ARCHITECTURE
Architecture for engineers who've been burned before.
No black boxes. No data exfiltration. No "trust us, it's secure." Here's exactly how Klaritics moves - and doesn't move - your data.
0 copies
of raw event data outside your warehouse
1 planner
that turns product questions into SQL
100%
deployment control in your infrastructure
audit-ready
query behavior for security review
Selected stage
Data Sources
SDKs, APIs, and existing warehouse tables feed the pipeline.
Notice what's missing? A second copy of your data - and a vendor cloud. Klaritics deploys inside your environment and queries your warehouse in place.
Sources: connect your existing event ecosystem.
Use Klaritics with your current instrumentation model. You can ingest new streams or start from tables you already trust.
- Web SDK (JS, lightweight, <12kb gzipped)
- Mobile SDKs (iOS, Android, React Native, Flutter)
- Server SDKs (Python, Go, Node, Java, Ruby)
- HTTP API for custom producers
- Existing warehouse tables (point and analyze with no re-instrumentation)
Ingestion: stream and batch, your choice.
Stream events through your existing pipelines or batch-load through current ETL workflows. We support Kafka, Kinesis, Pub/Sub, and direct warehouse writes.
Processing: identity, sessions, schema - handled.
Important: Processing runs as transformations inside your warehouse. You see every row. You can audit every step.
- Identity resolution (anonymous → user ID stitching)
- Sessionization (configurable session windows)
- Schema mapping (any event shape → Klaritics canonical model)
- Backfills and replays (retroactively reshape without losing history)
Storage: your warehouse, full stop.
We don't have a storage tier. There is no Klaritics cloud and no Klaritics data lake. Your warehouse is the storage layer; Klaritics deploys next to it inside your environment.
Analytics engine: where events become answers.
Our query planner translates UI interactions (build a funnel, a cohort, a retention curve) into optimized SQL that runs on your warehouse. We cache aggressively at the query layer; we never cache your raw data outside your warehouse.
UI: built for product and data teams together.
Drag-and-drop event builders, dashboards, alerts, embeds, and a SQL escape hatch when the UI gets in your way.
Where Klaritics runs
Klaritics deploys inside your infrastructure. The full stack - ingestion, query engine, dashboard UI - runs in your Kubernetes cluster, your Docker host, or directly on bare-metal Linux. We ship the software; you run it. There is no Klaritics cloud, no shared tenancy, and no outbound data path.
Want the full architecture PDF?
Get the technical brief with deployment diagrams, trust boundaries, and query lifecycle notes used in architecture reviews.