Intelligence

The Intelligence dashboard — internally called "God Mode" — transforms raw test execution data into actionable quality insights. It tracks trends, detects flaky tests, and performs AI-powered root cause analysis.

Dashboard Overview

The Intelligence view is divided into three zones:

Zone	Purpose
Success Trend Chart	Quality trajectory over time across configurable run windows
Quality Matrix	Heatmap classifying every test as Stable, Flaky, or Critical
Flaky Analysis Lab	Deep-dive diagnostics with AI root cause detection and Auto-Fix suggestions

Success Trend Chart

The main chart plots your project's pass rate across the last 15, 50, or 100 runs. A sustained drop signals a "Quality Drift" — a quality regression that may correlate with a recent merge or infrastructure change.

Run Window: last 50 runs
Pass Rate:  94.2% (trending down from 97.1%)
Quality Drift detected: -2.9% since run #38

Window Selection

Use a smaller window (15 runs) for active development feedback. Switch to 100 runs for release readiness assessment.

Quality Matrix

The heatmap grid represents every test in the project, color-coded by stability:

• Green (Stable) — consistently passing across the selected window
• Yellow (Flaky) — alternates between pass and fail without code changes
• Red (Critical) — currently failing and requires immediate attention
• Lightning icon — infrastructure alert: failure occurred during high system load (CPU > 80%)

Executive QA Summary

Click Generate Executive Report to produce an AI-written summary of the current quality state. This brief is designed for stakeholders who need release confidence without reviewing technical logs.

Sub-Pages

Page	Content
Quality Metrics	Score calculation, trend windows, pass rate details
Flaky Detection	Quality Matrix heatmap, classification logic, infra alerts
Root Cause Analysis	DNA Diagnostics, root cause categories, Auto-Fix suggestions

Data Source

Intelligence reads from .xyva/run-history/, which the Runner populates after every execution. At least 5 runs are needed before trend data becomes meaningful.