Laya Provider Guide

The second provider of decide — the same typed boolean / choice / score answers as TypeSafe's Jev, from an open-weights model you can run yourself. It emits no text at all.

Overview

Laya is Convai Innovations' Apache-2.0 "System One" decision model. You send one state plus named, typed questions; an encoder answers every question in a single forward pass and returns a probability distribution for each. NeuroLink calls whatever base URL you configure — a Laya server, or a LiteLLM proxy with a pass-through route to one. There is no built-in endpoint.

Key Facts


Inference type	`decide` only
Checkpoints	`typed-decisions` (default, fine-tuned for typed decisions), `multilingual`, `english`, or `auto`
Input window	1,024 tokens (`typed-decisions`, `multilingual`); 512 (`english`)
Questions per request	up to 64
Endpoint	`<base URL>/predict`; the base URL is required, with no default
Precedence	used automatically when its key and base URL are set and no TypeSafe key is

Quick Start

1. Get an endpoint and a key

Run a Laya server (see below), or use a LiteLLM proxy with a pass-through route to one. On LiteLLM, create a virtual key and add the route's /predict path (and /health, if you want to probe it) to the key's Allowed Routes; a key without them is refused with a 403 Key/team not allowed to access passthrough route.

2. Configure

Both the base URL and the key are required. Set them in the environment:

export LAYA_BASE_URL=https://your-proxy.example.com/laya  # NeuroLink calls <base>/predict
export LAYA_API_KEY=sk-...                                # the key your endpoint accepts
export LAYA_MODEL=typed-decisions                         # optional

or in the config passed to the SDK, exactly as for any other provider. Values passed per call override the constructor's, which override the environment:

const neurolink = new NeuroLink({
  credentials: {
    laya: {
      baseURL: "https://your-proxy.example.com/laya",
      apiKey: process.env.MY_LAYA_KEY,
    },
  },
});

3. Use it

import { NeuroLink, readDecisionChoice } from "@juspay/neurolink";

const neurolink = new NeuroLink();
const result = await neurolink.decide({
  provider: "laya",
  state: "We were billed twice for March. Please refund it today.",
  questions: {
    team: {
      type: "choice",
      instructions: "Which team should handle this?",
      criteria: {
        billing: "Payments and refunds",
        technical: "Bugs",
        sales: "Pricing",
      },
    },
    urgent: { type: "boolean", instructions: "Is this urgent?" },
  },
});
const team = readDecisionChoice(result.answers, "team");

From the CLI

neurolink decide "We were billed twice for March." --provider laya \
  --questions '{"urgent":{"type":"boolean","instructions":"Is this urgent?"}}'

When NeuroLink uses it

Every built-in consumer of decide — model routing, relevance-driven compaction, tool routing and RAG planning — asks for the default decision provider. That is the first one that is configured — in the environment or in the credentials passed to the SDK — and TypeSafe comes first. TypeSafe has two keys, TYPESAFE_API_KEY and AI_GATEWAY_API_KEY (its Vercel AI Gateway route), and either one counts. Laya counts only with both its key and its base URL. So:

A TypeSafe key, plus Laya's key and base URL: built-in features use TypeSafe; Laya runs only where a caller asks for provider: "laya".
Only Laya's key and base URL: built-in features use Laya.
A Laya key with no base URL: Laya is not configured. Built-in features ignore it, and provider: "laya" fails with Laya requires a base URL.
None of them: everything behaves exactly as it did without a decision model.

Limits

The window is small. Laya's encoders read 1,024 tokens, and 512 on the english checkpoint, against roughly 33,000 for Jev. Part of that is reserved for each question and its options, so NeuroLink allows about 768 tokens of state on typed-decisions and multilingual, and 320 on english, auto and any model name it does not recognise. Laya's server does not refuse a longer state — it answers from the start of it and says nothing — so NeuroLink refuses it instead, before any network call, with max_tokens_exceeded.

The size is an estimate, not Laya's tokenizer: about four characters per token for ASCII text, and 1.5 tokens per character for other scripts (0.6 on multilingual), calibrated against a live Laya 0.3.5 server. It errs toward refusing. Built-in consumers treat a refusal as "carry on as before", which means long-prompt model routing usually falls back to the heuristic when Laya is the only decision provider.

Many options degrade it. A choice's options share a fixed token budget, so accuracy drops past about 20 options; some Laya servers also reject a question whose options overflow the budget, as invalid_request.

Pick the checkpoint deliberately. Laya's own benchmark puts its general english and multilingual checkpoints close to chance on typed decisions without fine-tuning; typed-decisions is the fine-tuned one, which is why it is the default.

Errors

Reply	Kind	Retried
401 / 403 from the proxy	`authentication` — the provider instance stops retrying	no
400 / 422 from Laya	`invalid_request`, with Laya's reason	no
413 from Laya	`max_tokens_exceeded`	no
429	`rate_limit`	yes
503	`overloaded`	yes
other 5xx	`server`	yes
more questions or state than Laya reads (local)	`max_tokens_exceeded`, with no network call	no
no base URL configured (local)	`invalid_request`, with no network call	no

LiteLLM's 401 text echoes a masked copy of the rejected key and its hash; the provider drops that part before the message reaches a log or an error.

Running your own Laya server

Laya's example server exposes the same /predict route. Point LAYA_BASE_URL at it:

pip install "laya[serve]"
python examples/server.py            # from the Laya repository; serves :8000
export LAYA_BASE_URL=http://127.0.0.1:8000

Troubleshooting

Laya requires a base URL — set LAYA_BASE_URL, or pass credentials.laya.baseURL. There is no default endpoint.
Authentication failed — the proxy rejected the key. Check it in the LiteLLM Dashboard, including its Allowed Routes. The provider instance does not retry after a rejection.
A 403 with error code: 1010 — Cloudflare in front of your proxy blocked the client by its signature. Node's fetch is accepted; a proxy or agent that rewrites the user agent may not be.
exceeded the provider's token limit — the state is larger than Laya reads. Shorten it, or configure TypeSafe for long inputs.
Built-in routing never uses Laya — a TypeSafe key is also set and takes precedence, or Laya has no base URL.

Overview​

Key Facts​

Quick Start​

1. Get an endpoint and a key​

2. Configure​

3. Use it​

From the CLI​

When NeuroLink uses it​

Limits​

Errors​

Running your own Laya server​

Troubleshooting​

See also​