
Weekend Creative Agent Challenge: Daily Wisdom Canvas
An always-on autonomous agent that generates daily inspirational visual quotes using Amazon Bedrock (Nova Lite + Titan Image Generator), delivered to your inbox every morning via a serverless pipeline built with Lambda, EventBridge Scheduler, S3, and SNS.
π¨ Vision: An Agent That Creates Beauty Every Morning
What if you could wake up every morning to a freshly crafted piece of art β a thought-provoking quote paired with stunning AI-generated imagery, delivered straight to your inbox before your first cup of coffee?
That's exactly what Daily Wisdom Canvas does. It's an always-on autonomous creative agent that runs every morning, generates an inspirational quote themed to the day or occasion, creates beautiful artistic imagery inspired by that quote, and delivers the complete creation to subscribers via email β all without any human intervention.
Last month, I participated in the July 2026 Productivity Challenge where I built the AI Study Planner using Amazon Bedrock Nova Lite and Streamlit. That project taught me how powerful generative AI can be for structured content creation. But it was interactive β it needed a human in the loop. This time, I wanted to push further: could I build something truly autonomous? An agent that thinks, creates, and delivers on its own schedule. The Weekend Creative Agent Challenge was the perfect opportunity to explore this.
π§ What It Does
Daily Wisdom Canvas is a fully autonomous creative pipeline:
- Thinks β Every morning, it uses Amazon Bedrock Nova Lite to generate a unique, thought-provoking quote. The prompt is context-aware: it considers the day of the week, the season, notable occasions, and even global awareness days to craft something timely and resonant.
- Creates β It then takes that quote and uses Amazon Bedrock Titan Image Generator G1 v2 to produce a beautiful piece of artistic imagery that visually captures the essence of the words. Think watercolor sunrises, abstract compositions, and serene landscapes β all generated fresh daily.
- Stores β The quote text, image, and metadata are stored in Amazon S3 with a clean, date-organized structure β building a growing gallery of daily wisdom over time.
- Delivers β Finally, it sends an email notification via Amazon SNS with the quote text and a link to the generated image, so subscribers start their day with inspiration.
The "problem" it solves might sound simple β daily inspiration. But the engineering behind it demonstrates a powerful pattern: event-driven autonomous agents that combine multiple AI modalities (text + image) into a cohesive creative workflow.
ποΈ Architecture Overview
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ Daily Wisdom Canvas ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€β ββ ββββββββββββββββββββ ββββββββββββββββββββββββββββ ββ β EventBridge β β AWS Lambda β ββ β Scheduler ββββββββββΆβ (Python 3.12) β ββ β (Daily 6 AM) β β β ββ ββββββββββββββββββββ β ββββββββββββββββββββββ β ββ β β 1. Generate Quote β β ββ β β (Nova Lite) β β ββ β ββββββββββββββββββββββ€ β ββ β β 2. Generate Image β β ββ β β (Titan Image) β β ββ β ββββββββββββββββββββββ€ β ββ β β 3. Store in S3 β β ββ β ββββββββββββββββββββββ€ β ββ β β 4. Notify via SNS β β ββ β ββββββββββββββββββββββ β ββ ββββββββββββ¬ββββββββββββββββ ββ β ββ βββββββββββββββββββββββΌβββββββββββββββ ββ β β β ββ βΌ βΌ βΌ ββ ββββββββββββββββββββ ββββββββββββββββββββ βββββββββββββββ ββ β Amazon Bedrock β β Amazon S3 β β Amazon SNS β ββ β β’ Nova Lite β β /wisdom/ β β (Email) β ββ β β’ Titan Image G1 β β YYYY/MM/DD/ β β β ββ ββββββββββββββββββββ ββββββββββββββββββββ βββββββββββββββ ββ ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββο»Ώ
AWS Services Used:
Service | Role |
|---|---|
AWS Lambda (Python 3.12) | Core agent logic β orchestrates the entire pipeline |
Amazon EventBridge Scheduler | Daily cron trigger at 6:00 AM IST |
Amazon Bedrock β Nova Lite | Generates contextual inspirational quotes |
