One snippet. Structured

signal.





CavBot starts with two scripts in your stack: Analytics v5 for typed runtime signal, and the CavAi Brain for reasoning and assist workflows. Together they power one operating model: signals, intelligence, and execution, without requiring user-level identity data.


This page shows the full loop: what you install, what each layer does, what is collected, and how execution surfaces run from real signal.



Choose a Plan View Product




System overview

What becomes available after install

CavBot is a clear operating pipeline: Analytics v5 captures structured signal, and the CavAi Brain turns that signal into usable intelligence.




Analytics + Brain · multiple surfaces

Install layer

Install

Two script tags: Analytics v5 writes typed route and recovery events, while the CavAi Brain enables reasoning workflows.

Events

Typed

Append-only rows: eventName, route, timestamp, and a compact properties map.

Derived views

Rollups

Aggregation turns raw signal into route health, diagnostic trends, structure insights, and escalation-ready context.

Surfaces

Command Center

Dashboard views, CavAi workflows, the site badge, and arcade recovery all read from the same schema.



Step 1 · Installation

Install Analytics v5 + CavAi Brain

CavBot starts with two scripts. No build step, no SDK ceremony. Place both in your layout, template, or HTML shell so they load across the routes you want under guard.



Tracker snippet panel

Required attributes: data-project-key and data-site. Add the CavAi Brain script after Analytics v5. Optional: data-badge="true" to mount the badge.


Analytics + Brain snippet
<script
  src="https://cdn.cavbot.io/sdk/v5/cavai-analytics-v5.js"
  data-project-key="YOUR_PROJECT_KEY"
  data-site="YOUR_SITE_ID"
  data-badge="true"
  defer
></script>
<script
  src="https://cdn.cavbot.io/sdk/brain/v1/cavai.min.js"
  defer
></script>

  • Placement: in <head> with defer (recommended) or before </body>.
  • IDs: anonymousId is project-scoped; sessionKey groups activity windows.
  • Defaults: no emails, names, or account identifiers required to operate.




Step 2 · Surfaces


How CavBot surfaces connect in production

CavBot stays visible where it should and quiet where it should. Teams review priorities in Command Center, visitors see clear guard signals, and recovery moments stay measurable.


Surface 01


Command Center signal view

Operations

A calm view of operational signal — diagnostic drift, recovery behavior, structure health, and runtime quality — pulled from the event stream. Built to answer: “Where do journeys break?” and “What changed?”


Route health SEO snapshots
Runtime feel

cavbot_route_view cavbot_seo_snapshot cavbot_perf_sample


Command Center


Surface 02


CavBot Badge

Optional

A fixed, minimal guard presence that signals a session is under protection. It has its own interaction stream so it stays useful without becoming distracting.


Configurable Non-intrusive Guard signal

cavbot_badge_impression cavbot_badge_hover cavbot_badge_click




Surface 03


Arcade recovery surface

Recovery

A designed 404 surface that converts dead routes into measurable recovery signal — visibility into abandonment, referrers, and where the journey breaks.


Catch / miss Idle telemetry Recovery rate

cavbot_404_view cavbot_catch cavbot_miss cavbot_idle



Step 3 · Data model


From structured events to coordinated action

CavBot keeps the schema small on purpose. Most of the value comes from derived tables and typed insights — not from large payloads. The pipeline below mirrors how Command Center stays consistent across projects.



Pipeline

Layer 1 · Events


Append-only event stream keyed by anonymousId, sessionKey, route, and time.


events sessions

Layer 2 · Aggregates


Rollups for daily route counts, referrer breakdowns, SEO snapshots, and performance samples.


daily_page_aggregates referrer_aggregates

Layer 3 · Insights


Typed findings: 404 spikes, SEO regressions, and slow critical pages — designed for action.


404_spike slow_critical_page

Layer 4 · Surfaces


Dashboard views read the same tables; CavAi workflows can reason over insights.


Command Center /v1/assist

Example event (minimal payload)

{
  "timestamp": "2025-01-01T12:34:56Z",
  "anonymousId": "a6d0…",
  "sessionKey": "s3…",
  "eventName": "cavbot_route_view",
  "route": "/pricing",
  "properties": { "pageType": "marketing" }
}

Step 4 · Privacy & retention

Production-safe defaults

CavBot is built for operational signal. IDs are anonymous and project-scoped. Payloads are intentionally small. Retention can be configured per project to match your data posture.



IDs & identifiers


  • Project-scoped anonymousId (UUID) per visitor.
  • sessionKey groups events by activity window.
  • No names, emails, or account identifiers required by default.

IP & location


  • Optional IP handling; when enabled, truncation occurs before storage.
  • Coarse geography only for routing and anomaly detection.
  • No long-term storage of raw IPs as a default posture.

Retention controls


  • Raw events retained for an operational window (example: 90 days).
  • Aggregates can be retained longer for trends.
  • Project toggles can disable event types or performance sampling.

Support


Need rollout guidance?

The Help Center includes setup notes, troubleshooting, and a clear escalation path — built to match the same CavBot system language.



HELP CENTER