Skip to Content

How do I record config changes so an AI SRE can see them?

Tares records config changes through a second webhook source whose event type is config, so each change is stored as an event with the same labels as your deploys. You post one JSON object per change to the source’s ingest URL from your pipeline or tooling. The labels you declare, such as service, let you correlate a config change with deploys for the same service. The AI SRE guide covers how Tares uses the events you ingest.

How it works

The webhook connector accepts arbitrary JSON or NDJSON posted to a source’s ingest endpoint and stores the payload lossless. You declare labels to map payload fields into the envelope, and you mark one label as primary.

For config changes, you add a second webhook source next to the one that carries your deploys:

  • The source sets a fixed event_type of config, so every event it receives is typed as a config change and is easy to tell apart from a deploy.
  • A text_template renders a readable line from payload fields, for example the change and who made it.
  • A service label maps the payload’s service field into the envelope. Use the same label name your deploy source uses, so both kinds of event carry the same key to correlate on.

Because the payload is stored lossless, fields you do not map into labels, such as the old and new values, stay in the stored event.

Set it up

  1. In the console, go to Sources → Add source → webhook, or ask a connected agent to create a webhook source for your config changes. You can also POST /api/sources with the same object. The response includes the ingest_key.

  2. Declare the source. This one fixes the event type to config and labels each event by service:

- name: config-changes connector: webhook config: event_type: config text_template: '{change} by {sender}' labels: - { name: service, field: service, primary: true }
  1. Save the source and copy the ingest endpoint from the source page.

  2. Make your config tooling post one JSON object per change to /ingest/<ingest_key>. A Git hook, a deploy script or a cron job can all send it:

curl -X POST http://localhost:8787/ingest/<ingest_key> \ -H 'Content-Type: application/json' \ -d '{"change": "raised pool size to 50", "sender": "ci", "service": "web"}'
  1. If your Tares has auth on, the sender uses a Bearer Authorization header holding an ingest-scoped API key. On an instance without auth, you do not need it. Creating a push source in the console mints a key for it and shows it once.

  2. Check the mapping. Post a sample payload to the ingest URL, then open the source page and read Recent events and Fields. Recent events shows what arrived. Fields shows which payload fields arrive and how often, which helps you pick a reliable label. The test_source tool takes no payload for a push source and only returns the ingest endpoint, so it does not check the mapping.

  3. Give your deploy source a service label with the same name, so both sources use the same key for the same service.

What you get

  • Every config change is stored as its own event with the event type config, separate from your deploy events.
  • Each event carries the service label you declared, so you can correlate config changes and deploys by the same key.
  • A readable text line per event from your text_template, and the full payload kept lossless.
  • A source that anything able to send an HTTP request can feed, including CI systems, cron jobs and your own services.
  • A source page where Recent events and Fields show what arrives, so you can fix the label mapping early.

When something else fits

  • If the change is already a commit or pull request in a GitHub repo, the GitHub connector brings those in directly, and you do not need a webhook source for it.
  • If you want to record a one-off note from an agent rather than a stream of changes, the memory connector holds an agent’s own observations.

Still stuck? Get help

Last updated on