MomentsInHistory is an automated, AI-powered application that generates and posts historical event content (both text and images) to Facebook.
Leveraging the power of Amazon Bedrock foundation models, the system uses advanced AI to craft engaging historical narratives and generate accompanying visuals dynamically. The application is fully serverless, running on AWS Lambda with the latest Python runtime, and is orchestrated using AWS EventBridge for scheduled, hands-free execution. Upon successful completion of each posting cycle, it sends an email notification using AWS SNS. All infrastructure components are provisioned and managed using Terraform for streamlined deployment and infrastructure consistency. Deployment and CI/CD workflows are automated via GitHub Actions, ensuring seamless and repeatable releases.
- 📌 Generates posts about historical events with accompanying images using Amazon Bedrock foundation models
- 📌 Automatically publishes generated content to Facebook
- 📌 Scheduled execution via AWS EventBridge (e.g., daily posts)
- 📌 Email notifications sent via AWS SNS upon completion
- 📌 Fully serverless using AWS Lambda with the latest Python runtime
- 📌 Infrastructure managed with Terraform for easy deployment and consistency
- AWS Lambda – Python-based serverless backend
- Amazon Bedrock – foundation models powering automatic generation of historical posts and images
- AWS EventBridge – scheduled execution of Lambda
- AWS SNS – email notifications after execution
- Terraform – infrastructure as code (IaC)
- AWS Services – S3, CloudWatch, IAM, etc.
- Facebook Graph API – posting content to Facebook
- AWS account and configured AWS CLI
- Terraform installed
- Python 3.10+ (for local development and Lambda packaging)
- Facebook API credentials (App ID, App Secret, Access Tokens)
Symptom:
Posts and photos sent via the Facebook Graph API are successfully created (no error response), but they do not show up in the fanpage's visible feed or timeline.
Symptom:
If there are no distinguishing traits in a prompt, LLM tends to create the same content all over, which is unexpected taking into account high temperature