Ingest
PhyWare receives and retains normalised events.
Typical use Customer-owned telemetry and unrestricted operational systems.
Acquisition layer
PhyTrace Get your data into the model
SDKs and managed connectors that capture source data, normalise it into PhyUDM, and handle buffering, retry and transport — so integration is not the reason a project stalls.
Instrument once
PhyTrace is critical adoption infrastructure, not the whole value proposition. It exists so that normalising a new source is a day of work rather than a quarter.
Python SDK
Typed event builders, domain helpers, agent configuration, reliability and transport. Install with pip and start emitting validated PhyUDM.
Rust SDK
The same contract with a C-compatible FFI layer, for embedded and performance-sensitive integrations. Install with cargo.
ROS 2 bridge
Subscribe to existing ROS 2 topics and normalise them into PhyUDM without rewriting your robot stack.
Not every source can be copied
Third-party API terms frequently restrict bulk harvesting, storage and redistribution. Pretending otherwise produces integrations that must be torn out later, so the mode is selected per data class and contract — not per vendor.
Federated query
PhyCloud queries the source on demand and returns governed results.
Typical use Restricted vendor systems and records that cannot be bulk copied.
Reference
PhyCloud stores identifiers, metadata and authorised deep links — not the record.
Typical use Vendor-hosted media, evidence and sensitive records.
Derived event
PhyCloud stores a permitted normalised fact derived from a source record, with lineage.
Typical use Alerts, availability, health and correlation metadata.
A data lake is not permission to copy every connected record. Restricted sources contribute only permitted derived events or external references, while the underlying record stays with its owner and is retrieved on demand.
Every connector documents what it does
An undocumented integration is a liability. These are the properties each PhyTrace connector declares before it is considered production-ready.
- Supported source and event types
- Field-to-PhyUDM mapping and schema version
- Authentication and authorisation requirements
- Polling, webhook or streaming behaviour
- Expected latency and freshness
- Rate limits and retry semantics
- Data ownership and retention constraints
- Allowed storage, derivation, sharing and export behaviour
- Source timestamp and clock-quality characteristics
- Degradation behaviour and gap reporting
- Connector version and mapping provenance
Which source is blocking you?
Tell us the vendor and the data you need. We will tell you which connector mode their terms actually permit.