Limits
The sizes, counts and time windows of the service.
| Limit | Value |
|---|---|
| Size of one file | 200 MB |
| Files in one dataset | 500 |
| Public sources (Hugging Face, GitHub) in one dataset | 10 |
| Decisions built from one dataset | 12 |
| Length of one decision's question | 500 characters |
| Length of one adjustment to the agent | 1,000 characters |
| Text of one example | The base model's context (see Base models); longer examples are refused, never cut |
| Test examples measured against Jev | 20% of the labeled examples, up to 2,000 |
| Upload URL | Valid for 15 minutes |
| Download and audit links | Valid for 15 minutes; ask for a new one any time while the files are kept |
| Page of a list | 25 by default, up to 100 |
| Cost of a training | 1 credit, given back if the model does not beat Jev or the training fails. If you chose to keep the model even if it does not beat Jev, only a failed training gives it back |
| Coupons | One per purchase, typed under I have a coupon on Billing; it takes the place of the launch discount |
| Runs of the agent (structuring, suggestions) without a credit | 10 a day per account; with a credit, no limit; a run that fails on our side does not count |
| Examples pre-labeled before the training is paid | The held-out examples and a sample, up to 500; the rest when the training starts |
Images are not used in this version: they show up as Not used, with the reason.