01
Connect the run
Robot video, force, robot state, logs, and audio on one timeline.
Know what your robots physically did, and prove it.
Identify contact and impact events from the video, force, and logs you already record, with evidence attached.
haptica · run 0042local
00:08.70contact2 sensors · Vision Sensor + force thresholdHigh
00:12.41impact1 sensor · force thresholdHigh
00:17.05contact1 sensor · Vision SensorMedium
00:21.30unknownabstained · preservedUnknown
Confidence band, never a decimalsuggested region, not point
Evidence window · blue contact · pink impact
The record
One record of what physically happened.
02
Review contact and impact
Detections aggregate into one event per moment. The record states how it was found.
03
Share the evidence
Every event links its evidence: source windows, confidence band, provenance.
Jobs
What teams do with the record
Exception triage
Alert to reviewed event, no engineer rebuilding the timeline. Evidence attached.
Safety and behavior review
Every contact and impact across the fleet. One queue, one schema.
Physical regression
Did contact rates move run over run? Filter the ledger and see.
Training-data curation
Accepted events are labeled examples with provenance. Export the set your model needs.
Open source
Start with Haptica Core.
The open app, schema, and review workflow.
- Local-first Review Room for robot runs
- Open contact and impact event schema (physical-events v0.3)
- Manual, rule, threshold, acoustic, and plugin sensors
Production
Add the platform when it matters.
For cells where the event is visible but not directly sensed.
- Haptica Vision Sensor for video contact and impact
- Edge runtime and deployment calibration
- Evidence retention, controls, and support
Standards
Built to sit inside the stack you have.
Schemaphysical-events v0.3 · open repo, validators, typed SDKs
MCAPRides inside MCAP as a channel convention
LeRobotEpisode adapter for public datasets
Data rightsPrivate by default · contributor mode is explicit