Three datasets, three formats (CSV, JSON, JSONL). Filter by time range, by node, and by measurement source. The downloads stream straight from the SQLite store — even the 90-day raw export holds only one chunk in server memory at a time.

License: data is freely usable under the same MIT terms as the source code. Attribution to @steem-api-health is appreciated. If you build something on it, leave a comment under any daily report — we'll link back.

One row per polled tick: timestamp, node URL, success flag, latency, head-block height, error category, source location.

Pre-bucketed counts and average latencies — one row per (bucket, node). Use these when you don't need per-tick detail and want a much smaller download.

Each row: a contiguous run of failed ticks. Severity real means at least 2 minutes; short is a transient hiccup.

Every download here is a plain HTTP GET — automate it with curl, a daily cron, or a notebook. The complete endpoint reference lives at docs/API.md.

• GET /api/v1/export/measurements — raw rows
• GET /api/v1/export/aggregates — bucketed totals
• GET /api/v1/export/outages.csv / .json — outage log
• GET /api/v1/export/sources — list of source labels for the dropdown