Dashboard Quality

Monitor workspace health, research quality scores, and agent performance from the dashboard.

Overview

The Praxiom dashboard includes a Quality Overview panel that surfaces key metrics about your workspace health and agent performance. These metrics help you understand how well the agent is performing, where data quality issues exist, and how research quality trends over time.


Quality Overview

The Quality Overview panel displays three metric cards and a per-workflow breakdown table. Data is fetched from GET /api/analytics/workspaces/{id}/harness.

Contract Pass Rate

The percentage of agent runs that passed post-flight contract evaluation. Contracts are algorithmic checks that verify the agent met minimum output requirements for each workflow type (e.g., synthesis must produce at least 3 insights with citations).

ColourRangeMeaning
Green80%+Healthy — agent is reliably meeting quality contracts
Amber50–80%Warning — some runs are falling short
RedBelow 50%Issues — many runs are failing quality checks

The card also shows the verifier average score (0–100) from the independent Haiku verification agent.

Complexity Distribution

A donut chart showing how your queries break down by complexity:

SegmentColourDescription
SimpleEmeraldSingle-step, direct responses
MediumAmberMulti-turn sessions with tool use
ComplexIndigoMulti-agent missions with DAG execution

This helps you understand whether your usage patterns match your plan — complex queries consume more credits per run.

Harness Overhead and Cost

  • Harness overhead — Average milliseconds added by post-flight quality checks (contract evaluation + verifier). Typically 200–500ms.
  • Average cost per run — Mean USD cost across all agent runs in the period.

Per-Workflow Breakdown

Below the metric cards, an expandable table breaks down quality by workflow type:

ColumnDescription
Workflowsynthesis, recommendation, drafting, chat, pipeline
RunsTotal runs for this workflow
Contract %Pass rate for this workflow
Verifier ScoreAverage independent verifier score
RQSResearch Quality Score for this workflow

Workspace Health Card

The WorkspaceHealthCard shows the overall health status of your workspace data, powered by entropy scanning.

Health Status

  • Healthy (green check) — No significant data quality issues detected
  • Issues Found (amber warning) — One or more issues detected, broken down by severity:
    • High severity (red) — Critical issues that actively degrade research quality
    • Medium severity (amber) — Issues that may affect accuracy
    • Low severity (muted) — Minor issues worth noting

Workspace Maturity

The card displays your workspace's maturity level based on total entropy scans completed:

LevelScansDescription
New0–10Just getting started — limited quality signal
Developing10–50Building up quality history
Mature50+Rich quality data for reliable trending

Quality Trend Card

The QualityTrendCard shows your Research Quality Score (RQS) over the last 30 days as a sparkline chart.

RQS Grades

GradeScoreColourMeaning
A80%+EmeraldExcellent research quality
B60–80%BlueGood quality with room to improve
C40–60%AmberModerate — check data quality issues
DBelow 40%RedPoor — likely data quality problems

Trend Detection

The card compares the first half and second half of the 30-day window. If the difference exceeds ±5%, a trend indicator shows:

  • Trending up — Quality is improving
  • Trending down — Quality is declining
  • Stable — No significant change

Quality Badges

Throughout the UI, quality scores appear as colour-coded badges:

BadgeWhere It AppearsWhat It Shows
RQSDashboard, conversation messages, workflow breakdownResearch Quality Score
RecQSRecommendation cards, conversation messagesRecommendation Quality Score
DocQSDocument cards, conversation messagesDocument Quality Score

Badges use the same A/B/C/D colour scale as the trend card. They appear inline in conversation messages when the agent creates insights, recommendations, or documents.


What's Next

Was this helpful?