GitHub · infrastructure
ollaya-dev/ollaya
Anyone who wants the local decision stack to feel like the Ollama workflow they already know.
What it does
Runs open decision models locally behind the same /v1/systemone shape Jev uses, so a client changes one environment variable: `ollaya run laya --preset triage` prints calibrated labels with probabilities in milliseconds. Presets cover triage-style classification, and further models are pulled by name like container images.
The repository's own description: “Run open decision models locally: pull and serve Laya, decider, NLI and GLiClass behind a TypeSafe-compatible API. Ollama for decision models.”
How it works
A Rust daemon serves ONNX-exported decision models — Laya, decider, NLI, GLiClass — and speaks TypeSafe's wire format, which is what makes an existing Jev client work unchanged. The models are fetched from a registry by name, which is the part that makes this a workflow rather than a script.
The repository, by the numbers
Read from the GitHub API on 2026-09-27. Stars and the last push move daily — quote them with the date, the way we do.
What we checked
- the install script and the run example in the README
- the model list and the wire-format compatibility claim
- the licence (Apache-2.0), the stack and the last-push date
- the Hacker News thread: 587 points and 29 comments when we read it
What we did not check. We did not install the daemon or point a client at it, so drop-in compatibility with TypeSafe's API is the author's claim, and the millisecond figures in the README are unverified here.