dashboard  /  train

Compile a .kolm in your account.

Pick a base. Drop in your examples. Set the K-score gate. Hit compile. The artifact and receipt land in your account when it passes the gate.

privatesigned in as your-account·compile credits: view balance

Step 1Pick a base model

The pretrained weights you start from. Recipe is portable; you can swap the base later.

Step 2Upload your examples

JSONL, CSV, or a folder of text files. The compiler will dedupe, split, and report the K-score on a held-out fold.

Step 3Set the gate

The artifact will not sign if these thresholds fail. They go in the manifest and travel with every receipt.

Step 4Watch the run

Live training log streams here once you start. Preview run is a simulated walkthrough; Run for real hits POST /v1/compile against this account and streams the receipt chain back.

[--:--:--] waiting to start…