All projectsProject
Fev
Our own system-1 decision model
A fast, schema-driven model that turns free text into a set of typed decision variables.
What it is
Fev is our own implementation of a system-1 decision model. Where a general-purpose model reasons step by step, Fev is built to answer fast and consistently against a schema you define.
Instead of writing a free-form prompt and hoping for the right shape, you declare the variables you care about and the type of each one. Fev fills them in.
Variables and types
A Fev task is a set of variables. Every variable is an open field with a name of your choosing and one of three types:
- Score — a normalised numeric judgement between 0 and 1, plus a short label.
- Noul — a presence check: the model decides whether the field applies at all, or leaves it empty.
- Choice — a single value selected from a list of options you provide.
Why it matters
System-1 models are for the thousands of small decisions that do not need a reasoning chain — routing a ticket, scoring a lead, classifying an intent, flagging a risk. Declaring the schema up front makes those decisions auditable, testable and cheap to run.
Define your variables below and run the mock model to see the structured output.
Try it
Fev tester
Fev — define variables, get typed decisionsMock mode — no backend required
Task / input
Variables name + type
One of your options
One of your options
Normalised 0–1 judgement
Normalised 0–1 judgement
Presence check, may be empty
Temperature 0 = deterministic, 1 = varied
Define your variables and run the model to see structured output.