> ## Documentation Index
> Fetch the complete documentation index at: https://docs.slng.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Usage and billing

> Track spend, audio minutes, and requests across your account.

The **Billing & Usage** area in the dashboard reports what your account consumed:
how much you spent, how many audio minutes you processed, and how many requests
you made. Use it to see where spend concentrates, compare regions and models, and
export the raw numbers for your own reporting.

This page reports per-request model usage. To replay an agent call and see where
a session went wrong, use [Debug a call](/guides/agents/debug-a-call).

<Frame caption="The Usage dashboard, with the four scope tabs, headline metrics, and region breakdown.">
  <img src="https://mintcdn.com/slng-new-docs/q7VaV02pKEe0UuAm/heroshots/usage-overview.png?fit=max&auto=format&n=q7VaV02pKEe0UuAm&q=85&s=e237410e24a6e95284b55feed2f8f434" alt="The Billing and Usage dashboard showing total spend, audio minutes, and requests" width="2560" height="1600" data-path="heroshots/usage-overview.png" />
</Frame>

## Choose a scope

Four tabs across the top split usage by where it came from. Pick the one that
matches the question you are asking.

| Tab                 | Counts                                                                            |
| ------------------- | --------------------------------------------------------------------------------- |
| **Overview**        | Everything combined across Direct API, Managed Agents, and the Execution Layer    |
| **Direct API**      | Direct model and Batch API calls; excludes Managed Agents                         |
| **Managed Agents**  | Agents you deployed on SLNG                                                       |
| **Execution Layer** | Listen, Think, and Speak optimizations such as the context router and TTS caching |

## Read the headline metrics

Every scope leads with three numbers for the selected period:

* **Total spend** in your billing currency.
* **Total audio minutes** processed.
* **Total requests** made.

On the **Overview** tab each number breaks down by source, so you can see how
much of your spend came from direct model calls versus Managed Agents, the
Execution Layer, or the Batch API.

<Tip>
  Spend and audio minutes track separately. A cache hit from the Execution Layer
  can add audio minutes without adding model spend, so the two numbers move
  independently.
</Tip>

## Filter and slice the view

Narrow the report before you read the breakdowns:

* **Regions**, **projects**, and **models** dropdowns scope the numbers to one
  slice. The **Direct API** tab filters by all three; the **Overview** tab
  filters by region.
* **Day**, **Week**, and **Month** set the granularity of the trend charts.
* The **date range** picker bounds the whole report.

## Read the breakdowns

### By region

The **Region breakdown** table lists each region with its audio minutes, spend,
share of total spend, and request count. Use it to confirm that traffic lands in
the regions you expect. See [Regions](/guides/regions/regions-map) for what each
region covers.

### By model

On the **Direct API** tab, the **Model breakdown** table adds a per-model row
with:

* **Provider** — whether the model ran SLNG-hosted or on your own key (BYOK).
* **Type** — speech to text (STT) or text to speech (TTS).
* **Protocol** — `HTTP` or `WSS`.
* **Audio minutes**, **spend**, **% of spend**, and **requests**.

A separate **Batch API** breakdown lists model, audio minutes, jobs count, and
spend. It stays empty until you run a batch job.

## Read the trend charts

Two charts sit below the breakdowns:

* **Request Trend** plots request volume over the selected period.
* **Spend Trend** plots spend over the same period.

Each uses the **Day / Week / Month** granularity you selected and captions the
change against the previous period. The dashed section marks the current,
incomplete period.

## Inspect individual requests

Below the charts, the **Request log** lists individual model requests, one row
per call, with the date, type (STT or TTS), model name, transport, status code,
audio minutes, and price. Filter by **Type**, set **Status** to **Errors only**
to isolate failures, or search by model name.

The request log covers direct model requests. To debug an agent call session
with its transcript and events, use [Debug a call](/guides/agents/debug-a-call).

## Export the data

**Export** downloads the current view for your own reporting or reconciliation.
**Refresh data** pulls the latest numbers without reloading the page.