Skip to content

Run diagnostics

Two checks exist. A connection test is one fixed, repository-free provider round trip using a built-in read-only assist envelope, so it needs no profile choice. A validation run exercises a full implementation and independent review on an isolated sample using the profiles you pick. Both consume provider turns (requests against the runner’s provider subscription); heartbeats remain the passive health signal.

  1. Open Diagnostics and pick a Target: a provider (any compatible runner) or one exact runner. Only runners that are online, authenticated, and active are offered.
  2. Choose Run connection test.

Provider targeting checks scheduler behavior. Exact-runner targeting verifies one deployed process. One connection test runs at a time per workspace. Results appear under Recent connection checks with the job status, profile, provider, model, and runner.

  1. Open Runs. Pick an Implementation profile and Implement with provider, and a Review profile and Review with provider.
  2. Choose Run validation.

The run uses an isolated Python sample with deterministic tests, schema-validated output, and an independent review with fresh context. It never receives repository or deployment access. Recent workflows list each stage (Implementation, Independent review) with status, verdict, routing, and any error.

Job statuses are Waiting, Queued, Running, Succeeded, Failed, and Cancelled. A job that a Developer cancels from a change page stops at its next safe point. See Change and run states.