ollaya
Rust daemon and CLI that pulls open decision models like Laya and serves them locally behind a TypeSafe-compatible /v1/systemone.
Overview
ollaya runs open decision models locally the way Ollama runs LLMs. A decision model reads a state plus typed choice, score and noul questions and returns calibrated probabilities in one forward pass. ollaya pulls models by name, serves them from a local daemon, and exposes POST /v1/systemone, /v1/decisions and GET /v1/models wire-identical to TypeSafe, so the official SDK works by setting TYPESAFE_BASE_URL=http://localhost:11435. A native /api/decide adds routing information and timings.
Models include Laya, Von, Kev, Decider, Nimble, Winnow and others, each with its author's calibration. ollaya never re-hosts weights: it publishes small ONNX graphs that read the original weights from the author's Hugging Face repo, pinned and sha256-verified, or runs authors' GGUF files on llama.cpp. The README recommends winnow:e4b at 0.722 accuracy on typed decisions against Jev's 0.738, with 89 ms for five questions on an RTX 4090; Laya is suggested for CPU-only machines.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | intent | — |
| Yes / no | is_urgent / refund_requested / churn_risk | yes · no |
| Score | frustration | — |
How it works
- Install and ollaya pull a model by name.
- ollaya serve starts the daemon on port 11435.
- Point existing Jev clients at it via TYPESAFE_BASE_URL.
- Use presets like triage or your own questions.
Strengths
- Wire-identical to TypeSafe, so no client changes.
- Weights fetched from original authors and verified.
- Single binary with Ollama-style commands.
Caveats
- Open models trail hosted Jev slightly on the README's comparison.
- 4B-class models need an NVIDIA GPU for low latency.
Getting started
curl -fsSL https://ollaya.dev/install.sh | sh, then ollaya run winnow:e4b --preset triage "<message>".At a glance
- Type
- Service
- Status
- beta
- Language
- Rust
- Runtime
- daemon, CLI
- Has code
- yes
- Stars
- 1,305
Use cases and integrations
How curated lists describe it
Local runtime: serves open decision models behind a wire-identical `/v1/systemone` endpoint, so an existing Jev client only has to point `TYPESAFE_BASE_URL` at the daemon, and reports its recommended model at 0.722 accuracy against Jev's 0.738 on typed decisions.
yibie/awesome-jevOllama-style runtime for decision models: pulls and serves open encoders and decoders (Laya, Von, Kev, Decider, Nimble, Winnow, and more) with each author's calibration, behind `/v1/systemone`; the official TypeSafe Python SDK works against it unchanged. Site: [ollaya.dev](https://ollaya.dev).
AnotiaWang/awesome-decision-modelsRun open decision models locally: pull and serve Laya, decider, NLI and GLiClass behind a TypeSafe-compatible API. Ollama for decision models.
wh000wh000/awesome-jev-liveRust daemon and CLI that pulls open decision models (Laya, decider, NLI, GLiClass) by name and serves them locally behind TypeSafe-compatible `/v1/systemone`, `/v1/decisions`, and `/v1/models` endpoints, so existing Jev clients switch with `TYPESAFE_BASE_URL`; the `laya` router picks an English or multilingual checkpoint per request, and answers come from these open models, not from Jev.
cobanov/awesome-jevRun open decision models locally — pull, run and serve Laya and other open Jev alternatives behind a TypeSafe-compatible API.
hellogumbo/awesome-jevPulls and serves open decision models locally the way Ollama serves language models.
MrJev/awesome-jev
Frequently asked questions
Do answers come from Jev?
No, they come from the open models ollaya serves; only the API shape matches TypeSafe.
Which model should I start with?
winnow:e4b with an NVIDIA GPU, or laya on CPU, per the README.
Does ollaya host the weights?
No, it reads them from the authors' Hugging Face repos, pinned and checksum-verified.
Related projects
Go harness for System One models: hosted Jev, open Kev in-process via llama.cpp, and Cloudflare Clef, with typed helpers and a CLI.
ServicebetaGoTypeScript library for one-line local decisions with open models (kev-0.6b, kev-4b, open-jev), no API keys or cloud.
ProjectbetaTypeScriptSwift 6 bridge that runs System One decision models behind Apple's Foundation Models API, on device with Laya or via hosted endpoints.
ProjectbetaSwiftZero-dependency npm package that runs Laya natively or as WebAssembly in Node, Bun and browsers, drop-in for Jev's /v1/systemone.
ProjectbetaJavaScriptLocal-first desktop AI workbench for llama.cpp, vLLM and SGLang that includes a JEV typed-decision page and /v1/systemone gateway.
ServicebetaRuns the open LAYA decision model under Bun with native ONNX inference behind a TypeSafe-compatible API, plus an English/Persian playground.
LibrarybetaTypeScript