Hey, human!
HTTP/2 200 OK content-type: text/markdown; charset=utf-8 vary: accept, user-agent x-agent: [ai_agent]
Loading the agent response…
Write a recovery message for a recurring error
Attach next-step guidance to an error agents keep hitting, so it comes back with the failure.
A recovery message is guidance attached to a specific failure, so a caller that hits it is told what to do next instead of only what went wrong. They pay off most on the errors that repeat across many callers.
This identifies a recurring failure from real traffic, reads one instance closely enough to write accurate guidance, and attaches it to that error's fingerprint. These change what live callers see, so write them against a failure you have actually inspected.
Steps
Restless runs these seven in order with your stored key, and stops for your approval before each one that changes data.- 1
Confirm the identity and access behind the API key with GET /api/v1/me.
- 2
Pick the target project. If you only have an example slug, verify it with GET /api/v1/projects/{slug}, or browse options via GET /api/v1/projects.
- 3
Find the recurring failure with GET /api/v1/projects/{slug}/logs, filtering on status and urlContains, and confirm it is one repeated condition rather than several stacked together.
- 4
Read one instance in full with GET /api/v1/projects/{slug}/logs/{logId}, so the guidance you write matches the error the caller actually sees.
- 5
Check what is already attached with GET /api/v1/projects/{slug}/recovery, so you update the right fingerprint instead of adding a duplicate.
- 6
Write the guidance with POST /api/v1/projects/{slug}/recovery, sending fingerprintKey, message (what to do next, addressed to the caller), and strategy. Posting to a fingerprintKey that already exists replaces its message.
changes data - 7
If a message is no longer accurate, remove it with DELETE /api/v1/projects/{slug}/recovery, passing the fingerprintKey.
changes data