Existing databases

Keep your application’s migrations and match the Rust timestamp type to each existing column:

PostgreSQL column	Rust type
TIMESTAMPTZ	chrono::DateTime<chrono::Utc>
TIMESTAMP WITHOUT TIME ZONE	chrono::NaiveDateTime
Nullable timestamp	Option around the matching type

For example, an existing verification table with naive timestamps can use:

#[derive(Clone, Debug, serde::Serialize, sqlx::FromRow, better_auth::sqlx::AuthEntity)]
#[auth(role = "verification", table = "verifications")]
pub struct VerificationModel {
    pub id: String,
    pub identifier: String,
    pub value: String,
    pub expires_at: chrono::NaiveDateTime,
    pub created_at: chrono::NaiveDateTime,
    pub updated_at: chrono::NaiveDateTime,
}

Select this model in your application’s AuthSchema; see database concepts. The same mapping applies to user, session, and account timestamps.

Naive timestamps must already represent UTC wall-clock time. Generated accessors interpret them as UTC, and writes bind them without an offset. This does not convert local-time data or repair previously shifted values.

Bundled models use aware timestamps. Use custom rows for existing PostgreSQL TIMESTAMP columns. For handwritten SQLx implementations, pair SqlValue::NaiveTimestamp with ColumnKind::NaiveTimestamp so predicates and writes agree. SeaORM supports the same naive UTC convention; plugin tables use their bundled models.

PostgreSQL fixed-width IDs

Keep existing CHAR(n) columns and String fields. SQLx models opt into PostgreSQL’s bpchar wire type on each fixed-width field:

#[derive(Clone, Debug, serde::Serialize, sqlx::FromRow, better_auth::sqlx::AuthEntity)]
#[auth(role = "session", table = "sessions")]
pub struct SessionModel {
    #[auth(column_type = "bpchar")]
    pub id: String,
    #[auth(column_type = "bpchar")]
    pub user_id: String,
    pub token: String,
    pub expires_at: chrono::DateTime<chrono::Utc>,
    pub created_at: chrono::DateTime<chrono::Utc>,
    pub updated_at: chrono::DateTime<chrono::Utc>,
    pub ip_address: Option<String>,
    pub user_agent: Option<String>,
    pub active: bool,
}

Apply this to all CHAR(n) primary IDs, foreign keys, and optional fixed-width fields. Parameters and nulls use bpchar, leaving indexed columns unchanged. Handwritten implementations pair ColumnKind::BpChar with SqlValue::BpChar. Ordinary TEXT/VARCHAR fields retain text bindings; SQLite binds the exact string as text.

Both entity derives also accept id_generator = "path::to::function" on the model. Supply an application function returning a unique String that fits your existing columns. It runs only when no explicit ID is supplied; the default UUID is 36 characters and is never shortened automatically.

For SeaORM PostgreSQL fields, use column_type = "Char(Some(30))", save_as = "bpchar" alongside the usual primary-key attributes. The cast applies to parameters, preserving indexes. Omit save_as on SQLite/MySQL models.

PostgreSQL pads stored CHAR(n) values, ignores trailing spaces in equality, and rejects overlength non-space input. The adapter does not trim, pad, or truncate returned strings. See PostgreSQL character semantics. Keep application migrations in charge of existing schemas.