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

# Observability

> Track every request, measure every token, optimize every dollar.

Measuring every token, every request, every compression event, every routing decision, at session level locally and at team level in the managed Edgee console.

## Read usage in the console

Start in **Activity** for organization trends, **Team** for member or squad costs, **Sessions** for one agent run, and **Logs** for one model request. Select the same organization, period, and filters before comparing values.

### Activity

Review costs, requests, token volume, and savings over the selected period. Displayed usage cost is not necessarily a deduction from Edgee credits, including when using a provider subscription or BYOK.

![Activity showing organization usage over time.](https://mintcdn.com/edgee/ZJH-mHY-14RdSp8t/images/console/activity.png?fit=max\&auto=format\&n=ZJH-mHY-14RdSp8t\&q=85\&s=06acc7010cd20fda9585b83ea1dab948)

Local demonstration with fictional members and illustrative usage data.

### Logs

Filter to the affected period, key, model, or provider. Open a request to compare the model requested with the model served, token counts, latency, cost, and errors. A reroute explains a model difference; a fallback can explain a provider change after an error.

![Request logs with model, provider, tokens, latency, and cost.](https://mintcdn.com/edgee/ZJH-mHY-14RdSp8t/images/console/logs.png?fit=max\&auto=format\&n=ZJH-mHY-14RdSp8t\&q=85\&s=76807a6fb429afbc86e0cbff19c45bc3)

Local demonstration with fictional members and illustrative usage data.

Full payloads are available only when [Debug](https://www.edgee.ai/docs/features/debug.md) captured them. Enabling Debug later does not reconstruct earlier requests.

### Investigate an unexpected bill

1. Find the period with the increase in Activity.
2. Identify the member or squad in Team.
3. Open their sessions and inspect the most expensive requests.
4. Check routing, cache usage, model/provider, and the billing mode.
5. Adjust the appropriate [strategy](https://www.edgee.ai/docs/features/routing-strategies.md) or [limit](https://www.edgee.ai/docs/features/usage-limits.md), then verify new requests.

To send usage records to your infrastructure, configure an [observability destination](https://www.edgee.ai/docs/features/integrations/data-destinations.md).

## Application usage

For SDK response metrics and request tagging in application code, see [Application observability](https://www.edgee.ai/docs/llm-router/observability.md).

[Request tags](https://www.edgee.ai/docs/llm-router/observability.md#request-tags-for-analytics), [token usage](https://www.edgee.ai/docs/llm-router/observability.md#token-usage-tracking), and [compression metrics](https://www.edgee.ai/docs/llm-router/observability.md#compression-metrics) are covered in the LLM Router & SDKs section.


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