Dashboard

Overview

The Dashboard provides a comprehensive visual overview of platform activity, request metrics, token usage, cost analysis, and performance indicators. This centralized monitoring interface enables administrators and users to track key performance indicators, identify usage patterns, and make data-driven decisions.

Accessing the Dashboard

1. Navigate to the left sidebar navigation panel
2. Click on "Monitoring" in the menu
3. Select "Dashboard" from the submenu options (indicated by the arrow icon)
4. The Dashboard interface will load with all visualizations and metrics

Dashboard Interface Components

Time Range Selection

Located directly below the "Dashboard" title, this control allows filtering of all dashboard data by different time periods:

• 24H: Displays data from the past 24 hours (currently selected in the interface)
• 7D: Displays data from the past 7 days
• 1M: Displays data from the past month
• 3M: Displays data from the past 3 months
• All: Displays all historical data since platform inception

Requests Overview Panel

Positioned in the upper left section of the dashboard, this panel contains:

• Section Title: "Requests"
• Request Count: Displays the total number of requests (currently showing "5")
• Status Filter: Dropdown filter showing "Success" status with an option to clear (indicated by "×")
• Request Trend Graph: Shows request volume over time as a line chart
  • Green line with shaded area represents successful requests
  • X-axis shows time stamps (09:45 AM, 09:46 AM, 09:50 AM)
  • Green dot with "Success" label indicates successful request status
  • The graph shows the trend of requests, with a dip at 09:46 AM and increased activity at 09:50 AM

Token and Cost Metrics

Located in the upper right section, displayed as four separate panels with bar chart icons:

Average Cost per Request

• Title: "Average Cost per Request"
• Current Value: $0.0158
• Visualization: Bar chart icon indicating cost metrics

Avg Prompt Tokens per Request

• Title: "Avg Prompt Tokens per Request"
• Current Value: 1503
• Visualization: Bar chart icon indicating token volume

Avg Completion Tokens per Request

• Title: "Avg Completion Tokens per Request"
• Current Value: 24.4
• Visualization: Bar chart icon indicating token volume

Avg Total Tokens per Request

• Title: "Avg Total Tokens per Request"
• Current Value: 1527.4
• Visualization: Bar chart icon indicating total token usage

Model Usage Analysis

Located in the lower left section of the dashboard:

• Section Title: "Top Models"
• Table Format:

Name	Requests
Openai/gpt-4-0125-preview	3
Openai/gpt-4-turbo	2

• The table shows the distribution of requests across different models
• Model names are displayed with color-coded backgrounds (purple for preview model, blue for turbo model)
• Request counts indicate the number of times each model was called during the selected time period

Cost Analysis

Located in the lower middle-left section:

• Section Title: "Costs"
• Total Cost: $0.08
• Visualization: Bar chart showing cost distribution over time
  • X-axis: Time intervals from 09:45 AM to 09:50 AM
  • Y-axis: Cost values (not explicitly labeled but represented by bar height)
  • Blue bars with varying heights indicating cost fluctuation
  • Blue dot with "costs" label in the legend
  • Highest costs appear at 09:46 AM and 09:50 AM

User Activity

Located in the lower middle-right section:

• Section Title: "User Requests"
• Active Users: 1
• Visualization: Bar chart showing user activity over time
  • X-axis: Time intervals from 09:45 AM to 09:50 AM
  • Y-axis: User count (not explicitly labeled but represented by bar height)
  • Orange bars with consistent height indicating steady user activity
  • Orange dot with "users" label in the legend
  • Consistent height across all time periods indicates a single user has been active throughout

Latency Metrics

Located in the lower right section:

• Section Title: "Avg Latency"
• Current Value: 3 ms / req
• Visualization: Line chart showing latency trends over time
  • X-axis: Time intervals from 09:45 AM to 09:50 AM
  • Y-axis: Latency values in milliseconds
  • Teal line showing the pattern of response times
  • Teal dot with "latency" label in the legend
  • The line shows fluctuation in latency with a peak around the middle of the time period

Interpreting Dashboard Data

Request Metrics

• Total Requests: The count of 5 requests indicates low to moderate platform usage during the selected 24-hour period.
• Success Rate: The filter shows successful requests, suggesting all requests completed successfully.
• Request Pattern: The trend line shows varying activity levels throughout the time period, with the lowest activity around 09:46 AM and increased activity at the beginning and end of the monitored period.

Token Usage Analysis

• Prompt Tokens: The average of 1503 prompt tokens per request indicates relatively large input queries.
• Completion Tokens: The average of 24.4 completion tokens per request shows relatively concise responses.
• Token Ratio: The high ratio of prompt tokens (1503) to completion tokens (24.4) suggests large inputs with brief outputs, which might indicate efficient prompt engineering or query-focused tasks.
• Total Tokens: The average of 1527.4 total tokens per request represents the combined token consumption, useful for cost estimation and monitoring.

Model Distribution

• The GPT-4 preview model (openai/gpt-4-0125-preview) is used most frequently with 3 requests.
• The GPT-4 turbo model (openai/gpt-4-turbo) follows with 2 requests.
• This distribution indicates a preference for the newer preview model while still utilizing the turbo model for some requests.

Cost Analysis

• Average Cost: $0.0158 per request indicates the typical expenditure for each API call.
• Total Cost: $0.08 for the period shows overall platform expenditure during the selected timeframe.
• