Added DynamoDB to the extraction service for storing metadata. OpenSearch is now used solely for vector-related logic to reduce costs. Users can run the extraction service without OpenSearch.
Smart sampling now utilizes an in-memory vector DB (FAISS) to eliminate the OpenSearch dependency.
The evaluation service now supports storing historical requests in the database.
Included scene detection to the extraction workflow.
Additional configuration options exposed, allowing users to customize extraction settings via the UI and API, such as similarity thresholds, Amazon Rekognition API confidence thresholds, etc.
Bug fixes.
Folder structure following the AWS solution guidance standard.
Changes: