Ingest
The device gateway writes the frame to Kafka, and only then acknowledges the device.
fleet.raw.teltonika
HOW IT WORKS
Your data runs in your cloud.
Your data runs in your cloud.
We run everything around it.
StreamSource puts managed Kafka, connectors and your processing code inside your own AWS, GCP or Azure account. Our control tower handles deployments, upgrades and health, without ever touching your events.
Deploys in about 25 minutes
No inbound ports to open
Payloads never leave your VPC
01Architecture
One platform, two places
Everything that touches your data runs in your account. The only thing in our cloud is the control tower that tells it what to deploy.
STREAMSOURCE CLOUD
Control tower
- Dashboard & API
- Releases
- Health
- Billing
Sources
Telematics devicesTeltonika, iTriangle over TCP
IoT sensorsMQTT
DatabasesPostgres, MySQL change capture
Apps & APIsHTTP ingest
Your cloud account · AWS / GCP / AzureData stays here
Transform
No-code utilitiesTransform
Your own codeGit→Build→Test→Deploy
Managed Kafka3 brokers · replicated 3× · encrypted
RAWfleet.raw.teltonika
EVENTSfleet.evt.position
STATEfleet.state.latest
DEAD LETTERfleet.dlq.decoder
Schema registryBlocks breaking changes
AutoscalerScales on consumer lag
MonitoringGrafana & Prometheus
AgentApplies approved changes
Sinks
Databases & warehousesPostgres, BigQuery, ClickHouse
Object storageS3, Google Cloud Storage
Queues & streamsSQS, Pub/Sub, Kinesis
APIs & webhooksAny HTTP endpoint
Components never call each other. They only read and write topics, so the pipeline you see on the canvas is drawn from what's actually running.
02The life of an event
From a vehicle on the road to a row in your database
Follow one GPS position through a fleet pipeline. Every step is safe to retry, so a crash causes a repeat, never a gap.
Decode
A transform turns bytes into a typed, schema-checked event with a stable ID.
fleet.evt.positionEnrich
Geofence and alert rules add context and raise events like overspeed.
fleet.evt.alertsDeliver
The sink upserts by event ID, so retries update the same row instead of adding one.
postgres · positions
03The boundary
What runs where, and what crosses the line
Splitting the control plane from the data plane is what lets us manage the platform without ever holding your data.
CONTROL PLANE · OUR CLOUD
We coordinate
The dashboard, API and release manager that decide what should run.
- Workspaces, users and roles
- Deployment requests and versions
- Health summaries and alerts
Crosses the boundary
↓Deploy instructions
↑Health and status
✕Never your events
DATA PLANE · YOUR CLOUD
You own
Everything that reads, stores or moves your data, inside your VPC.
- Kafka brokers and topic storage
- Sources, transforms and sinks
- Secrets in your cloud's secret manager
04Deployment
Set up once.
Set up once.
Run everything from the dashboard.
Connect a cloud account with one template. After that you only open your cloud console for billing or account settings.
What gets created: Kubernetes clusterKafkaSchema registryAgentDashboards
05Reliability
Built for things going wrong
Networks drop, databases slow down and bad data arrives. Each case has a defined behaviour.
At-least-once delivery
Nothing is acknowledged until it's stored. A failure causes a retry, never a silent gap.
Stable event IDs
Every event carries an ID that survives retries, so sinks drop duplicates safely.
Dead-letter topics
Bad records are set aside with the reason attached, ready to inspect and replay.
Scales on lag
When a component falls behind, it gets more replicas. When it catches up, they go away.
06Shared responsibility
Know who handles what
You keep the decisions that belong to your business. We take on the work of running streaming infrastructure.
| Area | Your team | StreamSource |
|---|---|---|
| Cloud account & bill | Owns | — |
| Data & access policies | Decides | Enforces |
| Pipeline logic | Builds | Provides tools |
| Kafka & upgrades | Approves | Runs |
| Monitoring | Responds | Collects & alerts |
Does any of my data pass through StreamSource?
No. Events are produced, stored, processed and delivered inside your cloud account. The control tower only receives health metrics and deployment status.
What happens if your control tower is down?
Your pipelines keep running, because moving data never depends on us. You can't deploy changes from the dashboard until it's back, but the ssctl CLI still works against your cluster.
Can I keep my existing sources and destinations?
Yes. Pick a ready-made connector, or package your own as a container that reads or writes Kafka. It deploys and scales like a built-in one.
What do I need before deploying?
An AWS, GCP or Azure account, permission to create one scoped role, and a rough idea of your event volume and retention so we can size the cluster.
See it running on your own data
Start on the free sandbox, or book 30 minutes and we'll sketch your pipeline with you.