Detect Qdrant embedding drift and verify bounded re-embedding
Compare live embeddings with stored vectors and repair an approved batch only when held-out retrieval has degraded.
Lightweight cohort retention analytics in Kestra using in-process DuckDB SQL on input CSV files, featuring threshold evaluation without Python or external webhook requirements.
Automated cohort retention pipeline that computes monthly signup-to-order user retention metrics directly from user-uploaded event files using pure DuckDB SQL.
Unlike cohort-retention-duckdb (which requires Python scripting, Parquet datasets, and Slack webhooks),
this blueprint offers a zero-dependency, pure-SQL alternative that teams can drop in immediately
with standard CSV event logs.
user_id, event_name, signup_date, event_date).read_csv_auto() reads and parses the event file in-memory.If conditional gate checks whether the cohort retention meets the target threshold (>= 50%),
routing to success logs or alert notifications.user_id, event_name, signup_date, and event_date.read_csv_auto() directly to S3 or Google Cloud Storage buckets.