Skip to content

AI Analysis

Chris Bassey edited this page Aug 25, 2026 · 1 revision

AI Analysis (optional)

The plugin can generate a natural-language summary/triage note for an alert or case using an LLM provider you configure. It is off by default and bring-your-own-key.

Read Security Model before enabling this — it explains exactly what data does and does not leave your network.

Configure

Workbench → Settings → AI analysis:

  1. Toggle Enabled.
  2. Choose a provider: OpenAI, DeepSeek, Gemini, or Anthropic.
  3. Paste your API key (stored server-side; see below).
  4. Optionally set a model (e.g. gpt-4o-mini) and a custom base URL (for gateways / self-hosted-compatible endpoints).
  5. Save.

Using it

Open an alert or case flyout → AI Analysis tab → Generate. The result is stored on the document so it can be viewed later without regenerating.

Key storage

The API key is held server-side and is not returned to the browser (the field shows a masked placeholder once configured). For at-rest encryption of the stored key, configure the plugin's encryption settings; the Settings tab indicates whether encryption is configured.

What gets sent

Only a projected, allowlisted subset of an alert leaves the network — never the raw document. Comment bodies are never sent (only a count), and long fields such as full_log are truncated. See Security Model#ai-egress for the full boundary.

Clone this wiki locally