<!-- Markdown mirror. Canonical: https://empryo.com/docs/concepts/model-events -->

# Model events

> An opt-in debug view that records each model call's speed, token count and errors, so you can find the slow or failing model when you mix providers.

When you use several models at once, it can be hard to tell which one is slow or which one is failing. **Model events** is an optional debug view in the terminal UI. It records every model call Empryo makes, including the main agent, helper agents, compaction, code summaries and web search.

It is off by default and costs nothing while off.

## Open it

```
/model-events
```

Press `e` to start recording. Press `e` again to stop, which also clears what was recorded. The view stays empty until you start recording.

## The three tabs

Use `Tab` or the arrow keys to switch.

**Models** shows totals per model: calls, errors, average and most recent response time, and total tokens. The most recently used model is at the top.

**Recent** lists every call, newest first: when it ran, where it came from (main, helper agent, compaction or other), the model, status, duration, and tokens in and out.

**Errors** lists only failed calls, with the error message, the model and where the call came from.

## When to turn it on

- Things feel slower than expected and you want to know which model is the bottleneck.
- Helper agents fail without an obvious reason and you want to see which model errored.
- You're comparing two providers on the same task and want response times side by side.
- Cache savings look lower than expected and you want to see cache-read tokens per model.

For everyday work, leave it off. Cost and token totals are always available in `/context`.

## What it stores

- The last 500 calls, kept in memory only.
- Nothing is written to disk or kept between sessions.
- Stopping recording or restarting Empryo clears it.
- Recording is a per-session switch, not a config setting.

## Related

- [`/context`](/docs/concepts/cost-tracking) shows running totals, cost per model and helper-agent spend.
- [`/router`](/docs/recipes/task-router) lets you pick which model handles which kind of task.
