Software, LLMs, and robot chatbots

Talk to the robot: language in, the backend connected to AI, allowed actions out. Also web apps, dashboards, and vision — on a dedicated server in the EU we can run. Work is remote only from a laptop. Contact.

A vision model is as good as your labelled photos. We ship confidence scores and an explicit “unclear”. Models that must run on the chip, with no cloud, are on-device AI — same people, on the firmware page.

EU hosting

A dedicated server in the European Union. HTTPS, nightly backups that have been restored, monitoring, security updates, email support on working days. We can sign a processing agreement when we hold data for you.

New features are a separate milestone, quoted before we touch them. You own the application. If you leave, we move it to a server in your name and hand over the keys.

Talk to the robot

Talk-to-action: someone says or types what they want, the backend is connected to AI, the robot does something it is allowed to do. Chat over a serial port, BLE, or HTTP to an arm, a rover, or a fleet of ESP32 nodes. Tool calling with a tight allow-list: move joint, read sensor, stop. Not a chatbot that invents G-code.

The model can live on a machine you control in the EU, or on-device when the radio is down — same people as on-device AI. We do not paste customer data into a US API unless you ask for that in writing.

Payments, ledgers, latency

The same code when the job is a ledger or a hot path: something is too slow, something has to be exactly right, or customer data is somewhere it should not be. C++ and Rust in the path that counts. Local models on a machine you control in the EU when a US API is shipping text off the continent.

Contact

Email services@olivabot.com.

Write to us

To services@olivabot.com. NDA first if you prefer.