Monitoring health

A deployed instance is observable through the CLI’s status and health commands, plus Cloudflare’s own Workers logs and traces (enabled in wrangler.jsonc observability).

Instance status

notewright status

Reports the running core version, the composed notes and followUp provider ids, and a row per Durable Object marking each healthy or errored. A version skew between the CLI and the deployed core shows up here.

Per-agent health

notewright follow-ups health     # follow-up agent state + unresolved failure count
notewright speakers health       # speaker-context agent state + cached versions

Schedules

notewright schedules

Shows the registered cron per Durable Object and the next run time. If a schedule is missing, the agent has not started since deploy; run notewright start.

Failures

The follow-up agent records a failure row per meeting that could not be processed, with the retry count and last error:

notewright failures list                 # unresolved failures with error detail
notewright failures clear --meeting <id>  # clear one row
notewright failures clear --all --yes     # clear every row

To clear a failure and retry it in one step, use follow-ups run <id> --force, which clears the failure and run marker (keeping task mappings) and re-enqueues the meeting.

Cloudflare-side observability

wrangler.jsonc enables Workers logs (including invocation logs) and traces. Use the Cloudflare dashboard or wrangler tail to follow structured log lines such as follow_up_backstop_completed, follow_up_manual_run_requested, and noop_follow_up_created when diagnosing behavior beyond what the CLI surfaces.

Scripting with JSON

Every command accepts --json, emitting { ok, ... } or { error: { code, message } }. Combine it with the CI escape hatch to run health checks without a linked instance:

NOTEWRIGHT_DISABLE_KEYCHAIN=1 NOTEWRIGHT_URL="$URL" NOTEWRIGHT_SECRET="$SECRET" \
  notewright status --json