How dbtempo analyzes a query

Detect, analyze, and fix

Detect

Identify a query pattern and inspect the query text, workload statistics, schema, indexes, and execution plan that are available for it.

Analyze

Turn the matched rule and supporting artifacts into a diagnosis, confidence level, evidence, expected impact, and risk notes.

Fix

Propose a concrete next step such as migration SQL, suggested SQL, a query rewrite, a configuration change, or manual investigation.

The analysis cost ladder

Every analyzable query starts with the deterministic rules engine. When those rules fully resolve the finding, dbtempo returns the result without an AI call. More complex queries can move to a light or more capable AI tier according to the query's complexity, the available context, and the analysis level included in the current plan.

Deterministic rule evaluation does not consume credits. AI-backed analysis does, and the current share of the monthly allowance is shown in Settings → Billing. The guide does not publish fixed credit amounts because plan allowances can change.

Context and escalation

A ready database can add schema, index, statistics, and execution-plan context. If a required artifact is missing, the thread asks you to attach a connection, run a displayed read-only statement yourself, or continue with the context that is already available.

Queries observed repeatedly in monitored workloads can be escalated for deeper analysis. Automated plan probes use EXPLAIN, not EXPLAIN ANALYZE, so the probe does not execute the query. Source indexing can add call-site evidence when a connected source is available.

What the result means

A completed analysis may create a durable recommendation or report that the available evidence did not reveal an issue. Treat each recommendation as a reviewable engineering decision: read its evidence, fix zone, risk notes, and rollback guidance before acting.