The problem
Each integration failed silently in its own logs. Nobody had one view of health across EDI partners, the WMS sync and customer APIs.
What I built
- One ingest endpoint that every integration posts to, secured with a bearer token.
- Failure folding. Repeated failures collapse into one row with a count and first/last seen, and metrics count occurrences so success rates stay honest.
- Health rules per integration and subtype, with a heartbeat/staleness check. Suggested thresholds come from each feed’s volume tier and its longest observed quiet gap.
- KPI and failure-trend charts, single and bulk conflict resolution, and automatic discovery of new integrations.