Prior investigations

What it's for

After every ticket fix, whatever the AI figured out — the root cause, whether or not it led to a code fix — is embedded and stored, scoped to your organisation. On the next ticket, before reasoning, the most similar past discoveries are retrieved and handed to the model alongside the usual gathered context, so a recurring issue is recognised instead of investigated from nothing every time.

There's nothing to configure beyond the connector below, and nothing to review — this quietly improves what the model has to work with, the same additive role Grafana and Confluence context already play.

Requires an OpenAI connector — for embeddings, not necessarily for reasoning

This is powered by OpenAI's embeddings API specifically, since it's the one embedding endpoint with a stable, widely-compatible shape among this product's five AI provider types. Add an OpenAI connector to enable it, even if your organisation reasons with Claude, Gemini, or a private-hosted endpoint — the OpenAI connector only needs to exist and hold a valid key; it doesn't have to be the active reasoning provider.

Without an OpenAI connector configured, this feature is simply inactive: no discoveries are stored, none are retrieved, and every ticket fix behaves exactly as it would otherwise.

What's not built yet

Today the only source is your own ticket-fix history. Feeding this from your Confluence pages or repository structure directly — so it has something useful to say even before any tickets have been resolved — is a planned extension, not yet available.