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

# Telemetry Export

> Push your organization's request metrics to Datadog, Grafana Cloud or any OpenTelemetry collector over OTLP/HTTP

Telemetry export sends your organization's request metrics from Requesty to your own observability stack. Requesty pushes metrics over OTLP/HTTP to Datadog, Grafana Cloud, your own OpenTelemetry collector or any other OTLP/HTTP endpoint, so you can chart Requesty traffic next to the rest of your systems and alert on it with the tools you already use.

> [!NOTE]
>
> **[Set up telemetry export](https://app.requesty.ai/admin-panel/integrations)** in the Requesty Console.

> [!NOTE]
>
> Telemetry export is available to organization admins on the Enterprise plan. Each organization has one export destination.

## Set Up a Destination

1. Go to the [Admin Panel](https://app.requesty.ai/admin-panel) and open the **Integrations** tab
2. Find the **Telemetry export** card and click **Set up telemetry export**
3. Choose **Datadog**, **Grafana Cloud** or **Other OTLP endpoint**
4. Fill in the destination details (see below) and click **Save destination**
5. Click **Check connection** to confirm that the destination accepts exports

Requesty pushes metrics every minute while **Export enabled** is on.

### Datadog

1. Pick your **Datadog site**, for example `US1 (datadoghq.com)` or `EU1 (datadoghq.eu)`
2. Paste a Datadog **API key** from **Organization Settings → API Keys** in Datadog

Requesty sends metrics to `https://otlp.<your site>/v1/metrics` as delta metrics, which is what Datadog's OTLP intake accepts.

### Grafana Cloud

1. Copy the **OTLP endpoint**, for example `https://otlp-gateway-prod-us-east-0.grafana.net/otlp`, from your stack's OpenTelemetry tile in the Grafana Cloud portal
2. Enter the stack's **Instance ID**
3. Paste an **API token** from an access policy that can write metrics

Requesty authenticates with Basic auth built from the instance ID and token, and sends cumulative metrics, which Grafana Cloud's Prometheus storage needs.

### Other OTLP Endpoint

| Field | Description |
| - | - |
| **Endpoint URL** | Base URL of your OTLP/HTTP receiver. It must use `https://` and cannot contain a query string. Requesty appends `/v1/metrics`, so `https://otlp.example.com` receives exports at `https://otlp.example.com/v1/metrics`. |
| **Headers** | Up to 10 headers sent with every export, for example `Authorization`. Header values are write-only: after you save, the console shows only the header names. |
| **Temporality** | **Delta** sends what was added since the last export. **Cumulative** sends running totals. Choose what your backend expects: Datadog needs delta, Prometheus-style backends need cumulative. |

## Starter Dashboards

The Datadog and Grafana Cloud set-ups include a starter dashboard. Click **Download the Datadog starter dashboard (JSON)** or **Download the Grafana Cloud starter dashboard (JSON)** in the set-up dialog or on the destination card, then import the file in Datadog or Grafana.

## Manage a Destination

From the **Telemetry export** card you can:

* Turn **Export enabled** off to pause exports without losing the configuration
* Click **Check connection** to send a test export. The card shows whether the destination accepted it, or the error it returned.
* Click **Edit** to change the URL, temporality or headers. Saving keeps the stored headers unless you click **Replace headers**.
* Click **Delete** to stop exporting and delete the stored URL and headers

## Metrics

Metrics follow the [OpenTelemetry GenAI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/gen-ai-metrics/) where a convention exists, and use `requesty.*` names where it does not.

| Metric | Type | Unit | Description |
| - | - | - | - |
| `gen_ai.client.token.usage` | Histogram | `{token}` | Tokens per request, split into input and output by `gen_ai.token.type` |
| `gen_ai.client.operation.duration` | Histogram | `s` | Total request duration |
| `gen_ai.server.time_to_first_token` | Histogram | `s` | Time to the first token, for streaming requests only |
| `gen_ai.usage.cost` | Counter | `USD` | Request cost, split by `requesty.cost.type` |
| `requesty.requests` | Counter | `{request}` | Number of requests |
| `requesty.router.duration` | Histogram | `s` | Time spent in Requesty, excluding the provider |

## Attributes

Every metric carries these attributes:

| Attribute | Example | Description |
| - | - | - |
| `gen_ai.response.model` | `gpt-4o` | Model that served the request |
| `gen_ai.provider.name` | `openai`, `aws.bedrock` | Provider that served the request |
| `gen_ai.operation.name` | `chat`, `embeddings` | Type of request |
| `requesty.status_code` | `200` | HTTP status code returned |
| `requesty.stream` | `true` | Whether the request was streamed |
| `requesty.api_key.id` | `3f1c2a6e-...` | ID of the API key that made the request |
| `error.type` | `429` | Status code of a failed request. Only present when the status code is 400 or higher. |

Some metrics add one more attribute:

| Attribute | Metric | Values |
| - | - | - |
| `gen_ai.token.type` | `gen_ai.client.token.usage` | `input`, `output` |
| `requesty.cost.type` | `gen_ai.usage.cost` | `router`, `provider`, `plugin`, `server_side_tool` |

Provider names use the GenAI convention values where one exists, for example `aws.bedrock`, `gcp.vertex_ai`, `gcp.gemini`, `azure.ai`, `mistral_ai` and `x_ai`. Other providers use their Requesty provider name.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.