Adding a DB Column

How to add a column to an existing table.

Steps

In src/db/schema.sql, add the column to the CREATE TABLE statement and add an idempotent add_column_if_not_exists() call. Existing databases never rerun CREATE TABLE IF NOT EXISTS, so without the call the column reaches only fresh installs:
```
SELECT add_column_if_not_exists('table_name', 'column_name', 'BOOLEAN', 'false');
```
If existing rows need a backfill, or a column moves or is dropped, add a numbered migration src/db/migrations/NNN_name.sql with a matching .down.sql. Choose NNN after checking every branch (git ls-tree), because bun run check-migrations only sees the working tree.
Add the field to the Zod schema and types in src/types/db/schema.ts.
Read and write it through the owning repository in src/utils/db/repositories/ (see Raw SQL Boundary).
After a successful write, invalidate the affected caches in the same code path; never before the write and never on failure.

Runtime config columns: a column on a server_*_configs table that the bot reads through tomoriState.config must also be added to both config SELECTs in src/utils/db/repositories/PersonaRepository.ts (loadTomoriState and loadAllForServer; search for scaps.tool_use_enabled). assembledServerConfigSchema gives each field a .default(), so a column missing from the SELECT is silently replaced by its default and a check like config.flag === false never fires, with no type or test failure.
Descriptions: model and preset descriptions use the existing descriptions JSONB locale map. Do not add a column per language; add translations to the seed catalog’s i18n field and keep English in desc.

Verify

bun run check
bun run lint
bun run check-migrations   # if you added a migration
bun run db:lifecycle       # needs a local PostgreSQL user that can create databases

Schema reference: Database Schema. Cache map: Caching.