Logging

Logs are structured, correlated, and free of sensitive data.

Rules

Emit structured logs (JSON in deployed environments, readable text locally).
Every request gets a request ID, generated at the edge or accepted from an upstream header, and every log line within the request includes it. Include the user ID once authenticated.
Name business events <feature>.<entity>.<action>[.<detail>], for example story.created or auth.login.invalid_credentials. Stable names make logs searchable.
Log at boundaries: use case start/finish for important operations, external calls, and errors. Do not log inside tight loops.
Never log secrets, tokens, passwords, or full personal data. Log identifiers, not payloads.
Log an exception once, where it is handled, with the stack trace. Do not log and re-raise at every layer.
Batch jobs include batch_id in every entry.

Level	Use
`DEBUG`	Developer detail, off in production
`INFO`	Normal business events
`WARNING`	Recoverable surprises (retries, degraded dependencies)
`ERROR`	A request or job failed

Error tracking (Sentry), tracing, and metrics are in Observability.