> For the complete documentation index, see [llms.txt](https://cogna8-io.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cogna8-io.gitbook.io/docs/troubleshooting/performance.md).

# Performance

Cogna8 runs a multi-stage pipeline, so perceived slowness should be diagnosed by stage and workflow context rather than by total time alone.

***

## How to Assess

Start by observing Pipeline Status and identifying where time is being spent. Check whether the slowdown is one-off, tied to a specific scenario, tied to a particular browser, or consistently affecting all users.

***

## Common Contributors

* Large or ambiguous inputs that increase processing work
* Repeated conflict-heavy state updates
* Browser or device constraints
* Network variability
* Environment-specific configuration issues

***

## Practical Steps

{% stepper %}
{% step %}
**Reproduce with simpler input**

Isolate whether complexity is the factor.
{% endstep %}

{% step %}
**Compare browsers**

Rule out client-side issues.
{% endstep %}

{% step %}
**Use Demo Runner**

Test with controlled scenarios for comparison.
{% endstep %}

{% step %}
**Identify the slow stage**

Check whether one pipeline stage is consistently slow.
{% endstep %}
{% endstepper %}

{% hint style="warning" %}
**Optimize based on real bottlenecks.** Do not remove integrity checks for speed unless you have measured and accepted the risk tradeoff.
{% endhint %}