User repository on GitHub
The nano tier of local Jev — /v1/systemone-compatible typed decisions on a 7M-param CPU model (8MB, no GPU). jev-local drop-in.
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Models & Reimplementations
Jev Nano
The nano tier of local Jev: a /v1/systemone-compatible typed-decision server — noul, choice and score questions with probabilities — running a 7M-parameter model CPU-only from an 8MB tarball. Point any Jev/TypeSafe SDK client at it via base_url.
- Stars
- 0
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- javimosch
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceRuns locallySelf-hosted
Highlights
- Runs noul / choice / score typed decisions with probabilities on a 7M-parameter CPU-only model shipped as an 8MB tarball; no GPU needed.
- Exposes a /v1/systemone-compatible endpoint on 127.0.0.1:8097, so any Jev/TypeSafe SDK client can call it by changing base_url.
- Written entirely in MFL via machin: model runtime, server and decision heads compile to one static native binary, with no Python in the inference path.
- Install is a curl | bash script plus ./demo.sh, or a manual tarball from the mtlm-router v0.3.0 release for Linux x86_64.
Quickstart
bash
curl -fsSL https://raw.githubusercontent.com/javimosch/jev-nano/main/install.sh | bash
./demo.shWatch out
README reviewed 2026-10-17; GitHub reports Python as the repository language while the README insists there is no Python in the inference path, and the digest contains no benchmark or quality numbers beyond the MIT tag in metadata.
Questions about this entry
Short answers drawn from the linked project or page.
The nano tier of local Jev: a typed-decision server compatible with the /v1/systemone API, answering noul, choice and score questions with probabilities from a 7M-parameter model in an 8MB tarball.
It speaks the same /v1/systemone contract, so the README says to point any Jev/TypeSafe SDK client at it via base_url; it is a local, tiny stand-in rather than the hosted model.
A Linux x86_64 box and nothing else: no GPU, no Python and no dependencies beyond the appliance tarball. The model and server are a single static native binary.
Run curl -fsSL https://raw.githubusercontent.com/javimosch/jev-nano/main/install.sh | bash, then ./demo.sh; alternatively download the mtlm-router v0.3.0 Linux amd64 tarball, extract it and run ./start.sh.
POST JSON to http://127.0.0.1:8097/v1/systemone with a state, model jev-latest and a questions map, for example a noul question; the README's curl example shows the exact request shape.
GitHub reports Python as the repository language, but the README says the runtime, server and decision heads are written in MFL and compiled through machin's C backend, with no Python in the inference path.
Unknown from the digest: no benchmarks, model-quality results or comparisons are captured, and the repository has 0 stars. The author positions it as a drop-in nano tier for jev-local, nothing more.
More Jev resources like this
A LoRA plus 24-slot decision head over Qwen3.8-27B from Ming Xu's train-jev project: it serves TypeSafe's System One API with 0.851 Kev transfer-v4 accuracy and ECE 0.053, and needs an 80 GB GPU.
Jev Models & Reimplementations#community#open-models#local
Community
A 24 GB-GPU alternative from the train-jev project: Qwen3.5-4B plus a rank-16 LoRA and 24-slot decision head scoring 0.647 PK accuracy, 40.4 Decision Index and 43 ms p50 latency, ahead of Kev-4B at the same size.
Jev Models & Reimplementations#community#open-models#local
Community
A CPU-first decision model from Decision Models that answers the seven questions an LLM auto-router needs (category, difficulty, stakes, four flags) in one ONNX pass; MIT weights plus a Jev-compatible hosted API.
Jev Models & Reimplementations#community#open-models#local
Community
A Laya fine-tune for coding decisions: roughly 9k rows from SWE-Gym teach keep/drop file relevance, priority scoring and injection detection, reporting 0.833 accuracy on 1,150 held-out rows at 126 ms p50.
Jev Models & Reimplementations#community#open-models#local
Community
A Laya fine-tune from Convai Innovations on the 1,200-case typed-decisions benchmark: it reports 0.777 accuracy on the official 400-case test set, above TypeSafe Jev 1.13.0 at 0.727 and the teacher ceiling at 0.735.
Jev Models & Reimplementations#community#open-models#local
Community
A 9B bilingual Turkish and English Jev-class decision model from Berkay Hasret: a 173 MB LoRA adapter on Qwen/Qwen3.5-9B that answers up to 16 typed questions about one text in a single forward pass and reports 0.783 macro accuracy on full JevBench.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.