Use Fixtures and Runs

Save a fixture

1. Open Test in the Braid Inspector.
2. Enter the exact request body expected by API Input.
3. Save it with a scenario name such as valid-customer or missing-email.
4. Run it and record expected assertions in the Braid description or tutorial.

Fixtures are owned by one user and one Braid. They may contain sensitive business data, so use synthetic values whenever possible.

Read a run

The timeline separates run lifecycle events from node events. Each node reports status, duration, redacted slot summaries and structured errors. LLM usage may include token or cost metadata when the provider supplies it.

For a browser sink, the server captures a short-lived result and the editor renders it. The persisted audit trail stores availability and summaries rather than raw image or audio payloads.

Troubleshoot

• node_runtime_policy_violation: replace or remove a server-blocked Bead.
• braid_run_timeout: reduce external work or adjust an operator-controlled budget.
• 401 or 403 from a provider: verify the selected secret, provider access and model permissions.
• 204 with a successful run: the Braid intentionally has no public Output.