---
title: "Quality gates"
manual: "AI Assistant"
version: "main"
source: "Integrators/QualityGates.rst"
rendered: "2026-10-03T14:54:45+00:00"
---

# Quality gates {#quality-gates}

A quality gate reviews the generated answer before it is returned. It is useful
for high-risk content, strict source requirements or complex pipelines.

## Validation criteria {#validation-criteria}

A quality gate can check whether:

-   the answer is based on retrieved context;
-   the answer follows the requested language and tone;
-   required source references are present;
-   the answer contains unsupported claims;
-   the answer admits missing context where needed.

Use quality gates carefully: they add latency and model cost.