Changelog

Every narrative now carries a confidence score on its root-cause claim. High-confidence incidents can suggest a concrete action; lower-confidence ones say so plainly instead of guessing.

Fix-forward PRs. When a rollback isn't the right move, Inkvo can draft a forward fix as a pull request against your repo — you review and merge. Opt in per service.

Deploy correlation now reads deployment.environment from OTel resource attributes, so a staging deploy never gets blamed for a prod incident.

Resolved a case where simultaneous deploys to two services produced two separate narratives for one root cause. They now merge into a single incident.

Tone per environment. A prod outage reads like war; a staging hiccup reads like a heads-up; dev stays quiet. Severity thresholds and routing follow the tone automatically.

Custom environment remapping — point preview or canary at the tone that fits, in Settings → Detection.

Baseline learning window cut from 48h to 24h for services with steady traffic.

Slack threads occasionally duplicated the first narrative message under high event volume. Deduped.

Added 12 connectors across cloud & infra and source/deploys — including GitLab CI, CircleCI, and Cloudflare Workers logs. Browse them on the integrations page.

Opsgenie paging alongside PagerDuty, with on-call rotation sync.

OTLP ingest now accepts gzip-compressed payloads up to 4 MiB, halving egress for high-cardinality services.

Post-mortem Markdown export includes the full evidence timeline inline.

Mark-as-noise feedback. Flagging an incident as not-worth-paging feeds the per-service baseline, so the next one is quieter.

Seasonality modeling — a 9am traffic ramp is no longer treated as an anomaly. Weekly patterns are learned too.

Throughput-drop detection misfired during scheduled maintenance windows. Maintenance windows are now respected.

Narrative incidents. When something breaks, you get a story — what broke, what's affected, what changed, and what to do — instead of a wall of graphs.

OpenTelemetry-native ingest over HTTP and gRPC. One endpoint, one key per environment.

Slack, email, and PagerDuty delivery out of the box.