OORB RUNFP / ROBOT RELIABILITY SOFTWARE
Every run leaves evidence. Make it useful.
OORB RunFP is local-first robot reliability software. It records run provenance, compares MCAP-backed episodes, and supports outcome-rate and regression checks so robotics teams can make release decisions from recorded evidence.
Available through OORB onboarding. Bring your robot task and existing run data.
- Task label
- pick-place Matched
- Model identity
- baseline → candidate Changed
- Physical task outcome
- Not supplied Not scored
A change to investigate. Review the recorded evidence against your task criteria.
THE RUNFP WORKFLOW
From a recording
to a reviewable decision.
Keep the context. Inspect the changes. Choose the check that answers your release question.
Keep the run and its context together.
Capture MCAP-backed episodes with available software, hardware, robotics, and model provenance. Trace recorded outcomes alongside workspace commits.
Run provenanceSee what changed before you ship.
Compare baseline and candidate episodes. Inspect recorded environment and behavior differences with explicit, task-specific criteria.
Run comparisonsMake the release criteria explicit.
Report grouped outcomes with confidence intervals, inspect recorded state coverage, and gate new recordings against a named baseline.
Release checksA PRACTICAL FIRST WORKFLOW
Start with two recordings.
Build a baseline.
Initialize the robot project, import comparable runs under one label, then inspect the difference. Use the IDs returned by capture in the comparison command.
Follow the complete quickstart →runfp init --name robot-evaluation --root .
runfp snapshot baseline.mcap --label pick-place
runfp snapshot candidate.mcap --label pick-place
runfp diff <baseline-id> <candidate-id>WORK WITH YOUR EXISTING EVIDENCE
From recordings to your release workflow.
Bring telemetry into stored, comparable episodes.
RECORDED RUNSv → vModel provenance ↗Record local models or Weights & Biases artifacts.
MODEL IDENTITY↺Observed Real2Sim ↗Prepare recordings and robot assets for replay checks.
ISAAC SIM + ROSmcpCodex & Claude Code ↗Inspect saved run evidence through local MCP tools.
LOCAL MCPA comparison pass Recorded evidence met the comparison criteria.
A task success rate Needs validated outcomes and a defined task.
State coverage & replay Describe observed variables and recorded behavior.
CHOOSE YOUR WORKFLOW
Start with the question
your team needs to answer.
START WITH ONE ROBOT TASK