Connectivity Hub

Connected isn’t the same as
working.

Live health, throughput and freshness for every connection feeding Lumi. A pipe
that’s authenticated but silent looks fine in most tools — here it’s classified, flagged
with the reason, and fixable in one click.

⇅Connectivity9

⌕

Dashboard figures are illustrative and stream as a demo. Real thresholds are set per source.

STATE CLASSIFICATION

Five states, not two.

Up/down hides the failures that actually hurt. A connection can be authenticated,
reachable and completely useless — so Lumi classifies on what’s arriving, not just
whether the socket opens.

Flowing normally

Authenticated, delivering, and inside expected latency and volume.

freshness < 60s · errors 0

Flowing, but poorly

Still delivering, with elevated latency, retries or a throughput drop against baseline.

throughput < 60% baseline

Connected, silent

The connection is technically fine but no new data has arrived. The dangerous one.

freshness > 10m

Actively failing

Auth rejected, rate limited, or schema mismatch. The reason is named, not just the failure.

errors > threshold

No events at all

Nothing received in the window. Shown with a flat chart and no recent events so it can’t be mistaken for quiet-but-fine.

0 events / 6h

DATA FRESHNESS TRACKING

The failure nobody notices.

A stale feed doesn’t throw an error. It just quietly stops updating,
and every decision downstream is made on yesterday’s picture —
by an agent that has no idea it’s reading old data.

1

Age on every connection

Seconds since the last accepted record, not since the last successful handshake.

2

Per-source thresholds

Metrics should arrive every few seconds; a nightly CMDB sync shouldn’t be judged the same way.

3

Downstream blast radius

Which skills, agents and dashboards are reading this feed right now.

4

Stale data is marked, not used

Agents see the age and discount it, rather than reasoning confidently from stale input.

FRESHNESS BY CONNECTIONthreshold-aware

Datadog · metrics

3s ago
≤ 60s

Kubernetes · dc-west

8s ago
≤ 60s

ServiceNow · tickets

2m ago
≤ 5m

Dynatrace · traces

14m ago
≤ 60s ⚠

CMDB · nightly sync

6h ago
≤ 24h

PER-CONNECTION ERROR DETECTION

Named causes, not “connection
failed”.

Every failure class is detected separately, because each one has a different fix — and
only some of them are yours to fix.

Credential expired or revoked

Token lifetime ended, secret rotated, or the scope was withdrawn upstream.

→ one-click re-auth

Throttled by the provider

Requests rejected because the API quota is exhausted for the window.

→ auto backoff + resume

Payload shape changed

The upstream vendor changed a field and records are being dropped on parse.

→ mapping update flagged

Unreachable or TLS failure

Route, firewall, proxy or certificate problem between Lumi and the source.

→ path diagnostics run

Queue depth growing

Queries exceed the window; often the source is under load, not the connection.

→ retry with wider window

Queue depth growing

Data arriving faster than it’s processed — throughput is fine but lag is climbing.

→ scale ingest workers

CAPABILITIES

What the hub monitors.

◉

Connection health monitoring

Continuous checks on every feed, scored on delivery rather than handshake.

⇅

Live throughput tracking

Events per second and volume per connection, streaming in real time.

▦

State classification

Healthy, degraded, stale, error or offline — never just up/down.

⏱

Data freshness tracking

Age since the last accepted record, judged against per-source thresholds.

▤/div>

Integration status dashboard

Every connection, its state and its numbers on one screen.

⚠

Per-connection error detection

Auth, rate limit, schema, network, timeout and backlog identified separately.

⟳

One-click reconnect

Re-auth and resume from the row, with recovery confirmed by real throughput.

🔗

Downstream impact

Which skills and agents depend on the feed that’s struggling.

📈

Baseline comparison

Throughput judged against this connection’s own normal, not a fixed number.

WHY IT MATTERS

Bad data is worse than no data.

No data is obvious — someone notices the empty dashboard. Stale data looks
completely normal, and everything built on top of it quietly starts being wrong.

SILENT FEED FAILURE

found days later

flagged in minutes

Freshness thresholds catch what error logs never show.

TIME TO RECONNECT

a support ticket

one click

Re-auth from the row, verified by throughput resuming.

DECISIONS ON STALE DATA

unknowable

prevented

Agents see data age and discount it accordingly.

ROOT CAUSE OF A FEED FAULT

“connection failed”

named class

Auth, quota, schema or network — each with its own fix.

👤

For the ITOps team

DAY TO DAY

✓

Trust what you’re looking at

Every number carries its age, so you know whether it’s current before you act on it.

✓

Fix it from the row

Expired token? Reconnect, watch throughput resume, move on.

✓

Stop debugging the wrong layer

A named error class tells you immediately whether it’s your problem or the vendor’s.

✓

Know what’s affected

See which skills and agents go quiet when a feed drops.

◈

For the organization

ON THE BALANCE SHEET

✓

Automation you can rely on

Agents acting on stale data is the fastest way to lose trust in automation entirely.

✓

Protect the observability spend

You already pay for these tools — a broken feed means you’re paying and not receiving.

✓

Fewer false incidents

Stale inputs generate phantom problems that cost real hours to chase.

✓

Evidence for data quality

Freshness and completeness records, which regulated environments increasingly ask for.

Dashboard figures are illustrative and stream as a demo. Real thresholds are set per source.

Get started

Find out which of your
feeds is already stale

Connect your existing sources and we’ll show you their real freshness
and throughput on day one.

One control plane for hybrid and multicloud operations — agents, orchestration, and observability in one place.

Build

Orchestrate

Cowork

Resolve

About Us

Autonomous ITOps platform powered by an AI Coworker, SRE Orchestrator, and Agent Builder

Coworker

Chat instead of ten dashboards — your coworker does the work.

Orchestrator

One engine coordinating reliability, security, and cost.

1

2

3

4

Build

Orchestrate

Cowork

Resolve

About Us

About Us

Product

Coworker

Orchestrator

Product

Orchestrator

Coworker

Build

Platforms

Estate Models

Agents

Observability

Service Desk