[XB] User Fake Activity: AI Content, Online Users & Simulated Engagement 2.1.2

New Features​

• Added additional AI profiles with optional fallback AI profile used when the main profile cannot generate the requested content. Existing AI connection settings are moved to a new profile automatically when upgrading.
• Added batch deletion of AI content pool items
• Added XML import and export for selected activity triggers, with imported copies disabled by default for review

Improvements​

• Added an option to share member names with AI services, using anonymous labels by default
• Improved automated reply processing, counting performance, and enforcement of reply limits
• Improved activity scheduling, execution time limits, and protection against overlapping runs
• Improved AI content pool management, recovery, cleanup, and handling of failed content generation
• Improved fake user registration, simulated browsing, and permission-checking performance
• Increased custom and per-forum prompt limits to 10,000 characters, with validation instead of silent truncation
• Added a configurable probability for quoting replies, preserving existing quote settings
• Replies now always follow the discussion context and wait for a new message from a person before responding again
• Added a configurable interval between replies by the same user, shared across triggers
• Improved protection against duplicate replies during overlapping runs and replies to outdated discussions
• Improved background activity processing, execution time limits, error recovery, and prevention of overlapping runs
• Improved trigger scheduling across time zones, overnight windows, daylight saving transitions, and overdue activities
• Improved fake user selection, group targeting, and registration candidate generation
• Improved AI content pool management, initial generation, and daily request limit enforcement
• Improved ACP trigger management, filtering, sorting, dashboard charts, and activity reporting
• Improved PHP 8.4 compatibility and support for servers without the PHP intl extension
• Reduced unnecessary database queries and content generation during activity processing

Bug Fixes​

• Fixed generated threads being published without rechecking forum permissions on concurrent updates
• Fixed disabled or rescheduled triggers executing from outdated background jobs
• Fixed fake account registration settings, email domain validation, and website field handling
• Fixed automated reply tracking incorrectly counting regular replies or processing threads that reached their limits
• Fixed concurrent reactions, follows, and content watching producing incorrect activity results
• Fixed concurrent AI content generation, pool management, and recovery from failed operations
• Fixed activity processing exceeding execution time limits
• Fixed activity execution and scheduling issues involving concurrent runs, time limits, trigger settings, and action limits
• Fixed activity actions incorrectly handling content visibility, permissions, user selection, and targeting restrictions
• Fixed thread creation, replies, reactions, follows, content watching, and third-party integration activities

Smarter AI-Powered Fake Activity!​

This update brings AI-generated content, simulated thread and post creation, a new activity dashboard, and expanded controls for managing fake activity. Together, these additions make activity simulation more realistic, transparent, and configurable.

New Features​

• Added support for any OpenAI-compatible chat completions API, including OpenAI, DeepSeek, OpenRouter, Ollama, and LM Studio.
• Added fake thread and post creation using AI-generated content from the content pool.
• Added AI-generated profile information for newly registered fake users, including About, Location, Signature, and Custom Title.
• Added a background AI content pool that is automatically refilled every hour.
• Added an optional per-trigger AI prompt for thread and post creation activity, allowing content to be generated live instead of using the shared pool, with optional forum or thread context.
• Added richer thread context for AI-generated replies: the thread title, opening post and most recent messages are passed to the AI, so replies engage with the actual discussion instead of only its title.
• Added an option to open AI-generated replies with a quote of the thread's most recent message.
• Added a Global Prompt AI setting with extra instructions applied when refilling the shared background content pool.
• Added an AI Content Pool page in the admin panel, showing pool stock and usage rate per content type, with filtering by type and status, individual item viewing and deletion, on-demand generation, and a preview mode that shows what the current prompts produce without storing anything.
• Added optional per-forum prompt profiles for thread and post creation activity, so content for a given forum is generated from that forum's own topic instructions instead of the shared pool.
• Added a per-trigger content language, overriding the site-wide AI language setting.
• Added an Activity Dashboard with action summaries, daily charts, trigger statistics, and a recent runs log.
• Added activity logging for trigger runs, including actions performed, duration, and execution results (enabled by default with automatic cleanup).
• Added a per-trigger Maximum Actions Per Run limit.
• Added trigger cloning.
• Added a 14-day activity sparkline to the trigger list.

Improvements​

• Live AI generation now requests several items per API call and reuses them across the run, so a trigger posting multiple threads or replies costs a single request instead of one per post. The batch size and the per-run request ceiling are configurable per trigger.
• The background content pool refill is now adaptive: it measures how fast each content type is actually being used, spends a configurable request budget on whatever is closest to running out, and skips content types no active trigger draws from, including individual AI profile fields.
• Prevented fake users from reacting to the same content more than once during a single trigger run.
• Added a description for every activity type, shown when choosing or editing a trigger, so it's clear what each one actually does.

Bug Fixes​

• Fixed email generation during fake user registration so the configured fallback email domain is applied correctly.

Do you like this update?​

​

Please consider sharing your experience with a short review
Every bit of feedback matters!​

Click to expand...

Fixed missing controller parameters when triggering "content watch" activity

• Code refactor
• Added filter by content user for reaction activity
• Added user follow fake activity
• Added XFMG page view support
• Added content watch fake activity with XFMG, XenAddons/UBS and Truonglv/Groups support

2.0.4:

• Confirmed XF 2.3 support
• Improved trigger list display
• Fixed displaying incorrect execution time
• Fixed incorrect start / end time frame check
• Fixed reaction fake activity ignores first thread post

2.0.3 Patch Level 1:
Fixed inability to uncheck "randomize" option

• Code refactor & cleanup
• Fixed "max(): Array must contain at least one element" error in public page view fake activity handler (if no existing user meets the trigger conditions)
• Added new fake activity: user reactions for new content (requires enabled news feed)
• More informative online threshold option explain
• Added separate ACP admin permission to manage fake activity ("manage users" permission was previously checked)
• Added trigger test run button & elapsed time information
• Added CLI command to disable triggers:

php cmd.php xf-rebuild:ozzmodz-ufa-trigger-disable

2.0.1:

• Added filter by username for public page view handler
• Fixed add-on install error (BLOB "options" column can't have a default value)
• Removed unused "refine search" menu
• Fixed "any" usergroups criteria does not work for public page view
• Fixed handler class caching
• Added more detailed trigger options descriptions