system-one (ZulfiFazhar)
Self-hosted Jev-compatible /v1/systemone server using an encoder with decision heads for parallel noul, choice and score answers.
Overview
Zulvis System One is a self-hosted server that implements TypeSafe's POST /v1/systemone contract, so existing Jev clients can switch to it by changing base_url. It uses a bidirectional encoder with direct decision heads to answer multiple noul, choice and score questions over one input state in a single forward pass; the README cites about 33 ms. The list summary describes it as a System One API for Laya.
The server is aimed at teams that want Jev-style decisions on their own infrastructure, including air-gapped setups: weights download once on first boot and then the service runs in strict offline mode. It includes a web UI and live inspector, in-memory token-bucket rate limiting, constant-time API key checks, a non-root Docker container and a Pytest suite. Claims such as "0% hallucination" in the README reflect that no text is generated, not a measured accuracy figure.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this proposition true of the state? | — |
| Choice | Which of these discrete options applies? | — |
| Score | Where does the state fall on this ordered rubric? | — |
How it works
- Start the FastAPI service or Docker container; weights download on first boot.
- Point a Jev client at the server's base_url.
- The encoder scores all questions in one pass.
- Return calibrated probabilities, confidence and expected scores.
Strengths
- Drop-in Jev API compatibility.
- Offline operation after first download.
- Rate limiting and API-key guardrails included.
Caveats
- Accuracy versus Jev is not reported in the README excerpt.
- Marketing-style claims should be checked against your own data.
Getting started
uv sync && uv run fastapi run --port 8000, or docker compose up -d.At a glance
- Type
- Service
- Status
- beta
- Language
- Python
- Runtime
- FastAPI server / Docker
- Has code
- yes
- Has tests
- yes
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
System One Model API for Laya compatible with TypeSafe Jev.
wh000wh000/awesome-jev-liveSystem One Model API for Laya compatible with TypeSafe Jev.
hellogumbo/awesome-jev
Frequently asked questions
Can I use my existing Jev client?
Yes, it implements POST /v1/systemone so you change only base_url.
Does it need internet access?
Only on first boot to download weights; after that it runs offline.
Does it need a TypeSafe key?
No, it serves its own model; an optional LAYA_API_KEY bypasses rate limiting.
Related projects
Docker service exposing local Laya, Von, a CLM head or hosted Jev through one TypeSafe-compatible HTTP and MCP endpoint.
MCP serverbetaPythonConverts open Jev reproductions (Laya, Kev, NanoJev, PlayJev) to INT8 ONNX for offline CPU decisions at about 15.6 ms per question.
ServicebetaPythonNative Go runtime serving Gemma 4 12B as a Jev-like decision model with hand-tuned PTX kernels and no CGo or CUDA toolkit.
ProjectbetaGoGo harness for System One models: hosted Jev, open Kev in-process via llama.cpp, and Cloudflare Clef, with typed helpers and a CLI.
ServicebetaGoGo gateway exposing the Jev /v1/systemone contract over ordinary LLM backends by reading single-token label logprobs.
ServicebetaGoRuns Intern-Decision-0.8B on Apple Silicon behind a local /v1/systemone endpoint that accepts screenshots for agent decisions.
ServicebetaPython