AWS Bedrock: Easy 3 Step Image Generation App Guide

Hello Everyone! Today we will learn how to create an image generation Ai based application in just 3 simple steps.

Today we will be taking help of AWS Bedrock, Langchain, StreamLit and python based AWS Boto3 library to create the application.

Follow along and you will understand how easy to os to create such amazing application using the AWS Bedrock service.

So lets get started and learn some GenAI stuff.

Table of contents

Prerequisite

In this example we will using the AWS Bedrock service. We will be using the models provided by the StabilityAI Before getting stated with the example you must request access for the StabilityAI models.

Also since the bedrock service is not available in all the AWS regions we will be using the us-east-1 region for this example.

Action Items

1. Access for the StabilityAI models in the us-east-1 region.
2. Generate and configure access keys to execute the below python script. You will need the Bedrock service access.

Lets create the application

Step 1 - Create the Backend

Create an image_lib.py and paste the below code in it.

import boto3 #import aws sdk and supporting libraries
import json
import base64
from io import BytesIO

session = boto3.Session()

bedrock = session.client(service_name='bedrock-runtime') #creates a Bedrock client

bedrock_model_id = "stability.stable-diffusion-xl-v1" #use the Stable Diffusion model

def get_response_image_from_payload(response): #returns the image bytes from the model response payload

    payload = json.loads(response.get('body').read()) #load the response body into a json object
    images = payload.get('artifacts') #extract the image artifacts
    image_data = base64.b64decode(images[0].get('base64')) #decode image

    return BytesIO(image_data) #return a BytesIO object for client app consumption

def get_image_response(prompt_content): #text-to-text client function

    request_body = json.dumps({"text_prompts":
                               [ {"text": prompt_content } ], #prompts to use
                               "cfg_scale": 9, #how closely the model tries to match the prompt
                               "steps": 50, }) #number of diffusion steps to perform

    response = bedrock.invoke_model(body=request_body, modelId=bedrock_model_id) #call the Bedrock endpoint

    output = get_response_image_from_payload(response) #convert the response payload to a BytesIO object for the client to consume

    return output

Step 2 - Create the Frontend

We will be creating the UI of the application with the help of StreamLit library.

Create an image_app.py and paste the below code in it.

import streamlit as st #all streamlit commands will be available through the "st" alias
import image_lib as glib #reference to local lib script

st.set_page_config(layout="wide", page_title="Image Generation") #set the page width wider to accommodate columns

st.title("Image Generation") #page title

col1, col2 = st.columns(2) #create 2 columns

with col1: #everything in this with block will be placed in column 1
    st.subheader("Image generation prompt") #subhead for this column

    prompt_text = st.text_area("Prompt text", height=200, label_visibility="collapsed") #display a multiline text box with no label

    process_button = st.button("Run", type="primary") #display a primary button

with col2: #everything in this with block will be placed in column 2
    st.subheader("Result") #subhead for this column

    if process_button: #code in this if block will be run when the button is clicked
        with st.spinner("Drawing..."): #show a spinner while the code in this with block runs
            generated_image = glib.get_image_response(prompt_content=prompt_text) #call the model through the supporting library

        st.image(generated_image) #display the generated image

Step 3 - Run the application

Open the terminal and paste the below command. A server will be running at port 8080. You can access the application by visiting the url - http://localhost:8080

streamlit run image_app.py --server.port 8080

Conclusion

In conclusion, creating an image generation app with AWS Bedrock in just three simple steps is both efficient and powerful.

By integrating LangChain, streamlining processes with Boto3, and leveraging the flexibility of Python, you can harness the full potential of GenAI. This approach not only simplifies development but also ensures robust and scalable solutions for your image generation needs.

Start building today and experience the transformative power of these cutting-edge technologies.