# `TokenUsage`

The tokens one call to a model used: what it read, what it wrote, and how many of those it read its provider's cache served, where it says.

[AI](types.md#ai); not ordered; since Jadpo 0.1.0.

## Make one

Build one from its fields, `TokenUsage { input: ...  output: ... }`.

Its parts keep these rules; a value that breaks one fails with the rule's name:

- `cached_within_input`: `cached <= input, where it is given`

## Fields

| Field | Type |
|---|---|
| `input` | [`TokenCount`](type-token-count.md) |
| `output` | [`TokenCount`](type-token-count.md) |
| `cached` | [`TokenCount`](type-token-count.md)`?` |

## Text, JSON and storage

It has no single text form, so `text.from` and `text.parse` don't take it. JSON sends an object of its parts.

| JSON |
|---|
| `{"input":1200,"output":350,"cached":null}` |
| `{"input":1200,"output":350,"cached":800}` |
| `{"input":0,"output":0,"cached":0}` |

It is stored in 3 columns, each named for the field, `_` and its suffix. This is `{"input":1200,"output":350,"cached":800}`:

| Column | Type | Holds |
|---|---|---|
| `<field>_input` | `bigint` | `1200` |
| `<field>_output` | `bigint` | `350` |
| `<field>_cached` | `bigint` | `800` |

## Shown to people

`text.format(value, locale:)` shows it to people in their locale.

With `locale: Locale.en_gb`:

| Value | Shown |
|---|---|
| `{"input":1200,"output":350,"cached":null}` | `input 1,200, output 350` |
| `{"input":1200,"output":350,"cached":800}` | `input 1,200, output 350, cached 800` |
| `{"input":0,"output":0,"cached":0}` | `input 0, output 0, cached 0` |

## Operations

It has the contract's own operations: `text.format`, `json.read`, `json.write` and `==`. It isn't ordered, so `<` and sorting don't take it. [Every standard type meets one contract](../topics/every-standard-type-meets-one-contract.md) says how they work.