Settings

Everything is namespaced under pitta.. Open Preferences: Open Settings and filter on pitta, or edit settings.json directly.

Settings apply per user by default and can be overridden per workspace — except the guard settings, which VS Code refuses to take from an untrusted workspace so a cloned repository cannot turn them off. See Safety.

Safety

Setting	Default	What it does
pitta.query.confirmDestructive	prod	Which environments confirm before DROP / TRUNCATE / unqualified DML. The typed confirmation on prod cannot be turned off.
pitta.query.blockUnqualifiedDml	true	Refuse UPDATE/DELETE with no WHERE, on any connection.
pitta.query.statementTimeoutMs	0	Server-side statement timeout. 0 disables it.
pitta.query.autocommit	true	Run statements outside an explicit transaction.
pitta.redis.allowKeysCommand	false	Allow KEYS. It walks the whole keyspace and blocks the server while it does; the tree uses SCAN instead.
pitta.redis.allowMonitor	false	Allow MONITOR. It makes the server report every command to this client and measurably slows it for everyone. Refused on production whatever this is set to.
pitta.redis.monitorSeconds	10	How long MONITOR runs before stopping itself.

Results and memory

Setting	Default	What it does
pitta.defaultFetchSize	500	Rows fetched per page from a server-side cursor.
pitta.query.streamResults	true	Keep the cursor open after the first page and read on as you scroll. Turning it off reads up to pitta.maxRowsInMemory before the first row appears.
pitta.maxRowsInMemory	50000	Rows a result keeps before it is marked truncated. Costs roughly 0.5 KB per row per 20 columns, so the default caps an accidental SELECT * at about 27 MB.
pitta.gridCacheBudgetMB	200	Memory budget for cached result pages; the oldest are evicted first.

Exports and imports stream to and from disk and are not bounded by any of these. A table larger than memory exports fine.

Explorer

Setting	Default	What it does
pitta.explorer.pageSize	500	Objects per level before a Show more entry. VS Code virtualizes rendering, but handing it tens of thousands of items still costs a call each as you scroll.
pitta.explorer.showSystemObjects	false	Show pg_catalog and information_schema.
pitta.explorer.rowCountMode	estimate	estimate reads the catalog's row estimate and is free. exact runs COUNT(*) — accurate, and a full scan on a large table.
pitta.metadataCacheTtlSeconds	300	Schema cache lifetime. 0 re-reads the catalog every time. The cache is also invalidated eagerly when Pitta runs DDL itself.

Editor

Setting	Default	What it does
pitta.editor.completion	true	Suggest tables, columns and functions from the bound connection.
pitta.editor.diagnostics	semantic	off, syntax (grammar only), or semantic (adds unknown-name and type checks against the catalog).
pitta.editor.diagnosticsDelayMs	400	Pause after typing before diagnostics run.
pitta.editor.formatOnSave	false	Format .sql files bound to a Pitta connection on save.
pitta.editor.keywordCase	upper	Keyword case the formatter produces.
pitta.editor.maxLineLength	100	Length past which the formatter breaks a comma-separated list.

History and saved queries

Setting	Default	What it does
pitta.history.enabled	true	Record executed statements locally.
pitta.history.excludeEnvironments	["prod"]	Environments kept out of history entirely.
pitta.history.maxEntries	5000	Entries kept before the oldest are dropped.
pitta.savedQueries.folder	.pitta/queries	Workspace folder holding saved .sql files.

Diagnostics and privacy

Setting	Default	What it does
pitta.logLevel	info	Verbosity of the Pitta output channel: trace, debug, info, warn, error, off. Set debug before reporting a bug.
pitta.telemetry.enabled	false	Off. Nothing is collected or transmitted.

Per-connection overrides

Some of these are also profile fields, and the profile wins for that connection: statement timeout, fetch size, autocommit, search_path, TimeZone and application_name. Set them on the profile when they belong to the server rather than to you.