Extension catalogue

Extension catalogue src/extensions is a searchable catalogue of 124 real database extensions, plugins and licensed features across the supported dialects. It is metadata, not integration code: nothing in it runs SQL or talks to a driver. It answers "does this engine have something that does X, and what is it called?" — then hands you the install statement to run. Actually installing an extension is a dialect operation: Only the PostgreSQL family has a genuine CREATE EXTENSION mechanism. Entries for other engines describe the closest real equivalent — a plugin, a licensed server option, a feature flag — and say so with isNativeSqlExtension: false. Querying the catalogue An entry: Note what that description does: it records that CockroachDB accepts the statement while providing none of the functionality. That distinction — between "it works" and "it does not error" — is the reason the catalogue exists. installCommand is omitted where there is no single statement, because the thing is a licensed server option toggled outside SQL. Categories full-text-search, data-types, json, spatial-gis, cryptography, uuid-generation, vector-search, indexing, foreign-data, scheduling, partitioning, monitoring-stats, utility, procedural-language, time-series, replication, security-audit, high-availability, connectivity, compression, in-memory, machine-learning. Coverage Source file | Entries | Covers | --- | --- | --- | postgres-extensions.ts | 51 | Real CREATE EXTENSION extensions — pgtrgm, postgis, pgcrypto, hstore, uuid-ossp, pgvector, … | mysql-mariadb-plugins.ts | 20 | MySQL/MariaDB plugins (INSTALL PLUGIN) and storage engines | oracle-db2-features.ts | 21 | Oracle and IBM Db2 licensed options and built-in features | mssql-features.ts | 13 | SQL Server features — In-Memory OLTP, Always Encrypted, PolyBase, … | cloud-warehouse-features.ts | 19 | Snowflake / Redshift / ClickHouse / BigQuery-style capabilities | Every entry was verified against product documentation rather than generated to hit a number. 124 is what could be verified; padding it would mean inventing plausible-sounding entries, which is worse than a shorter list. A worked example You want vector search and do not know what your engine offers: On PostgreSQL that finds vector (pgvector), whose install statement you can run directly: after which DataTypes.VECTOR(1536) maps to a real VECTOR(1536) column — see Data types. Related reading extensions — the full entry shape and per-file breakdown Schema objects — createExtension and friends Dialects — per-engine capability notes Database types