Julia-1 Decision Lab

Real Julia-1 ONNX

Finite-choice decision inference using the official SupersonicLabs/Julia-1-ONNX graph and external weights.

Julia-1official ONNX export

WebGPUbrowser inference

577 MBexternal weights

2–20options per decision

Test scenarios

Municipal service request

Ownership, urgency and human-review decisions.

Security access event

Response path, escalation and automation gate.

Procurement request

Owning function, processing path and review requirement.

Agent tool selection

Use Julia as a System-1 action/router decision layer.

Runtime path
State + question + choices → Julia encoder → Julia-1 ONNX → ONNX Runtime WebGPU → option logits → probabilities.

One observed state is evaluated against multiple explicit decisions. The model is loaded once and kept resident in browser memory.

SUPersonicLabs / JULIA-1-ONNX

Observed state / context

Ready for model load

Press “Load Julia-1”. The first load downloads the official ONNX graph, external weights and tokenizer assets.

Raw Julia output / diagnostics

No inference yet.