QAN AI Insights¶
QAN AI Insights provides query optimisation and tuning guidance based on Query Analytics (QAN) data for a specific query pattern. It is focused on performance analysis (plans, indexes, rewrites) rather than general infrastructure chat.
Use ADRE Chat for open-ended questions; use QAN AI Insights when you are already in QAN and want AI help on that query.
Behaviour (conceptual)¶
- PMM sends QAN-related context to the configured HolmesGPT backend and returns an analysis.
- The server may cache the latest analysis per query identifier and service so repeated views do not always re-run a full analysis. Cached content has a timestamp; your UI may offer a control to refresh.
Settings¶
- Administrators can adjust the QAN insights prompt (within size limits enforced by PMM) so the model follows your organisation’s preferred format or policies.
- ADRE must be enabled and HolmesGPT URL configured for insights to work.
Privacy¶
Query text and metrics sent for analysis may contain schema, SQL, and application identifiers. Ensure HolmesGPT and network paths meet your compliance requirements.
API details for operators: see the QAN insights rows in the ADRE API table in dev/adre/README.md.