@tanstack/ai-ollaya weekly downloads statistics

<div align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://tanstack.com/api/readme/ai.png?theme=dark" /> <source media="(prefers-color-scheme: light)" srcset="https://tanstack.com/api/readme/ai.png" /> <img src="https://tanstack.com/api/readme/ai.png" alt="TanStack AI" width="900" /> </picture> </div> <br /> <div align="center"> <a href="https://npmjs.com/package/@tanstack/ai-ollaya" target="_parent"> <img alt="NPM downloads" src="https://img.shields.io/npm/dm/@tanstack/ai-ollaya.svg" /> </a> <a href="https://github.com/TanStack/ai" target="_parent"> <img alt="GitHub stars" src="https://img.shields.io/github/stars/TanStack/ai.svg?style=social&label=Star" /> </a> </div>

@tanstack/ai-ollaya

Ollaya adapter for TanStack AI. Call the open-source laya decision models with decide() and typed questions, running fully local. This package talks HTTP with fetch. It does not add an SDK.

Install

pnpm add @tanstack/ai @tanstack/ai-ollaya

Setup

Ollaya runs open decision models locally — think Ollama, but for the laya decision models instead of chat LLMs. No API key required.

1. Install Ollaya. On macOS or Linux:

   curl -fsSL https://ollaya.dev/install.sh | sh
   

   On Windows, in PowerShell:

   irm https://ollaya.dev/install.ps1 | iex
   

   Downloads and the desktop app are on ollaya.dev/download.

2. Start the server. It listens on http://127.0.0.1:11435 by default — the same URL this adapter defaults to.

   ollaya serve
   

3. Pull a laya model (in another terminal):

   ollaya pull laya:latest        # router: picks laya:en or laya:multilingual
   # or a specific one:
   ollaya pull laya:en            # English (ModernBERT-large)
   ollaya pull laya:multilingual  # 100+ languages (mmBERT-base)
   

4. Verify the model is available and the server answers:

   ollaya list
   ollaya run laya:latest --preset triage "My payouts have been failing for 3 days"
   

That's it — the adapter below talks to the running server over HTTP.

Usage

import { boolean, choice, decide, score } from '@tanstack/ai'
import { ollayaDecider } from '@tanstack/ai-ollaya'

const result = await decide({
  adapter: ollayaDecider('laya:latest'),
  state: ticket,
  questions: {
    queue: choice({
      instructions: 'Which team should handle this ticket?',
      options: {
        billing: 'Payments, invoices, refunds',
        tech: 'Bugs, outages, integrations',
        sales: 'Pricing, upgrades, new accounts',
      },
    }),
    urgency: score({
      instructions: 'How urgent is this ticket?',
      levels: ['low', 'medium', 'high'],
    }),
    refund: boolean({
      instructions: 'Is the customer asking for a refund?',
    }),
  },
})

result.queue.value
result.meta.model
result.meta.usage

Point the adapter at a non-default host with the baseURL option:

ollayaDecider('laya:latest', { baseURL: 'http://my-host:11435' })

Supported models

• laya:latest — routes each request to laya:en or laya:multilingual by the text's script and language
• laya:en — English decision model (ModernBERT-large)
• laya:multilingual — decision model for 100+ languages (mmBERT-base)

You can also pass any other laya tag as a string.

License

MIT