Metrics and Variables is the administrator screen where you tune the values that Ender Turing's analytics run on. It is not a day-to-day screen: the settings here apply to the whole workspace and change how conversations are measured, how quality figures are displayed, and how generative AI features connect to model providers.

The screen is organised into three collapsible panels:

Each panel is saved on its own.

If you do not see the Metrics and Variables tab, your role does not meet the access conditions below — the page opens on the Settings tab instead.

Reaching this tab requires three conditions at once. All of them must be true:

If any condition is missing, the tab is hidden and the page falls back to the general Settings tab. There is no read-only mode: anyone who can open the tab can edit and save every field that is visible to them.

The EnderGPT panel appears only if you can manage system settings and either your workspace allows AI model selection or your account has extended platform access. In workspaces where Ender Turing manages model selection centrally, the whole panel stays hidden and models are maintained by Ender Turing.

Every panel has its own Save button at the bottom. Saving one panel does not save the others, so save each panel you edited before leaving the page.

Fields are validated as you type. If a required value is missing or out of range, the error appears under the field and Save does nothing until you correct it. A confirmation message appears once the save succeeds.

There is no recalculate or reprocess action on this screen.

The first panel holds the general analytics variables.

Two fields behave differently from plain inputs:

Client search aliases is also a JSON editor. It expects a mapping such as {"name1": ["phone1", "phone2"]}. Very large alias lists switch to a plain text box instead of the highlighted editor so that typing stays responsive.

This panel connects Ender Turing to your AI providers and decides how the generative AI features behave. It is hidden entirely in workspaces where Ender Turing manages these values for you.

Provider credentials

Keys are masked when the page loads: only the first and last few characters are shown, with asterisks in between. That masked value is safe to leave alone — save the panel without touching the field and the stored key is preserved. Use the eye icon to reveal what you are typing while entering a new key.

AutoQA re-scoring

Usage reporting

LLM cost and provider-token reporting are always enabled. There are no reporting checkboxes in this panel. Users who can manage both System and Automation settings can view USD, input, cached input, output, and total tokens in Settings > System > Usage. See System for how the metrics are used and how missing historical token data is displayed.

Privacy and logging

Categorisation limit

Models per task

Each AI task has its own model slot, so you can pair a fast model with a routine job and a stronger model with a job that needs it:

Each slot takes a provider and a model name. Depending on the model you pick, one extra control appears:

Changing the provider clears any reasoning or thinking value you had set, so re-check it after switching.

Prompt templates

These are the base templates behind the corresponding AI features. For everyday prompt work, use Gen AI Studio instead — it offers versioning, a usage panel, and a safer editing experience.

The last panel defines when a conversation should be treated as a repeat contact — a customer coming back about something that was not finished the first time. Ender Turing looks at the previous conversation with the same caller and links the two when it matches your rules.

The two required fields are enforced on save: a duration of 0 or more must be present, and at least one direction must remain selected.