Skip to content

Release 2.1

[2026-06-17]

Major feature release — advanced detectors, special-date handling, connector improvements, and platform hardening since Release 1.1.

Advanced anomaly detectors

New and extended detector types in data view YAML. See Detectors for the catalog and per-detector pages.

Data quality & freshness

  • Bounds — flags values outside an allowed [min, max] range (mode: below, above, both).
  • Stale — flags series whose last data point is older than maxAgeDays relative to the expected latest date; optional maxReportAgeDays stops reporting very old stale series.
  • Flatline — detects near-zero or effectively constant values over minDurationDays.
  • Zero rate — flags windows where the share of near-zero days exceeds rateThreshold.

Trend & seasonality

  • DoPY (dopy) — day-over-same-day-prior-year comparison.
  • Sustain — run-length deviations where values stay above or below baseline for at least minDurationDays.
  • Seasonal residual (sr) — trend + seasonal baseline with robust z-score on residuals.
  • Volatility — regime shifts in rolling variability (mad or std) vs. a longer baseline window.

Business & structural comparisons

  • Target — compares actuals to explicit business targets (static daily/weekly/monthly or variable monthly), with optional where scope and valuesBy per dimension.
  • Peer — compares each series to peers sharing groupBy dimensions while differing on varyBy; baselines precomputed at load time.
  • Mix — flags anomalies in a child series' share of its parent aggregate (parentBy, childVaryBy, daily/weekly/monthly granularity).
  • Cardinality — flags anomalies in distinct-value counts of a countColumn grouped by groupBy; all granularities computed from raw data.

Detector behavior

  • Direction mode — periodic, quantile, and advanced detectors support mode: above | below | both.
  • Aggregate mode — aggregated detectors (mom, mpy, wow, wpy, qoq, qtw, qtm, target weekly/monthly) support aggregateMode: sum | average.
  • Series filters — detectors can be scoped with where filters and optional maxValue thresholds; skipped detectors report reasons in data preview.
  • Baseline precomputation — peer, mix, and cardinality detectors store per-fact baseline tables at load time for fast detection.

Special-date suppression

Known calendar events can suppress expected anomalies instead of generating noise.

  • Configure per detector with useSpecialDates: true.
  • Define impact in data view YAML:
  • specialDates — explicit dates with optional impact (positive, negative, both).
  • recurringRules — rules such as month_end, month_start.
  • specialPeriods — date ranges (including year-crossing periods like 12.15–01.14).
  • Matching anomalies are excluded based on anomaly direction and configured impact.

Connectors

  • Microsoft Fabric SQL (fabric) — service-principal authentication via structured config (server, database, tenant_id, client_id, client_secret).
  • Structured config JSON — connector settings stored as encrypted JSON instead of a raw connection string; UI adapts fields per engine type.
  • Legacy compatibility — API requests may still send connectionString for SQL-based connectors; it is converted to config JSON on save.
  • Query preview — improved connector query sampling for data preview runs.

Filters & LLM summarization

  • Saved filters — full admin UI with Aggregates, Anomalies, and Summary result tabs.
  • Pinning & sharing — per-user pin state and shared filter access across users.
  • LLM summarization — configurable templates in Setup; refresh on demand or when data views change; settings redacted in API responses.
  • Comment indicators — filter results show linked operational comments on matching dates.

Security & operations

  • API hardening — public API gate middleware, security response headers, API rate limiting.
  • Credential protection — centralized ConnectionStringCrypto; SQL query safety checks.
  • Data view security keys removed — per-data-view secret keys dropped; access controlled via standard auth.
  • Worker task queue — optimized task queue schema and processing for load/anomaly jobs.
  • Anomaly index — IsCurrent flag on anomaly records; recalculation respects current vs. historical rows.

Platform & packaging

  • AnomalyGuard CLI — cross-platform anomalyguard tool for REST operations, config, and CI/CD workflows; downloadable from the Master UI.
  • API keys — admin UI and API for key management with optional description.
  • Docker & CI/CD — Master and Worker container images; Azure Pipelines for app build (ACR) and documentation publish.
  • Azure Marketplace — ARM templates and managed-app packages for Start, Small Team, Growing Team, and Enterprise Team SKUs.
  • Versioning — application version centralized in version.props (2.1.0).

User interface

  • Home — redesigned result cards with configurable max tiles (5, 15, etc.) and Summary / All views.
  • Comments — dedicated add/detail pages; list view streamlined.
  • Monitoring — list and detail pages split; detail shows load, range, anomaly, and task stats in one row.
  • Data views — enhanced YAML editor, load-mode persistence, license usage display ({count} from {limit}).
  • Data preview — chart-only runs (no detectors), skipped-detector reasons, Plotly chart builder improvements.
  • UI error logging — consistent UiLog error reporting across Blazor pages.

Documentation

  • MkDocs site with Admin Guide, User Guide, Tutorials, and Release Notes.
  • New pages: Detectors catalog with per-detector docs, updated Connectors, DataViews, and sample YAML with all new detector types.