Skip to content
Documentation
Connections

Connect integrations

Manage website connections, server keys, deployment markers, and optional crawler collection from Profile.

Website connections

Open Profile and select a website. Integrations, API keys, and imports belong to that website. Check the selection before connecting a provider or creating a key.

Signup tracking, revenue tracking, Polar, GitHub, and crawler collection have separate setup controls. Turning a tracking integration off stops its collection; it does not delete existing reports.

Server keys

Create a key under Event API keys or in a tracking setup sheet. Keys are website-scoped, their secrets appear only once, and Talvess stores their hashes. Use server environment variables for the secret.

Creating a replacement does not revoke an existing key. Update each installation first, then revoke the old key explicitly. Revoking a shared server event key stops the event and crawler integrations using it.

GitHub deployments

Connect a GitHub App installation and choose one repository for the selected website. The App needs deployment access. Talvess records successful deployments after the connection begins and shows release markers beside traffic.

Deployments sync through webhooks and a rate-limited manual refresh. Disconnecting keeps the timeline already recorded. A before/after change describes an observed association; it does not prove the deployment caused the change.

Crawler collection

Open crawler setup from Profile and use its agent prompt or server examples. The browser tracker cannot see bots that never run JavaScript. Send request details from your server or edge to POST /api/crawlers using a website-scoped server key.

The endpoint expects a unique requestId, a pathname without query or fragment, a userAgent, and the original occurredAt timestamp. The key selects the website; do not add browser visitor context or a website ID to this payload.

Collect each request at one layer, use a supported background delivery hook, and keep failures separate from serving your website. Follow the examples in the setup sheet.

Bot labels come from supplied user agents and can be spoofed. A recorded request does not prove indexing, training, a citation, or a conversion. AI visitor referrals remain separate.