What's New

Pricing model update · Faster everywhere · Platform install guides

• New

  Pricing model: Trial · Team · Growth · Enterprise

  Replaced the free-forever tier with a 14-day full-feature trial — every detector, full AI root cause, no credit card. Team is now $500/mo flat for 1.5 TB/month and 30-day retention. New Growth tier at $1,500/mo for 5 TB/month, 30-day retention, SSO, audit log, and priority support. Enterprise from $5,000/mo with PrivateLink, SAML, custom retention, and dedicated support. AI root cause is included on every tier.

• Improved

  Edge caching for every read-heavy surface

  The app, demo, and read-only API endpoints (alerts, incidents, streams, service health, SLO burn-down, log histograms) now serve from the nearest CDN edge. Cold-load TTFB for visitors outside our origin region dropped from ~1 s to ~150 ms. Dashboards feel snappier on every repeat refresh.

• New

  Five platform-specific install guides

  New step-by-step setup walkthroughs at /docs/install for AWS, Google Cloud, Kubernetes, Railway, and Vercel. Each is copy-paste ready: prerequisites, the exact configuration, verification steps, and the gotchas we've seen on real deployments. 3–10 minutes to first log on every platform.

• New

  Log Processing: field-value rules

  When your log shipper writes the level under a non-standard field — `severity` (GCP, K8s events), `loglevel` (Bunyan), `priority` (syslog), `status_code` for 5xx surges — you can now remap it from the UI without writing a regex against the message body. Six one-click templates for the most common shippers. Live preview as you build the rule.

• New

  SLOs: per-SLO alert routing

  Each SLO can target a specific notification channel. Payment-SLO breaches go to #on-call-payment, search-SLO to #sre-log. Previously every SLO alert went to every enabled channel for the tenant.

• Improved

  Tier limits surfaced upfront

  SLO and rule counts now display as “X of N defined” in the UI before you hit the cap. Pricing page lists exact counts per tier — trial and Team include 5 SLOs, 20 threshold rules, unlimited notification channels; Growth and Enterprise are unlimited.

• Improved

  Settings polish + accessibility pass

  Every destructive action in Settings now names the entity being deleted in the confirm dialog. Every icon-only action button has an accessible label. Error banners are announced to assistive tech. Form labels are correctly paired with inputs. Mobile tables scroll horizontally instead of squishing.

Dashboards v2 · Investigation page · Detector intelligence

• New

  Dashboards v2: query-first canvas

  New drag/resize grid layout with cross-filter context — click any value on any panel and the whole board re-filters to it. Six panel types: timeseries, bar, table, stat, alert timeline, live tail. Markdown panels for documentation alongside the data. Auto-created “Getting started” board on first ingest.

• New

  Investigation page

  New analyze view that surfaces probable root cause, blast radius, cascade timeline, and what-changed analysis (recent deploys, new errors, volume shifts) in one panel. Every Slack and PagerDuty alert deep-links to it.

• New

  Domain-specific detector packs

  Four detector categories now run out of the box on every tier: Kubernetes (70+ rules for OOMKilled, CrashLoopBackOff, image pull failures, evictions), AWS service patterns (RDS, S3, DynamoDB, ECS, EKS, IAM, KMS, Lambda — 28 rules), web/HTTP (4xx and 5xx surges, slow endpoints, gateway timeouts), and infrastructure (disk pressure, memory pressure, kernel events).

• Improved

  Adaptive auto-resolve for flapping alerts

  An alert that fires, resolves, and re-fires repeatedly gets a progressively longer auto-resolve window — 15 min → 30 min → 45 min — so you stop seeing the same noisy condition reopen every few minutes. Stable alerts keep the default short window.

• Improved

  Honest demo framing

  The demo at app.getepok.dev/demo now describes itself as “synthetic data, real detectors” — the log stream is example data from a 5-service microservices app, but every detector, every analysis, every alert is the real engine working on it. No video, no canned screenshots.

Public availability · 20 detectors on every plan

• New

  Every detector on every tier

  All 20 detectors — anomaly detection, new error fingerprinting, silence detection, golden signals, SLO monitoring, pattern clustering, K8s/AWS/web/database intelligence — enabled by default on every tier. No upsell wall on detection.

• Improved

  Flat pricing, no surprises

  Predictable monthly invoice regardless of traffic shape. No per-host, per-event, or cardinality fees. Current pricing on the Pricing page.

• New

  Universal ingest

  Accept logs from any source that can speak HTTP: Elasticsearch Bulk, Loki Push, OTLP HTTP, Fluent Bit native, Fluentd, syslog (RFC 5424 / 3164 / Cisco / Fortinet / Palo Alto), CloudWatch subscription filters, GCP Cloud Logging via Pub/Sub, plain JSON. No proprietary agent required.