Operations
Health
GET /health always answers HTTP 200 with:
{"status":"ok","model_ready":false,"port":8080}The status code is deliberately unconditional. PULSE serves every non-AI
feature without a model, so a readiness probe here should catch a wedged or
unbound server, not an AI outage. Read model_ready for the model. It is true
only when the engine is running and has confirmed it can serve a request.
Self-healing
An in-process watchdog probes local resources and the inference and voice
endpoints, and a remediator reacts to what it finds.
SelfHealService/GetStatus shows it working: whether the loop is healthy, the
last event, recent events, and how many remediations were tried and failed.
healthy is false only when the most recent event is critical and was not
remediated. The service only observes. It never triggers remediation itself.
Remediation is automatic by default, from a fixed list of non-destructive actions: restart one of PULSE’s model runners, add bounded capacity to one, open the in-process circuit breaker, roll back the loop’s own most recent change to a runner when that change was made shortly before the fault, or notify. A plan outside that list or below the confidence floor becomes a notification. Nothing deletes data, rolls back a release, or touches a security control. See the self-healing operations agent card.
To turn automatic remediation off, set SELF_HEAL_AUTONOMOUS=false. Plans are
still reasoned and recorded, but none is applied.
Logs
Logs are OpenTelemetry-shaped JSON on stdout, with service.name
pulse-backend and the version from APP_VERSION. Lines worth watching:
| Log line | Meaning |
|---|---|
refusing to start: ... | A fatal startup condition. The rest of the line names it. |
AUTHENTICATION DISABLED — every request will be rejected as Unauthenticated | The session key is one that authz refuses. Fix PULSE_JWT_SECRET. |
PUBLISHCFG ... | What each outward channel is allowed to do |
grounding: no data/corpus directory | The product fact files are missing from the image; whitepapers are written without product facts |
gRPC reflection ENABLED | Reflection is on. It should not be on a deployed instance. |
security config validated, or a warning listing problems | The result of the production gate |
The production gate
At startup PULSE checks for insecure configuration:
- The session key is missing or refused.
CORS_ALLOWED_ORIGINSis unset or contains a wildcard. The backend serves no web bundle and the image no longer setsPULSE_WEB_DIR, so there is no same-origin exemption unless someone sets it by hand (do not).
With PULSE_ENV=production, either problem stops the server with refusing to start in production. Otherwise it is a warning.
A missing or refused session key is the exception: it stops startup on every
deployment, whatever PULSE_ENV says, because a server that cannot verify its
own sessions has nothing safe to do. What production adds is that the CORS
problem stops it too.
PULSE_ENV=production on the PulseControlPlane, so a
deployment applied from it refuses to start on an insecure configuration.
Before applying it to a running instance, confirm the current pod logs
security config validated; otherwise the new pod will not start. PULSE_ENV
has no other effect: it does not touch billing or plans, and reflection has its
own opt-in, PULSE_GRPC_REFLECTION.Audit pipeline health
Each stage of a one-line audit (Audit a domain)
reports to agent health as its own agent, pulse-audit-<stage>. Reports go
to TIDE_HEALTH_URL; with it unset there are none. A
stage that wedges stops reporting, so it shows as stale in the TIDE console
rather than idle.
Every run ends with one log line, audit <run id> (<domain>): <completed | partial | failed> <reasons>. A run is bounded at 45 minutes end to end, and its
diagnosis also keeps the PULSE_AUDIT_DEADLINE_SECONDS budget. Time a stage
spends waiting for model capacity counts against both.
TLS
Pick one:
| Setup | How |
|---|---|
| TLS at the edge (usual) | Nothing to set. The edge terminates TLS and speaks plain HTTP to PULSE. |
| Direct exposure | PULSE_TLS_CERT and PULSE_TLS_KEY. TLS 1.2 minimum, forward-secret AEAD suites only. |
| Service-plane mutual TLS | MESH_MTLS=true, with the mesh CA. Every client needs a mesh certificate. TLS 1.3 only. If the CA is unavailable, the server refuses to start rather than downgrade. |
Rate limiting
PULSE_RATE_RPS and PULSE_RATE_BURST cap calls per peer IP. Over-limit calls
get ResourceExhausted before any other work. Off by default.
Licence gate
With a licence configured (LICENSE or LICENSE_FILE, verified with
LICENSE_PUBLIC_KEY), every non-public call is refused with license expired or invalid when its signature is bad or it has expired. With no licence
configured, nothing is gated.
Where the data is, and backups
Everything PULSE keeps is in its appfs root, bucket pulse on the in-cluster
object store. Every object there is sealed under PULSE’s own key, derived from
TIDE_ENVELOPE_KEK. Back up that bucket with the object store, and keep the
KEK: without it the backup cannot be read.
Once encrypted rows in the new envelope format exist, do not roll back to a build from before that format: it cannot read them.