| 开发者 | benkalsky |
|---|---|
| 更新时间 | 2026年9月9日 22:26 |
| PHP版本: | 7.2 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
rest, cron, cli or xmlrpc.digitizer-ai-agent-log/v1/activity, which requires manage_options. There is deliberately no route that deletes: a log that can be erased through the API is a log an attacker erases on the way out.
Multisite
Each site keeps its own log, in its own table, and a run that switches between sites files each change under the site it happened on.
Changes are held in memory during the request and written once, at shutdown, and only when the request came in on a channel worth recording. A browser request writes nothing at all.
No. Nothing leaves the site.
Old rows are pruned automatically, at most once an hour.
Yes, with a filter. Nothing is filtered out by default, because what is absent from this log is supposed to mean it did not happen over an API:
add_filter( 'digitizer_ai_agent_log_record', function ( $record, $entry ) { if ( 'elementor_library' === $entry['object_subtype'] ) { return false; } return $record; }, 10, 2 );
The entry carries object_type, object_subtype, object_id, object_name, action, fields, blog_id, channel, app and user_id, so a rule can match on who made the change as well as what it touched. On a network the filter runs inside the site the change happened on, so a callback may read that site's own options to decide.
Note that such an entry is not a mistake: WordPress refuses an identical meta write before the plugin ever sees it, so a change that reaches the log did alter the row. It is simply a change you may not care about, and only your site can say which those are.
The table is dropped and both of its options are deleted, on every site of a network. Nothing is left behind.
blogdescription) is recorded when it changes, beside the site name.digitizer_ai_agent_log_record filter, applied to every entry once the channel and the application name are known, so a site can silence a writer it does not care about. Nothing is filtered by default.