Can I vibe code HighTouch?

Build a lightweight Reverse-ETL sync worker in Python using FastAPI, SQLAlchemy, and Celery.

1. DATA MODEL:
- Connections: stores database credentials (Postgres, BigQuery, Snowflake) using encrypted strings.
- Destinations: stores target API settings (Klaviyo API key, Meta CAPI token).
- Syncs: ties a SQL Query to a Destination, storing field mappings (SQL column -> API parameter) and schedule (cron string).
- SyncRuns: stores execution metadata (status, rows_processed, rows_failed, log_output, started_at, completed_at).
- RecordState: stores primary key and record hash (MD5) from the previous run to detect diffs.

2. CORE FUNCTIONALITY:
- Execute SQL query against target Connection.
- Compute MD5 hash of each record's mapped values.
- Compare against RecordState to filter only added or modified rows.
- Batch modified rows into groups of 100.
- Format batches according to destination API schema (e.g. Klaviyo profile batch update payload format) and issue POST requests.
- Update RecordState with new hashes upon successful HTTP response.

3. ERROR & RATE LIMIT HANDLING:
- Intercept HTTP 429 and 5xx response status codes.
- Apply exponential backoff with random jitter using Celery retry mechanisms.
- Log specific row failures to SyncRuns table while allowing valid rows in the batch to complete.

4. OUT OF SCOPE:
- Visual drag-and-drop segment builders.
- Multi-tenant enterprise SSO / RBAC.
- Generic visual UI for schema mapping; use simple JSON schema definitions in the DB.

$ each button prefixes agent-specific run instructions · build your own product, never copy proprietary code, trademarks or designs