Postgres query-optimization tooling
The Optimization section of awesome-postgres, all fourteen entries, as catalogued on
2026-08-20, plus dexter which the catalog files under Monitoring and which plainly belongs
here. pev2 and hypopg are written up from primary sources; the rest are at the catalog's word
(T3).
This is the section that aims directly at growth edge 9, and reading it as a group shows the field splitting by which artifact each tool reads.
Reading a single plan
| Entry | Catalogued as |
|---|---|
| pev2 | Online Postgres EXPLAIN visualizer. Paged from primary sources |
| pgMustard | Modern EXPLAIN UI with performance tips (commercial) |
| pg_flame | Flamegraph generator for query plans |
Three renderings of the same input, the plan. pg_flame's choice of a
flamegraph is the interesting variation — it makes time attribution down the tree the primary axis,
where pev2 ranks by estimate error.
Reading the accumulated workload
| Entry | Catalogued as |
|---|---|
| PoWA | Workload analyzer with real-time charts |
| pg_web_stats | Web UI for pg_stat_statements |
| PgHero | PostgreSQL insights made easy |
| pganalyze | Performance monitoring (commercial) |
| Metis | Observability and performance tuning (commercial) |
All five sit on pg-stat-statements. pg_web_stats says so in its own catalog line.
Proposing schema changes
| Entry | Catalogued as |
|---|---|
| hypopg | Hypothetical/virtual indexes. Paged from primary sources |
| dexter | Automatic indexer detecting slow queries |
| EverSQL | Automated query optimization and indexing recommendation (commercial) |
| pgassistant | Developer tool for optimization with LLM integration |
dexter is the natural pair to hypopg and the higher-value next read: it closes the loop by taking slow queries from the logs, generating candidate indexes, and testing them using HypoPG rather than building them. That makes it the first fully automatic instance of the find-it/diagnose-it/fix-it chain the spoke has been assembling one page at a time.
pgassistant is the only entry in the entire catalog that mentions an LLM, which is a fact about the catalog's vintage as much as about the tool.
Changing configuration rather than queries
| Entry | Catalogued as |
|---|---|
| pgtune | Configuration wizard (listed twice — a tool and an online version) |
| pgconfig.org | Online configuration tool |
| TimescaleDB Tune | Tuning program for available resources |
| aqo | Adaptive query optimization extension |
aqo is the one that does not fit the section. Adaptive query optimization means feeding actual row counts back so the planner learns from its own errors — which attacks the independence assumption at its root rather than working around it per query. If any entry in this catalog deserves a primary read for growth edge 9, it is this one.
Next candidates
aqo (feedback-driven estimation — the deepest answer available to edge 9), then dexter (automation over hypopg), then PoWA.
Related
postgresql-planner-statistics · pev2 · hypopg · query-execution-plan · pg-stat-statements · postgres-monitoring-tooling · awesome-postgres · operational-databases