Amazon Bedrock β Titan Image Generator G1 v2 | Creates artistic imagery from quote themes |
Amazon S3 | Persistent storage for quotes and images |
Amazon SNS | Email delivery to subscribers |
IAM | Least-privilege roles for Lambda execution |
π¨ How I Built It
Step 1: The Quote Generation Engine
The heart of the agent is the quote generation prompt. I spent time crafting a prompt that produces varied, meaningful output rather than generic platitudes:
import boto3import jsonfrom datetime import datetimeο»Ώ
bedrock = boto3.client('bedrock-runtime', region_name='us-east-1')ο»Ώ
def generate_quote(today): """Generate a contextual inspirational quote using Nova Lite.""" day_name = today.strftime('%A') month_name = today.strftime('%B') day_of_month = today.day prompt = f"""You are a creative philosopher and wordsmith. Today is {day_name}, {month_name} {day_of_month}.ο»Ώ
Generate ONE original, thought-provoking inspirational quote that:- Feels fresh and authentic (not a clichΓ©)- Is relevant to the energy of a {day_name}- Is between 15-30 words- Could stand alone as art on a wallο»Ώ
Also provide a brief visual description (2-3 sentences) of an artistic scene that captures the emotion of this quote. Think: watercolor, oil painting, abstract art, nature photography styles.ο»Ώ
Return JSON: {{"quote": "...", "visual_description": "...", "theme": "..."}}"""ο»Ώ
response = bedrock.invoke_model( modelId='amazon.nova-lite-v1:0', body=json.dumps({ "messages": [{"role": "user", "content": [{"text": prompt}]}], "inferenceConfig": {"maxTokens": 300, "temperature": 0.9} }) ) result = json.loads(response['body'].read()) content = result['output']['message']['content'][0]['text'] return json.loads(content)ο»Ώ
I intentionally set the temperature to 0.9 β I want creative variety, not predictable outputs. Each day should feel different.
Step 2: Image Generation
The visual description from Step 1 feeds directly into Titan Image Generator:
def generate_image(visual_description, theme): """Generate artistic imagery using Titan Image Generator G1 v2.""" image_prompt = ( f"Beautiful artistic illustration, {visual_description}. " f"Style: fine art, vibrant colors, emotional depth, " f"theme of {theme}. High quality, suitable for framing." ) response = bedrock.invoke_model( modelId='amazon.titan-image-generator-v2:0', body=json.dumps({ "textToImageParams": {"text": image_prompt}, "imageGenerationConfig": { "numberOfImages": 1, "height": 1024, "width": 1024, "cfgScale": 8.0 } }) ) result = json.loads(response['body'].read()) return base64.b64decode(result['images'][0])ο»Ώ
Step 3: Storage and Notification
The S3 structure keeps everything organized for easy retrieval and potential future use (like a web gallery):
def store_and_notify(today, quote_data, image_bytes): """Store creation in S3 and send SNS notification.""" s3 = boto3.client('s3') sns = boto3.client('sns') date_path = today.strftime('%Y/%m/%d') bucket = 'daily-wisdom-canvas' # Store the image image_key = f"wisdom/{date_path}/artwork.png" s3.put_object(Bucket=bucket, Key=image_key, Body=image_bytes, ContentType='image/png') # Store the quote metadata metadata_key = f"wisdom/{date_path}/metadata.json" s3.put_object(Bucket=bucket, Key=metadata_key, Body=json.dumps(quote_data, indent=2), ContentType='application/json') # Generate presigned URL for email image_url = s3.generate_presigned_url( 'get_object', Params={'Bucket': bucket, 'Key': image_key}, ExpiresIn=604800 # 7 days ) # Send notification message = ( f"π¨ Your Daily Wisdom Canvas\n\n" f"β¨ \"{quote_data['quote']}\"\n\n" f"πΌοΈ Today's artwork: {image_url}\n\n" f"Theme: {quote_data['theme']}\n" f"Have a wonderful {today.strftime('%A')}!" ) sns.publish( TopicArn='arn:aws:sns:us-east-1:ACCOUNT_ID:daily-wisdom-notifications', Subject=f"π
Daily Wisdom Canvas β {today.strftime('%B %d, %Y')}", Message=message )ο»Ώ
Step 4: The Lambda Handler
Everything ties together in a clean handler:
