Documentation Index
Fetch the complete documentation index at: https://mintlify.com/HypathStack/model-scribe/llms.txt
Use this file to discover all available pages before exploring further.
Audit logs grow without bound unless you have a deliberate strategy for pruning them. ModelScribe’s database driver ships with three retention modes that cover the most common operational needs: keep everything forever, discard records older than a configurable number of days, or maintain a fixed-size rolling window of the most recent records per table. Retention is declared in config/model-scribe.php and enforced on-demand when you run model-scribe:prune.
Retention modes
The default mode. ModelScribe never deletes records automatically — every
audit entry is kept indefinitely. Use this when your compliance or legal
requirements mandate a full immutable history, or when you manage archival
and deletion outside of ModelScribe.// config/model-scribe.php
'drivers' => [
'database' => [
'driver' => 'database',
'table' => 'model_scribe_logs',
'retention' => [
'type' => 'permanent',
],
],
],
model-scribe:prune while type is 'permanent' is a safe no-op
— it returns 0 deleted records and touches no rows. Deletes all records whose created_at timestamp is older than days days
when model-scribe:prune runs. This is the most common choice for
applications that need a rolling compliance window (e.g. 90-day GDPR
retention, 1-year SOC 2 retention).// config/model-scribe.php
'drivers' => [
'database' => [
'driver' => 'database',
'table' => 'model_scribe_logs',
'retention' => [
'type' => env('MODEL_SCRIBE_RETENTION', 'days'),
'days' => (int) env('MODEL_SCRIBE_RETENTION_DAYS', 90),
],
],
],
| Variable | Purpose | Default |
|---|
MODEL_SCRIBE_RETENTION | Sets the type | 'permanent' |
MODEL_SCRIBE_RETENTION_DAYS | Number of days to retain | 90 |
MODEL_SCRIBE_RETENTION=days and MODEL_SCRIBE_RETENTION_DAYS=365 in
your .env to keep one year of history, for example. Keeps only the latest keep records per table, deleting everything
older. This is ideal for high-frequency models — such as a heartbeat or
sync-status model — where you care about the recent trail but not the full
history.// config/model-scribe.php
'drivers' => [
'database' => [
'driver' => 'database',
'table' => 'model_scribe_logs',
'retention' => [
'type' => env('MODEL_SCRIBE_RETENTION', 'rotating'),
'keep' => (int) env('MODEL_SCRIBE_RETENTION_KEEP', 500),
],
],
],
| Variable | Purpose | Default |
|---|
MODEL_SCRIBE_RETENTION | Sets the type | 'permanent' |
MODEL_SCRIBE_RETENTION_KEEP | Number of records to retain | 500 |
id DESC, skips the first keep rows, and reads the id of the next row.
Every record with an id less than or equal to that cutoff is then deleted
in a single DELETE statement. If the table contains fewer than keep
records, nothing is deleted. Retention is enforced at prune-time
ModelScribe does not delete records on insert. Entries accumulate in the
database as they are written, and the configured policy is only enforced when
you explicitly trigger a prune. This means your tables may grow beyond the
retention threshold between prune runs — plan your schedule accordingly.
Run the prune command manually at any time:
php artisan model-scribe:prune
# Prune a specific driver only
php artisan model-scribe:prune --driver=database
Schedule the prune command in routes/console.php so it runs automatically.
Daily is a sensible default for most applications:use Illuminate\Support\Facades\Schedule;
Schedule::command('model-scribe:prune')->daily();
rotating retention you may want to prune
more frequently — for example ->hourly() — to keep table size tightly
bounded. The file driver’s prune() method is intentionally a no-op and always
returns 0. Log file rotation for the file driver is handled entirely by
Laravel’s built-in logging system. Configure it in config/logging.php using
the daily channel’s days option or whichever channel you have pointed
MODEL_SCRIBE_LOG_CHANNEL at.
The rotating retention mode counts records per table, not per subject
model. If a table is shared by multiple model types — for instance, several
models all using the default model_scribe_logs table — the keep limit
applies to the combined total across all of them. A single high-frequency
model can therefore push older entries for other models out of the window
ahead of schedule. Use named stores to
give each model type its own table before applying rotating retention.