Custom Analysis

Pass raw matching events to the LLM for freeform analysis.

Custom analysis is not available in the AI-assisted tracker builder. Create custom trackers using the YAML editor.

Overview

The custom analysis type passes a filtered list of raw events directly to the LLM for freeform analysis. Unlike structured types, it does not run pre-computed queries or inject metric variables — only the matching events and universal base variables are available. Use custom when none of the structured types fit your use case.

Configuration Reference

Prop

Type

RawEventsConfig Properties

Prop

Type

Available Template Variables

custom analysis only has access to the universal base variables — no pre-computed metrics:

VariableDescription
{{windowPeriod}}Human-readable analysis window, e.g. "last 7 days"
{{dataAsJson}}Raw event list serialised as JSON
{{executedAt}}ISO 8601 execution timestamp

Example

.journium/trackers/user-engagement.yml
apiVersion: journium.app/v0Beta
kind: InsightTracker
metadata:
  name: user-engagement
  displayName: User Engagement
  description: Tracks engagement patterns via raw event inspection
spec:
  type: LLM
  trigger:
    mode: automatic
    schedule: daily
  window:
    period: last_7d
  analysis:
    type: custom
    rawEvents:
      events:
        - signup_completed
        - habit_created
        - session_started
      minEvents: 1
      maxEvents: 50
  llm:
    promptTemplate: |
      Analyze user engagement for {{windowPeriod}}.
      Event data: {{dataAsJson}}
      Generate a one-sentence insight summarizing engagement patterns.
    maxOutputTokens: 200

How is this guide?

Last updated on

On this page