What exactly is Laya Goish?

It is a Laya inference engine and HTTP server written in Rust (Goish Rust) that turns English and Thai text into structured decisions — routing tickets, scoring urgency, detecting refund requests — running entirely on your own infrastructure.

Which question types can I ask?

Three: choose returns a choice from your options with probabilities and confidence, score returns a score against your ordered criteria, and check returns the probability that a condition is true.

How do I start the server?

Download the v0.1.0 Linux release, run bash scripts/download-models.sh f16, then ./laya serve models/laya_english_f16.gguf --port 8080 --threads 4. The HTTP API answers on /v1/decide, and docs/http.md plus examples ship in the repository.

What are the hardware requirements?

Linux x86-64 with AVX2 and FMA. Model weights are not bundled — they are downloaded separately by the provided script — and the engine ports Laya inference from ggmlc.

Can it handle Thai as well as English?

Yes. Upstream Laya handles English, and OpenThai-SystemOne v0.3 covers Thai and English. Both can be loaded into one server and selected per request, so one API serves both languages.

Can I integrate it without HTTP?

Yes — the README lists a command line and a persistent JSON-lines process alongside HTTP, and batch requests let you apply the same questions to many inputs. Optional bearer authentication can protect the decision API.

Does it need an API key or a hosted service?

No key is needed: inference runs on your machine's CPU with no hosted service required. The repository also declares no license file, which is worth checking before production use.

Jev directory

CommunityJev Repos & SDKs

Laya Goish

Laya GGUF inference and an HTTP server written in Goish Rust: one Linux binary runs upstream Laya or OpenThai-SystemOne on your CPU and serves typed choose, score and check decisions over HTTP, the CLI or a JSON-lines process.

See highlights Open github.com

Stars: 17
Language: Rust
Reading time: 2 min
Published by: Community
Author: cogentica-ai
Last verified: Oct 2026

cogentica-ai/laya-goish

Organization repository on GitHub

View on GitHub

Laya GGUF inference and HTTP server in Goish Rust

Forks: 7
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import madewithjev gguf rust cpu-inference

Works forJev for Routing Jev for Scoring Jev for Classification

Used inJev for Customer Support Jev for Software & DevTools

Languages: Rust
Platforms: CLI API
Capabilities: Open sourceRuns locally

Highlights

A single Linux executable performs inference on your CPU, on your own infrastructure, with no hosted inference service in the loop.
Serves upstream Laya for English and OpenThai-SystemOne v0.3 for Thai and English; load both models and pick one per request.
Answers typed questions: choose (a choice with probabilities and confidence), score (against ordered criteria) and check (probability a condition is true).
Integrates over HTTP at /v1/decide, the command line or a persistent JSON-lines process, with batch requests and optional bearer authentication.
Requires Linux x86-64 with AVX2/FMA; the v0.1.0 release matches a script that downloads the f16 GGUF weights separately.

Quickstart

bash

curl http://localhost:8080/v1/decide \
  -H 'Content-Type: application/json' \
  -d '{"model": "laya", "state": "I was charged twice. Please refund the duplicate payment.", "questions": {"refund": {"type": "noul", "instructions": "Does the customer ask for money back?"}}}'

Watch out

Grounded in the repository README and GitHub metadata reviewed on 2026-10-15; the repository declares no license file, publishes no benchmark numbers, and targets only Linux x86-64 with AVX2/FMA, with model weights downloaded separately.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

The Open Reimplementations: Local Models That Speak System One

Sep 22, 2026

Closed weights have not stopped the ecosystem from rebuilding the shape in the open. API-compatible servers, retrained heads on small models, and inference-time tricks that skip generation entirely.

Jev Repos & SDKs#open-models#research#local

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai