Analytics

Analytics is the daemon’s current operational view. It is not a replacement for source data, workspace backups, or a monitoring system.

Operational endpoints

These endpoints require analytics permission in authenticated deployments:

Endpoint	Use
`GET /api/analytics/usage`	Request, actor, provider, and connector counters.
`GET /api/analytics/errors`	Recent errors; supports stage, time, and limit filters.
`GET /api/analytics/latency`	Latency summary.
`GET /api/analytics/logs`	Recent structured daemon log entries.
`GET /api/analytics/memory-safety`	Memory safety metrics.
`GET /api/analytics/continuity`	Continuity state.
`GET /api/analytics/continuity/latest`	Latest continuity summary.
`GET /api/telemetry/events`	Recorded telemetry event view.
`GET /api/telemetry/health`	Telemetry collector state.
`GET /api/timeline/*`	Timeline investigation routes.

Start with a bounded query and correlate it with daemon status and diagnostics:

curl -fsS http://127.0.0.1:3850/api/analytics/errors?limit=20
curl -fsS http://127.0.0.1:3850/api/analytics/latency
curl -fsS http://127.0.0.1:3850/api/diagnostics

Counters and in-memory buffers are runtime observations. They can reset when the daemon restarts. Use durable logs, database state, and the affected source when investigating a persistent incident.

Telemetry

Telemetry is configured at memory.pipelineV2.telemetryEnabled and defaults to enabled. Persistently opt out with:

memory:
  pipelineV2:
    telemetryEnabled: false

For a process-local opt-out without editing configuration:

SIGNET_TELEMETRY_OPTOUT=1 signet daemon start

The daemon writes recorded telemetry events to:

$SIGNET_WORKSPACE/.daemon/telemetry/events.jsonl

PostHog delivery is best-effort and depends on telemetry configuration. The local JSONL audit is the inspection surface; it lets an operator see what was recorded without treating a remote dashboard as the source of truth.

Telemetry is designed to avoid memory content, user identity, file paths, and raw secrets. The audit log is still workspace-private operational data: do not commit it or paste it into a public issue without review.

Investigation workflow

Check daemon state and /health/ready.
Pull a bounded recent error and latency view.
Identify the affected request, session, source, provider, or queue.
Inspect the original source and daemon logs.
Use Diagnostics for health or repair decisions.
Re-run the same bounded query after a change to verify the outcome.

Related: Daemon, Diagnostics, Configuration.