Changelog

Agent and API ship from the same repository and share a version number. Newest first.

0.9.7 — 2026-08-24

Incident thresholds can now be set per host, overriding the metric default.
Fixed a rate-limit counter reset bug on POST /v1/ingest around leap-second handling.

0.9.6 — 2026-06-19

GET /v1/incidents supports state=acknowledged.
Agent retry backoff capped at 60s, was previously unbounded.

0.9.5 — 2026-05-01

Ingest batches accept up to 500 metrics per request, was 100.
Dashboard: added a 90-day uptime view per component.

0.9.4 — 2026-03-11

Fixed host tags containing commas being split incorrectly during registration.
Added mem.used_pct to the default collect.metrics list.

0.9.3 — 2026-01-22

POST /v1/incidents/{id}/ack now returns acknowledged_at.
Removed deprecated GET /v1/ping; use GET /v1/hosts instead.

0.9.2 — 2025-11-30

Agent binary published for arm64.
Config validation on startup now rejects unknown keys instead of ignoring them.

0.9.1 — 2025-10-08

429 responses on /v1/ingest now include a Retry-After header.
Fixed clock-skew handling for collected_at timestamps more than 5 minutes in the future.

0.9.0 — 2025-08-19

First stable release of the ingest batching API.
Breaking: request field values renamed to metrics.

0.8.3 — 2025-06-25

GET /v1/hosts now includes host tags in the response.
Notification webhook retries switched to exponential backoff.

0.8.2 — 2025-05-02

Fixed a memory leak in the metrics query cache under sustained load.

0.8.1 — 2025-03-14

GET /v1/metrics gained since / until range parameters.

0.8.0 — 2024-11-20

Initial public API: GET /v1/hosts, POST /v1/ingest.
Agent gained a collect.interval option; previously fixed at 60s.

0.7.0 — 2024-09-15

First tagged release. Single-host ingest only, no batching, no incidents yet — metrics were push-only with no read API.