Inference Clients and Policy Server Setup

August 7, 2026 · View on GitHub

RoboLab uses a server-client architecture: your model runs as a standalone server process, and RoboLab connects to it through a lightweight inference client during evaluation.

For writing your own inference client, see Evaluating a New Policy. For the full run CLI reference, see Running Environments.

Shipped policy clients

BackendPolicy / modelReferences
pi0_family/π0, π0-FAST, π0.5, PaliGemma, PaliGemma-FAST (select with --policy)Code, Papers: π0, FAST, π0.5
cosmos3/Cosmos3-Nano-PolicyWebsite, Paper
gr00t/GR00T N1.7 DROID / GR00T N1.6 DROIDWebsite, Code, Paper
dreamzero/DreamZero-DROIDCode, Paper
volo/VoLoAgentWebsite, Code, Paper