blink
Open one-pass typed-decision model with a dependency-free C99 runtime that also builds to a 66 KB WebAssembly module.
Overview
Blink, from SQLite AI, puts the System One interface behind a C library you can link into a service, daemon or device, or load as WebAssembly in a web page. It answers questions of the form "given this state and this criterion, which of these options?" in a single forward pass, with no generation, decoding loop or JSON, returning one probability per option declared at call time.
The runtime is C99 with no dependencies beyond libc and libm. Weights are mapped read-only and never copied, and scoring performs zero allocations, which the test suite checks mechanically. The same sources build to a 66 KB WebAssembly module that runs in browsers and Node without a file system or server. The API encodes state and the option menu once and then scores, which suits repeated decisions over the same inputs.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Given this state and criterion, which of these options? (e.g. which queue for this ticket) | options declared at call time |
How it works
- Open a .blink model file
- Initialize a session in a caller-provided arena
- Encode state and option menu once
- Score to get one probability per option
Strengths
- No dependencies; embeddable anywhere
- Zero-allocation scoring, tested
- Runs in browsers via a 66 KB Wasm module
Caveats
- Model quality versus hosted Jev not stated in the excerpt
Getting started
Include blink.h, open blink-tiny.blink with blink_model_open_file and score options in a session.At a glance
- Type
- Service
- Status
- beta
- Language
- C
- Runtime
- embedded library / WebAssembly
- Has code
- yes
- Has tests
- yes
- Stars
- 22
Use cases and integrations
How curated lists describe it
An open-source, high-performance System One Model for one-pass typed decisions, with an embeddable C runtime and WebAssembly support.
wh000wh000/awesome-jev-live
Frequently asked questions
Does it need a server?
No, it links into your program or runs as Wasm in a browser.
Does it use the Jev API?
It offers the same kind of interface as an embeddable model, not the hosted API.
How big is the Wasm build?
About 66 KB.
Related projects
Mangaba AI's local, calibrated typed-decision model for Portuguese text, trained to optimize the Brier score; proprietary since v5.
ProjectproductionPython- Foq★ 11
Local open alternative to Jev using a 1.58-bit ternary 8B model to answer booleans, choices, scores and Pydantic extraction in ~25 ms.
ProjectbetaPython Runs the open Laya decision model entirely in a browser tab with WebGPU and ONNX Runtime Web: no server, no API key.
ServicebetaTypeScriptllama.cpp-based C++ server that turns any LLM into a System One decision model, reading answers from next-token logits.
ProjectbetaC++Three open 4B decision models for English, Turkish and German that serve the Jev /v1/systemone API on a CPU in about 3 GB of RAM.
ServicebetaPythonJapanese System One decision model on ModernBERT-ja (314.6M) returning typed choice, score and bool distributions with no generation.
ProjectbetaPython