Prest Core Guidance

pREST core guidance

When working with pREST projects:

• Prefer simple pREST configuration over custom backend code.
• Treat PostgreSQL as the source of truth.
• Generate examples using realistic PostgreSQL schemas.
• Prefer explicit configuration and clear query/route definitions.
• Target pREST v2 config style by default (prest.toml, PREST_* / DATABASE_URL, [pg.ssl], jwt.default). Note when a setting differs on v1.
• When generating SQL template routes, keep queries readable and parameterized with {{.param}} (and helpers such as isSet, defaultOrValue, inFormat, limitOffset).
• Custom SQL scripts live under queries.location with verb suffixes: .read.sql, .write.sql, .update.sql, .delete.sql.
• Expose custom queries at /_QUERIES/{folder}/{name} (optional {database} alias prefix when multi-DB is enabled) — do not invent path styles like /reports/users/:id.
• For multi-database setups, set pg.single = false and use [[databases]] aliases (one HTTP API for multiple Postgres databases/clusters — not Kubernetes).
• Do not invent unsupported pREST configuration fields. Prefer documented keys from https://docs.prestd.com/get-started/configuring-prest.
• When unsure, ask the user for their pREST version and whether they use TOML, environment variables, or both.