laya

Name: Laya AI
Author: DevBeast

laya is devbeast's open-weights system one decision model: it answers typed yes/no, choice and score questions about a piece of state in a single non-autoregressive forward pass, across 100+ languages, and never generates text.

devbeast/laya

input · usd / 1m tok

$ 00 free

output · usd / 1m tok

$ 00 free

context window

8.19k tokens

modalities in → out

T →

// live decision execution

test single forward-pass decision

state / context to evaluate

question / decision query

decision type

public mode • sign in for unique hits

Ready to execute. Sub-second response on this server...

// open-weights community · early access program

Join as Free Contributor — Free Future AI Access

Free Contributor Tier ● 100% Free Forever Early Access

Want to join our developer team or access our next-generation AI models for free? Register below as a Free Contributor. We are opening pre-release checkpoints, specialized routing weights, and future multimodal models to contributors at zero cost.

Register your email below. We will send you confirmation and direct email invitations with free API access keys whenever new AI model weights or features are deployed.

Contributor Benefits

✦ 100% Free access to future AI models

✦ Pre-release checkpoints before public launch

✦ Higher rate limits on decision endpoints

✦ Direct email announcements with credentials

// api reference & integration guide

laya Decision API

Single non-autoregressive forward pass decision model. Evaluates state context against boolean, choice, and score queries in milliseconds with zero hallucinations.

Core Endpoints

                    Host Base URL: http://187.52.117.2:8000
                

Method	Endpoint	Description	Auth
POST	`/query`	Execute decision query on state context	Header `X-API-Key` or Session
GET	`/test`	In-memory benchmark test (health check)	Public
POST	`/v1/systemone`	Raw typed multi-question decision engine	Header `X-API-Key`
GET	`/settings`	Model configuration & context window specs	Public

Implementation Code Examples

Supported Decision Types

1. noul / boolean (Boolean Probability): Calibrated Yes/No probability for binary statements. Returns summary.verdict ("Yes / True" or "No / False"), probability (0.0 – 1.0) and confidence score.

2. choice (Multiple Choice): Selects the best match from custom comma-separated options (e.g. option_a, option_b, option_c) with full probability distribution.

3. classification (Routing & Categorization): Directs tickets, leads, or content into categories (e.g. billing, tech_support, sales, account_security).

4. score (Ordinal Scoring & Severity Rubric): Assigns a continuous calibrated rating across ordered rubric tiers (e.g. very low, low, medium, high, critical or numeric scale).

5. sentiment (Sentiment Analysis): Classifies customer sentiment instantaneously (positive, neutral, negative) with zero text generation delay.

6. risk (Risk & Urgency Assessment): Evaluates operational or security threat level (critical, high, medium, low) for instant dispatch.

7. intent (Intent Classification): Maps user input to actionable intents (inquiry, complaint, upgrade, cancellation, feedback).