Laya English
laya-english is for concise English messages. Its total budget is 512 tokens per question.
LAYA AI · GETTING STARTED
Laya AI documentation
Use Laya AI to make a focused decision about a short piece of text. This guide covers model choice, input size, answer shapes, and the path from a browser trial to your application.
Pick a model for the text you have
Laya AI offers two hosted checkpoints. Choose before writing the question: the input budget applies to each question, including the state, instructions, and labels.
Laya Multilingual
laya-multilingual handles multilingual short text with a 1024-token budget per question.
If a request exceeds the selected budget, the service rejects it. Shorten the state or labels rather than relying on silent truncation.
Try a short-text decision
Paste a compact message in the playground, select a model, then ask one focused question. For a damaged-item request, a choice question could ask whether the next step belongs to returns, delivery, or general support.
- Keep the state to the text that should influence the answer.
- Write labels that do not overlap; make uncertain cases visible in your review process.
- Read the selected answer together with its probabilities and token usage before you use it in a rule.
Three ways to ask
Yes or no
A noul question returns a yes probability for a single condition, such as whether a message needs manual review.
One of several labels
A choice question selects an option you provide and shows the distribution across all options.
Ordered levels
A score question rates text against an ordered scale you define, such as low, medium, and high urgency.
Use Laya where it fits your stack
Hosted API
Send model, state, and questions from your backend with a Laya AI key. The API guide includes a request you can run and a place to manage keys.
Local models
For an installation you operate yourself, follow the upstream code and model weights. Local setup is separate from this site's hosted account and credits.
Read the local setup guide →Check availability and credits
The playground checks whether the Laya GPU service is connected before a run. A successful hosted request uses one credit. If the service is disconnected, it returns an error before charging; a failed reserved run returns its credit. See pricing for the current account offer.