Decision API · one forward pass

Ask a bounded question. Get a typed verdict.

Verdict Decision takes state and a question and returns a yes/no probability, a choice between options, or a score. It doesn't generate text, so you have nothing to parse and can branch on the answer straight away.

Get an API key Try the playground

pip install verdict · npm i @verdict/decide

request

{{ code }}

response● {{ latency }}

{{ r.label }}
{{ r.pct }}

{{ json }}

01 — why decide

Stop parsing prose to make a choice.

a generator

“Looking at the command, it appears to delete the build folder, which is generally safe, although it could be risky if the path is wrong, so I would probably suggest…”

→ regex it, retry it, hope it parses

a decision model

{ "answer": "yes", "p": 0.94 }

→ if (r.p > 0.9) run()

02 — the api

Three answer types. One endpoint.

Every call goes to POST /v1/decide. You set the type, and the response comes back in the matching schema every time.

Binary

Yes or no, with a calibrated probability. Use it for approvals, checks and anything you'd gate on.

{ answer: "yes" | "no"
  p: number }

Choice

Pick one of the options you list and get a probability for each. Use it for routing, labelling and ranking.

{ answer: Option
  distribution: Record<Option, number> }

Score

A level on a scale you define. Use it for risk, urgency, quality and priority.

{ score: number
  distribution: number[] }

03 — use cases

Every place your code asks “which one?”

Swap out a prompt-and-parse call for one typed decision. You get lower latency, a lower bill and one fewer thing to break.

{{ u.q }}

{{ u.type }}

04 — deploy

Run it next to your data, or call it from anywhere.

Local

Comes with Verdict and runs on your own hardware when it's powerful enough. Nothing leaves your machine and there's no charge per call.

verdict serve --decide --port 7070

Verdict Cloud

The same model and the same schema over one HTTPS call. It scales without your GPUs and gets updates as the model learns.

https://api.verdictdecision.com/v1/decide

observe

Every Verdict run records the task, the configuration it used and what happened.

label

Tests and Smart Harness checks mark each outcome as a pass or a fail.

train

The model is retrained on those real outcomes, not on anyone's opinion of what should happen.

05 — pricing

Pay per decision.

local

Included with every Verdict seat. Runs on your hardware.

Download Verdict

cloud● most used

$0.40/ 1k decisions

Metered API with no minimum. All three answer types are billed at the same rate.

Get API key

enterprise

Custom

Dedicated capacity, fine-tuning on your own outcome data, and on-prem deployment.

Talk to us

Your code has questions. Give it verdicts.

Get an API key Read the docs