Decision 2.0 STUDIO Checking connection

INTRODUCING

Decision 2.0

Open Foundation Decision Models

One pass. Many decisions.

vLLM Semantic Router
Open the studio

Every scale.
One family.

Ask many questions about one input. Decision turns your options, yes-or-no questions and ordered scales into probabilities you can use.

Choice
Pick an option
Noul
A probability of yes
Score
A level on your scale
THE STUDIO

Your input. Your decisions.

Route a request, ask your own questions, or evaluate a batch of contexts.

Route each request to the right model.

Model
Try an example 1 prompt · 6 routing decisions

01User prompt

Fast pathQuick arithmetic

Shared by every question

Complete input per question. Nothing is truncated.

02Decisions

—
Ready when you are

Meet Decision 2.0.

Every model answers the same Choice, Noul and Score questions. Larger models read more nuance; smaller ones answer faster.

ModelSizeComplete inputReleasedStatusActions

Run Studio with your own model

Connect Studio to a verified Decision model. The model stays loaded between requests.

export DECISION_NATIVE=/path/to/native
export DECISION_MANIFEST_SHA256=<trusted manifest SHA>
export DECISION_RUNTIME_PATH=/path/to/release
uvicorn app:app --host 127.0.0.1 --port 7860

See the Studio README for the deployment contract. Credentials stay on the server. No model is downloaded automatically and no paid API is called.