AI Index Health monitors the freshness and coverage of AI vector indexes built on Search API. It finds the silent failures that quietly degrade AI Search and RAG results, and lets you fix them with a targeted re-embed instead of a costly full reindex.

The problem it solves

When content changes after it was last embedded, or when indexable items are never tracked, your vector index drifts out of sync with your content — and the AI answers built on it become quietly wrong, with no error to warn you. Search API's own tools don't surface this; AI Index Health does.

What it detects

Stale embeddings — items whose content changed after they were last indexed.
Coverage gaps — indexable items that were never tracked and so will never be embedded.
Dimension and model mismatches — when the embedding model or vector size the index was built with no longer matches the configured one, with optional end-of-life signals from AI Model Registry.

Features

Targeted re-embed — queues only the affected items through the Search API tracker, not a full reindex.
Drush command — drush ai-index-health:report prints per-index stale, coverage and mismatch counts plus the affected item ids; CI-friendly.
Admin dashboard — the same report in the UI, with a one-click re-queue action.
Works on any Search API index; AI-specific checks activate when AI Search is present.

Requirements

Drupal 10.3+ or 11
Search API

Optional integrations

AI Search — enables embedding dimension and model mismatch detection.
AI Model Registry — end-of-life and dimension metadata for the models an index was built with.

Installation

Install with Composer:

composer require drupal/ai_index_health

Enable the module, then run drush ai-index-health:report or visit the dashboard to review index health and queue targeted re-indexing.

AI Index Health