Changelog

PromQL conformance with the Prometheus specification

SigNoz's PromQL implementation had several gaps where its behavior diverged from Prometheus. This release brings it into line with the Prometheus spec.

⚠️ Action may be required: regex matchers are now anchored, matching Prometheus. env=~"foo" is now treated as env=~"^foo$". Previously SigNoz did substring matching, so a pattern could over-match — job=~"api" also matched api-server. If you relied on substring behavior, wrap the pattern explicitly: use =~".*foo.*" for contains, =~"foo.*" for prefix, and write glob-style habits like name* as name.*. This is the one change that can alter the results of existing queries.

Also fixed, all bringing the PromQL implementation in line with Prometheus semantics:

• Selectors without an exact metric name now work — nameless selectors ({job="api"}), regex metric names ({__name__=~"prod_.+"}), and negations previously returned empty results silently.
• Series identity now follows Prometheus semantics, so without() aggregations no longer return duplicate series and vector matching joins correctly.
• Series first seen exactly at a query window's end are no longer missed.
• User labels beginning with __ (e.g. __address__) are preserved; only SigNoz-internal storage keys are stripped.

Investigate and optimize telemetry costs with Noz or your AI agent

You can now investigate telemetry costs using Noz in SigNoz Cloud or the telemetry cost investigation skill from any AI agent.

Ask “Why is my bill high?” or “Help me optimize costs.” The skill analyzes Cost Meter data, ranks telemetry by ingestion volume, checks whether the data is used in dashboards or alerts before recommending what to drop, and identifies further optimization opportunities.

In SigNoz Cloud, you can run the investigation directly from Noz. Cloud and Community Edition users can also invoke the same skill using a SigNoz plugin or an individual agent skill.

When an AI agent has access to your codebase, it can apply the recommended instrumentation and configuration changes directly.

Table export in the Logs and Traces explorers

You can now export results from the table tabs in the Logs and Traces explorers as CSV or JSONL, bringing them to parity with the other explorer views.