def lambda_handler(event, context): """Main entry point β triggered daily by EventBridge.""" today = datetime.now() print(f"π¨ Daily Wisdom Canvas β Generating for {today.strftime('%Y-%m-%d')}") # Step 1: Generate quote quote_data = generate_quote(today) print(f"π Quote: {quote_data['quote']}") # Step 2: Generate artwork image_bytes = generate_image(quote_data['visual_description'], quote_data['theme']) print(f"πΌοΈ Image generated: {len(image_bytes)} bytes") # Step 3: Store and notify store_and_notify(today, quote_data, image_bytes) print("β
Stored in S3 and notification sent!") return { 'statusCode': 200, 'body': json.dumps({ 'message': 'Daily wisdom generated successfully', 'quote': quote_data['quote'], 'theme': quote_data['theme'] }) }ο»Ώ
π§© Key Decisions & Challenges
Why Lambda over ECS/EC2? The entire pipeline executes in under 60 seconds. Lambda's pay-per-invocation model is perfect β this agent runs once per day, making it essentially free-tier friendly.
Challenge: Image prompt engineering. My first attempts produced generic stock-photo-style images. The breakthrough was chaining the outputs β having Nova Lite generate the visual description specifically for the image generator, rather than trying to craft image prompts directly from the quote. This "agent thinking about art direction" pattern dramatically improved quality.
Challenge: Lambda timeout. Bedrock image generation can take 15-20 seconds. Combined with text generation and S3 uploads, I needed to set the Lambda timeout to 90 seconds (default 3 seconds would fail).
IAM Least Privilege: The Lambda role has only the specific permissions it needs β Β
bedrock:InvokeModelΒ for the two specific model IDs, Β s3:PutObjectΒ scoped to the specific bucket/prefix, and Β sns:PublishΒ to the specific topic ARN.π What I Learned
- Multi-modal chaining is powerful. Using one AI model's output as creative direction for another creates results that feel more intentional than either model alone.
- Autonomous agents need observability. I added CloudWatch structured logging so I can review what the agent creates without checking email β this helped me iterate on prompt quality.
- EventBridge Scheduler > EventBridge Rules for this use case. The newer Scheduler service offers timezone support natively, which matters when you want the email at 6 AM IST specifically.
- Serverless is perfect for autonomous agents. Zero infrastructure management, zero cost when idle, automatic scaling if I ever want to generate multiple variants.
- From interactive to autonomous is a mindset shift. My previous AI Study Planner waited for user input. Building Daily Wisdom Canvas taught me to think about edge cases differently β what happens on holidays? What if Bedrock has a transient error? I added retry logic with exponential backoff.
π What's Next
I'm planning to add:
- A static S3 website that displays the growing gallery
- Weekly "best of" selections using Nova Lite as a curator
- Multiple artistic style variants per day
- Social media auto-posting integration
π Try It Yourself
The complete source code, CloudFormation template, and setup instructions are available here:
ο»Ώ
π Final Thoughts
The Weekend Creative Agent Challenge pushed me to think beyond chatbots and into truly autonomous AI systems. Daily Wisdom Canvas might seem like a simple project, but it demonstrates a powerful pattern: event-driven, multi-modal, autonomous agents that create value without human intervention.
From my AI Study Planner last month to this autonomous creative agent, my journey with Amazon Bedrock has been incredibly rewarding. The combination of Nova Lite's creative text generation and Titan Image Generator's visual artistry opens up possibilities I'm only beginning to explore.
Every morning now, I wake up to a new piece of AI-generated art with words that make me pause and think. That's the magic of building agents that create while you sleep. β¨
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article