SYSTEM Cited by 3 sources
PlanetScale Insights¶
What it is¶
Insights is a Postgres extension developed by PlanetScale and available exclusively on PlanetScale Postgres clusters. It provides per-query performance telemetry and hosts Traffic Control, the feature that enforces workload-class resource budgets on tagged queries.
Capabilities¶
- Per-query performance telemetry — query duration, plan shape, resource consumption aggregated per statement for "find slow SQL" workflows.
- Traffic Control — tag queries with
SQLCommenter metadata
(e.g.
action=analytics) and apply Resource Budgets capping concurrent workers, server share, or per-query cost. See systems/planetscale-traffic-control. - AI-powered index suggestions — Insights periodically
uses its own per-query-pattern telemetry (rows read,
rows returned, runtime share, frequency) to pick
candidate queries, feeds them to an LLM that proposes
CREATE INDEXstatements, and validates each candidate by runningEXPLAINwith a hypothetical index via HypoPG. Only suggestions that materially reduce planner-estimated cost are surfaced. See concepts/llm-generated-database-changes, patterns/llm-plus-planner-validation, patterns/workload-aware-llm-prompting.
Why it exists¶
Upstream Postgres ships observability primitives
(pg_stat_statements, pg_stat_user_tables) but no first-class
mechanism for classifying queries by workload and enforcing
class-level concurrency limits. Insights fills that gap with
PlanetScale-specific extension code running inside the
PlanetScale Postgres binary.
Seen in¶
- sources/2026-04-11-planetscale-keeping-a-postgres-queue-healthy
— Introduced as the vehicle that makes Traffic Control possible on
PlanetScale Postgres. The article's concrete Traffic Control
configuration (cap
action=analyticsto 1 worker, 25% ofmax_worker_processes) uses Insights' SQLCommenter tag recognition as the classification substrate. - sources/2026-04-16-cloudflare-deploy-postgres-and-mysql-databases-with-planetscale-workers — Called out as one of PlanetScale Postgres's headline differentiators on the Cloudflare partnership launch, alongside database branching and agent-driven SQL tooling.
- sources/2026-04-21-planetscale-ai-powered-postgres-index-suggestions
— Insights gains an AI-powered index suggestions
surface. Rafer Hazen's launch post canonicalises Insights'
per-query-pattern telemetry (rows-read vs rows-returned ratio,
≥ 0.1% aggregate runtime share, minimum frequency) as the
input oracle that makes LLM-generated index suggestions
useful — and HypoPG+
EXPLAINas the output oracle that makes them safe. Positions Insights' workload data as the structural moat: "to use the unique data and capabilities available in Insights to produce the best overall results."
Source¶
Related¶
- systems/planetscale-traffic-control
- systems/planetscale-for-postgres
- systems/planetscale
- systems/postgresql
- systems/hypopg
- patterns/workload-class-resource-budget
- patterns/llm-plus-planner-validation
- patterns/workload-aware-llm-prompting
- concepts/llm-generated-database-changes
- concepts/hypothetical-index-evaluation
- concepts/index-candidate-filtering
- companies/planetscale