Structured JSON logging¶

RTB logs through tracing. For machine-parseable output, install the JSON formatter; for opt-in usage events, use rtb-telemetry. Both redact secrets automatically.

JSON-formatted logs¶

rtb-cli wires tracing-subscriber for you. To force JSON output (e.g. for log shipping), the subscriber honours the standard RUST_LOG filter plus the tool's JSON-output flag. Spans and fields serialise as one JSON object per line:

{"timestamp":"2026-05-24T09:00:00Z","level":"INFO","fields":{"message":"deploy started","target":"prod"},"target":"mytool::deploy"}

In tests, never install the subscriber globally — use tracing_subscriber::fmt().with_test_writer() or the tracing-test crate.

Telemetry events¶

rtb-telemetry emits opt-in usage events to pluggable sinks, keyed by a salted, hashed machine ID (the raw machine ID never leaves the host). It applies rtb-redact automatically to the args and err_msg fields of every event, so a command line carrying a token is scrubbed before it reaches any sink.

Redaction is automatic — but extend it for free-form strings¶

The telemetry path and the HTTP middleware redact known fields for you. For any other free-form string you send to an external surface, route it through rtb_redact::string first:

tracing::info!(url = %rtb_redact::string(&raw_url), "fetching release");

rtb-redact strips URL userinfo, credential query params, Authorization headers, and well-known token prefixes (sk-, ghp_, AIza, sk-ant-…, …). See its component page for the full rule set.

OTLP / distributed tracing¶

rtb-telemetry can export to an OTLP collector for distributed tracing — endpoints and sinks are configured per the telemetry spec.