← Back to all sparks
S

SigNoz

INFRA · APIS
Velocity6.3

SigNoz rebuilt its dashboards for agents to drive, and is now pruning the API surface behind them.

observabilityagent-operable dashboardsprometheus compatibilityopen specsrbacapi deprecation
Current state
The dashboard rewrite is the anchor of the current cycle: custom views, tagging, per-user pins, and a search query DSL, all structured so an AI agent can operate on dashboards as reliably as a person. Around it, SigNoz has been closing compatibility and access gaps — PromQL brought into line with the Prometheus specification including anchored regex matchers, timeseries export from every explorer plus CSV and JSONL download in the Trace View, six more onboarding integrations, and fine-grained access control in beta for Cloud and self-hosted Enterprise. The newest entry is a deprecation notice: the v1 alert history endpoints are being removed over a security vulnerability, and API callers have to migrate before they start returning 404.
Where it's heading
The pattern is conformance over invention. Rather than differentiate through proprietary query behaviour, SigNoz is matching Prometheus semantics exactly, exposing structured schemas agents can rely on, and letting data leave through export. That trades lock-in for substitutability, which is the trade an open-source observability vendor makes when the bet is on being easy to adopt and hard to argue against. The breaking changes follow the same logic — anchored regex matchers and removed v1 endpoints both break existing usage in order to leave a cleaner contract behind.
Prediction
Expect the structured, agent-operable treatment applied to dashboards to spread to alerts and their history next, since that is where the API surface is being rebuilt anyway, and expect fine-grained access control to leave beta once those surfaces settle.

Recent moves

  1. 7d ago

    ⚠️ Action required (API users) - migrate off the v1 alert history endpoints

    The v1 alert history endpoints are deprecated and will be removed to close a security vulnerability, after which calls to the stats, timeline, top_contributors, and overall_status routes return 404. Anyone with scripts or integrations against them has to migrate — the second breaking change this cycle, after the anchored PromQL regex matchers.

    View source ↗
  2. 14d ago

    A Revamped Dashboard Experience for Humans and Agents

    ⚡ SPARK

    Dashboards were rebuilt to be navigable by hand and operable by agents at the same time, with a query DSL that reads the same either way. It is the clearest statement of the strategy the PromQL and export work also serve: make the surfaces conform to a spec rather than to SigNoz.

    View source ↗
  3. 21d ago

    PromQL conformance with the Prometheus specification

    SigNoz's PromQL implementation is brought into line with the Prometheus specification, closing behaviour gaps. The change is breaking: regex matchers are now anchored, so a pattern that previously substring-matched will match less than it used to and queries need review.

    View source ↗
  4. 28d ago

    Data export: timeseries from all explorers, download in the Trace View

    Timeseries data can be exported from the Time Series view in every explorer, and the Trace View gains CSV and JSONL download. Letting data out easily is consistent with the conformance strategy — the product competes on being good to use, not on holding the data.

    View source ↗
  5. 1mo ago

    More Onboarding Integrations

    Six more data sources — NGINX Ingress, FluxCD, Cassandra, Cloudflare Workers, PlanetScale, and the Hermes Agent — get guided onboarding, alongside streamlined Kubernetes navigation and host disk usage. Adoption friction work rather than a change in what the product does.

    View source ↗
  6. 1mo ago

    Fine-grained access control (beta)

    Custom roles with scoped permissions reach beta for Cloud and self-hosted Enterprise, replacing an all-or-nothing model where only admins could manage service accounts and API keys. It is the delegation groundwork any team needs before letting automation hold its own credentials.

    View source ↗