-generator-v2:0
records:
automation/automation_invokeModel:
fields:
name: cerb.integrations.aws_bedrock.profile_images.invokeModel
extension_id: cerb.trigger.automation.function
description@text:
script@raw:
inputs:
text/prompt:
required@bool: yes
start:
set:
config@json: {{cerb_workflow_config('cerb.integrations.aws_bedrock.profile_images')|json_encode}}
decision/model:
outcome/stabilityCoreV1:
if@bool: {{'stability.stable-image-core-v1:0' == config.aws_model_id}}
then:
set:
invoke_body:
prompt: {{inputs.prompt}}
mode: text-to-image
aspect_ratio: 1:1
output_format: png
outcome/awsTitanV2:
if@bool: {{'amazon.titan-image-generator-v2:0' == config.aws_model_id}}
then:
set:
invoke_body:
textToImageParams:
text: {{inputs.prompt}}
taskType: TEXT_IMAGE
imageGenerationConfig:
cfgScale@int: 8
seed@int: 0
width@int: 512
height@int: 512
numberOfImages@int: 1
outcome/unsupported:
then:
return:
http.request/invoke:
output: http_response
inputs:
method: POST
url: https://bedrock-runtime.{{config.aws_region}}.amazonaws.com/model/{{config.aws_model_id}}/invoke
headers:
Accept: application/json
Content-Type: application/json
body@key: invoke_body
authentication: cerb:connected_account:{{config.aws_account_id}}
response:
resource:
expires@date: 15 mins
on_success:
outcome/ok:
if@bool: {{200 == http_response.status_code and 'application/vnd.cerb.uri' == http_response.content_type}}
then:
# Extract images from the JSON response
file.read:
output: json_file
inputs:
uri: {{http_response.body}}
length: 4096000
on_success:
set:
json_data@json: {{json_file.bytes}}
file.write:
output: image_file
inputs:
mime_type: image/png
expires@date: 2 hours
content@text: {{json_data.images|first|base64_decode}}
on_success:
var.unset:
inputs:
key@csv: json_data, json_file
return:
url: {{cerb_url('c=ui&a=image&token=' ~ image_file.uri|split(':')|last)}}
policy_kata@raw:
commands:
http.request:
deny/url@bool:
{{inputs.url is not pattern ('https://bedrock-runtime.*.amazonaws.com/model/*/invoke')}}
allow@bool: yes
file.read:
allow@bool: yes
file.write:
allow@bool: yes
automation/automation_interaction:
fields:
name: cerb.integrations.aws_bedrock.profile_images.interaction
extension_id: cerb.trigger.interaction.worker
description@text:
script@raw:
start:
while:
if@bool: {{'use' != prompt_action}}
do:
outcome/hasPrompt:
if@bool: {{prompt_text}}
then:
function/invoke:
uri: cerb:automation:cerb.integrations.aws_bedrock.profile_images.invokeModel
output: results
inputs:
prompt: {{prompt_text}}
await/prompt:
form:
title: Image Generation
elements:
sheet/prompt_image_url:
hidden@bool: {{results.url is empty}}
label: Preview:
data:
image:
url: {{results.url}}
limit: 5
schema:
layout:
headings@bool: no
paging@bool: no
style: grid
columns:
text/image:
params:
value_template@raw:
<img src="{{url}}" style="width:512px;height:auto;">
textarea/prompt_text:
label: Describe the image:
default: A profile picture of a tech worker
max_length@int: 2048
required@bool: yes
submit/prompt_action:
buttons:
continue/generate:
label: Generate
icon: magic
continue/use:
label: Use this image
hidden@bool: {{results.url is empty}}
icon: right-arrow
icon_at: end
style: secondary
value: use
return:
image:
url: {{results.url}}
policy_kata@raw:
commands:
function:
deny/uri@bool: {{uri != 'cerb:automation:cerb.integrations.aws_bedrock.profile_images.invokeModel'}}
allow@bool: yes
toolbar_section/toolbar_profile_image:
fields:
name: Image Generation (Amazon Bedrock)
toolbar_name: record.profile.image.editor
priority: 200
is_disabled: 0
toolbar_kata@raw:
interaction/bedrock:
label: AI Image Generation
uri: cerb:automation:cerb.integrations.aws_bedrock.profile_images.interaction
icon: magic