ai_insights

Overview

AI Insights brings analytics and notifications to your Drupal site's AI
assistant conversations. It uses LLM-powered classification to automatically
discover what users are asking about, tracks topic trends over time, and
can alert you when things change.

Built on top of
AI Chatlog and the
AI module ecosystem.

Submodules

AI Insights Topics

Automatically classifies AI conversations into topics using your
configured LLM provider. A configurable "granularity prompt" controls
whether topics are broad categories or fine-grained sub-areas. Topics are
stored as taxonomy terms, created on-the-fly as the classifier discovers
them.

• Topic dashboard — top topics ranked by volume,
  trend comparison between date ranges, weekly timeline, and a chat vs
  search split view.
• Chatlog integration — topics appear as a column
  and filter on the conversation list, and in the conversation detail
  header.
• Admin tools — re-tag conversations after
  changing the prompt, merge duplicate topics, delete unwanted topics.
• Drush command —
  drush ai-topics:tag for manual or catch-up runs.

AI Insights Alerts

Configurable notification rules that evaluate on cron and send email
alerts when topic activity matches your criteria.

• Topic Alert — notify when specific topics
  appear above a threshold.
• Volume Spike — notify when a topic trends up
  compared to its baseline.
• New Topic — instant notification when the
  classifier discovers a previously unseen topic.
• Digest — periodic summary of all topic
  activity.

Each rule supports configurable frequency (hourly, daily, weekly),
recipient lists (email addresses and/or Drupal roles), source filtering
(chat only, search only, or both), and a daily email cap to prevent
flooding. A notification log tracks all sent alerts.

How classification works

1. On cron, the TopicTagger finds conversations not yet classified.
2. User messages are sent to the LLM with a prompt that includes the
   granularity instructions and the list of existing topics (to prefer
   reuse).
3. The LLM returns a JSON array of 1–3 topic names.
4. Topics are matched to existing taxonomy terms or created as new
   ones.
5. Mappings are stored in a custom table linking thread IDs to term
   IDs.

Requirements

• Drupal 10.3+ or 11
• AI Chatlog
• AI (with a
  configured chat provider)

Installation

1. Install with Composer or place in
   modules/contrib/.
2. Enable:
   

   drush en ai_insights ai_insights_topics
       ai_insights_alerts

3. Configure the granularity prompt and AI provider at
   Configuration > System > Topic Intelligence.
4. Run drush ai-topics:tag to classify existing
   conversations.
5. View results at Reports > AI Topics.
6. Create notification rules at Configuration > System >
   Topic Notifications.