> ## Documentation Index
> Fetch the complete documentation index at: https://bifrost-dev.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Decisions

> Evaluate state against named questions using /v1/decisions.

Use decisions to get structured judgments - probabilities, choices, and rubric scores - from judgment models like TypeSafe's jev family. A decision request carries a `state` (what to evaluate) and a map of named `questions`; every question produces one typed answer.

## Provider Model Examples

* TypeSafe: `typesafe/jev-1.13.0`, `typesafe/jev-latest`, `typesafe/jev-preview`

## Basic Request

```bash theme={null}
curl --location 'http://localhost:8080/v1/decisions' \
--header 'Content-Type: application/json' \
--data '{
  "model": "typesafe/jev-1.13.0",
  "state": "Customer message: I was double charged and nobody replied to my two emails. I want a refund today.",
  "questions": {
    "is_frustrated": {
      "kind": "noul",
      "instructions": "Is the customer frustrated?"
    },
    "category": {
      "kind": "choice",
      "instructions": "Pick the ticket category",
      "criteria": { "billing": "charges and refunds", "bug": "product defects", "other": "anything else" }
    },
    "urgency": {
      "kind": "score",
      "instructions": "Rate how urgently this needs a human reply",
      "criteria": ["can wait a week", "should be answered soon", "needs a reply today"]
    }
  }
}'
```

## Question Kinds

| Kind     | Answer value                                           | Criteria                                                |
| -------- | ------------------------------------------------------ | ------------------------------------------------------- |
| `noul`   | Probability between 0 and 1                            | Optional `true`/`false` descriptions                    |
| `choice` | One option string                                      | Required map of option to description (max 255 options) |
| `score`  | Probability-weighted rubric score, including fractions | Required ordered array of 2-10 level descriptions       |

`state` and `instructions` accept a plain string or structured data (object or array), preserved losslessly.

## Response

```json theme={null}
{
  "model": "jev-1.13.0",
  "answers": {
    "is_frustrated": { "kind": "noul", "value": 0.98 },
    "category": {
      "kind": "choice",
      "value": "billing",
      "confidence": 1,
      "probabilities": { "billing": 1, "bug": 0, "other": 0 }
    },
    "urgency": {
      "kind": "score",
      "value": 2,
      "confidence": 1,
      "probabilities": { "0": 0, "1": 0, "2": 1 },
      "legend": { "0": "can wait a week", "1": "should be answered soon", "2": "needs a reply today" }
    }
  },
  "usage": { "prompt_tokens": 437, "completion_tokens": 72, "total_tokens": 509 }
}
```

Every declared question produces an answer of its declared kind. `confidence`, `probabilities`, and `legend` are included when the provider supplies them.

## Fallbacks

Standard fallback syntax applies:

```json theme={null}
{
  "model": "typesafe/jev-preview",
  "fallbacks": ["typesafe/jev-1.13.0"]
}
```

## Validation

Requests are validated before dispatch and rejected with a `400` rather than silently approximated: unsupported kinds, missing instructions, missing choice criteria, score criteria outside 2-10 levels or with non-string entries, and states or instructions that are neither string, object, nor array.

<Note>
  TypeSafe's native API is also available 1:1 at `POST /typesafe/v1/systemone` - see the [TypeSafe provider page](/providers/supported-providers/typesafe) for the mapping between the two.
</Note>