Organization repository on GitHub
High Performance OpenAI and System One Compatible CPU Edge Device Service
- Forks
- 4
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
OpenJev-Cactus
A FastAPI backend from SiliconLabAI that serves an OpenAI-compatible chat and tool-calling API plus a Jev-style POST /v1/systemone endpoint for typed decisions, powered by cactus-needle on CPU edge devices. The first request downloads the Needle weights once.
- Stars
- 6
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- siliconlabai
- Last verified
- Oct 2026
Highlights
- Two API generations in one app: OpenAI-compatible tool calling at /v1/chat/completions and Jev-style typed decisions at /v1/systemone.
- Tools are passed in the request body — no tools.json file needed.
- Powered by cactus-needle; the first request downloads the Needle weights automatically, one time.
- Quick start is a venv, pip install -r requirements.txt, and uvicorn main:app on port 8000.
- MIT-licensed, includes /v1/models and /health, and works with a TanStack/React front end or plain curl.
Quickstart
bash
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 8000 --reloadWatch out
README reviewed 2026-10-17; the digest truncates the tool-calling curl example mid-JSON, so the exact tool schema is not shown, and no latency, memory or accuracy figures appear for the CPU edge target. 'Jev-style' is the repository's own wording for a System One-shaped API, not the TypeSafe service.
Questions about this entry
Short answers drawn from the linked project or page.
A FastAPI backend from SiliconLabAI that exposes two APIs: an OpenAI-compatible chat and tool-calling surface, and a Jev-style POST /v1/systemone endpoint for typed decisions. It is powered by cactus-needle and aimed at CPU edge devices.
It replicates the Jev-style typed-decision request shape listed in the README's endpoint table: POST /v1/systemone returns structured answers rather than generated text. The digest does not show a complete request example for it, only the endpoint's row and purpose.
Create a Python virtual environment, activate it, pip install -r requirements.txt, then run uvicorn main:app with the host, port and reload flags from the README. The server listens on localhost:8000, and the first request downloads the Needle weights.
The README positions it for CPU edge devices and mentions no GPU, and there is no API key because the model runs locally through cactus-needle. The captured text gives no memory or latency requirements, so headroom on small devices is unverified.
In the request body — the README notes no tools.json file is needed. The example shows a chat completion request with a tools array, but the digest truncates that JSON mid-schema, so the full tool definition format is not captured here.
OpenJev is an Express and Vite playground with swappable backends; OpenJev-Cactus is a standalone FastAPI service serving an OpenAI-compatible API plus the System One endpoint on local hardware. Neither is the official TypeSafe service.
The SiliconLabAI organisation, shared with OpenJev. GitHub metadata: Python, MIT, 6 stars, 4 forks, created and pushed 2026-09-24, so the repository has not been updated since its first day.
More Jev resources like this
A System One layout optimizer for decision-model workloads: it reorders rows and fields so a prefix-caching backend reuses input computation, then returns per-record decisions in original order. Its bundled backend serves AutoTrust JEV-9B through vLLM.
Jev Repos & SDKs#community#github#python
Community
0814wdwd/solo_jevPython
0814wdwd/solo_jevPythonHTTP server for the Mica v0.5 4B Preview decision model: GGUF weights run through llama.cpp and answer noul, choice and score questions with calibrated probabilities over the TypeSafe /v1/systemone format.
Jev Repos & SDKs#community#github#python
Community
akivet/MicaPython
akivet/MicaPythonServing code for the Jeff 1.0 family of Jev-style decision models: a state, a typed question and its options go in, and one forward pass returns a probability for each option. Four open models from 4B to 31B parameters are served the same way.
Jev Repos & SDKs#community#github#python
Community
j-geuter/jeff-servePython
j-geuter/jeff-servePythonA frozen decision server that serves Gemma 4 31B — quantized to FP8 at startup — in TypeSafe's /v1/systemone format, reading the next-token distribution instead of generating tokens. It needs one FP8 NVIDIA GPU with 80 GB and has no authentication.
Jev Repos & SDKs#community#github#python
Community
krishna-gogineni-765/deck31bPython
krishna-gogineni-765/deck31bPythonAn open-source, non-autoregressive System One decision engine inspired by TypeSafe AI: it evaluates state context against typed questions in a single forward pass, returning probability distributions and calibrated confidence without generating text.
Jev Repos & SDKs#community#github#python
Community
Atulmishra22/jev-litePython
Atulmishra22/jev-litePythonOpen infrastructure from the SimpleJev project for training, evaluating and deploying System 1 decision models across language and multimodal backbones: one API takes a state, question and candidate options and returns a choice with probabilities.
Jev Repos & SDKs#community#github#python
Community
SimpleJev/JevAnyPython
SimpleJev/JevAnyPythonFrom 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.