Creating Your First Dashboard Workflow

Build a repeatable dashboard-based operating routine for model and platform health

This tutorial helps you establish a practical review workflow for platform operations using the Dashboard module.

What You’ll Build

A reusable operating checklist that:

1. Reviews demand and capacity indicators
2. Flags latency or throughput regressions
3. Validates model quality and token usage shifts
4. Produces a clear action list for owners

Step 1: Prepare a Baseline Window

1. Open Dashboard.
2. Set all core charts to LAST 7 DAYS.
3. Record current values for API calls, latency, throughput, and top model usage.

Step 2: Run a Demand and Capacity Pass

1. Review Total Requests trend and delta.
2. Compare running endpoints with recent load behavior.
3. Check inactive clusters to identify resilience risk.

Step 3: Run a Performance Pass

1. Inspect API Calls by project to find heavy consumers.
2. Inspect Latency for project-level outliers.
3. Inspect Throughput for efficiency regressions.

Step 4: Run a Model Quality and Cost Pass

1. Review Model Usage top models.
2. Review Accuracy benchmark spread across those models.
3. Review Token Metrics for input/output shifts.

Step 5: Produce an Action Register

Create a short table for your team:

Priority	Issue	Owner	Action	Due
P1	Example: Latency regression in Project A	Platform Eng	Inspect deployment scaling and logs	24h
P2	Example: Output tokens up for Model X	Applied AI	Tighten prompt and max token policy	48h

Step 6: Verify Outcomes

1. After changes, revisit Dashboard in the same window.
2. Confirm target metrics moved in the expected direction.
3. Capture before/after evidence for weekly review.

Operational Cadence Recommendation

• Daily (10–15 min): anomaly detection and assignment
• Weekly (30 min): trend review and capacity planning
• Monthly (45 min): quality/cost strategy adjustments