Roadmap. On-device and NPU routing are not implemented. The tested loop uses a scripted reasoner.
OhhO Mind
Give your robot a mind of its own.
- Category: Intelligence
- Accent: violet
- Live app: Open the console
Mind on this site is a prototype console plus the agent_engine and learning_engine packages in the public OmniBot repository. Cloud, on-device, and NPU routing is roadmap.
Overview (hero)
Mind is a prototype console. On-device and NPU routing are roadmap, not current behavior.
Highlights
- Prototype agent loop in agent_engine tests
- Scripted reasoner, not an on-device NPU
- Hardware-limit checks in the learning-engine verifier
What you get
- The public packages are
omnibot-ai-engines/agent_engineandomnibot-ai-engines/learning_enginein the OmniBot repository. - What you can run today is those unit tests and the simulated console.
- Actions are grounded on a position-controlled 6-DOF servo arm with a parallel jaw gripper; the safety gate checks kinematic joint limits, velocity bounds, and reachability. There are no force sensors or tactile skins on the arm today. Force limiting, tactile feedback, and contact-rich in-hand manipulation are Vision: not built. Pick-and-place task evaluation is In progress (OHH-101) with no fabricated success rates.
- On-device and NPU routing are roadmap. Over-the-air delivery to a fleet is roadmap.
Features
- The agent loop — A perceive → reason → verify → act loop exists as package tests, with a scripted reasoner.
- A safety gate it can't skip — Every low-level action is checked against kinematic limits — velocity, joint deltas, and reach on the position-controlled arm — before it reaches a motor. Force limiting and tactile feedback are Vision: not built.
- Hybrid reasoning — Roadmap. On-device and NPU routing are not implemented. The tested loop uses a scripted reasoner.
- Memory and learning — Roadmap beyond the package tests. Fleet OTA of a learned policy is not implemented.
How it works
- Read the packages — agent_engine and learning_engine are in the public OmniBot repository.
- Run the tests — That is the checked behavior.
- Open the console — The site console is simulated.
- Do not expect an NPU route — On-device and NPU routing are roadmap.
Specs
| Spec | Value |
|---|---|
| Reasoning | Roadmap: cloud, on-device, and NPU routing |
| Safety | Kinematic joint-limit + reachability gate (force/tactile sensing is Vision: not built) |
| Arm hardware | Position-controlled 6-DOF servo + parallel jaw (no force sensor / tactile skin) |
| Tested loop | Scripted reasoner in agent_engine |
| Delivery | Roadmap: fleet over-the-air update |
Plans
There is no paid plan. The console demo is free. Hosted cloud is a waitlist, not a checkout.
FAQ
How is this different from OhhO Autonomy? Autonomy and Mind are both prototype consoles. Mind does not decide a mission on a robot, and it does not route to an NPU.
Does it need the cloud to think? The tested loop can run with a scripted reasoner and no cloud key. A hybrid of cloud, on-device, and NPU models is roadmap.
Is it safe to let a robot decide for itself? The learning-engine verifier has hardware-limit checks in tests for kinematic joint and reach limits. Force limiting, tactile feedback, and in-hand manipulation are Vision: not built.
Can Mind feel grip force or perform in-hand adjustments? No. The physical robot has a position-controlled 6-DOF servo arm with a parallel jaw gripper and no force sensors or tactile skin. Force limiting, tactile sensing, and in-hand manipulation are Vision: not built.
How would a robot receive this later? Over-the-air delivery is roadmap.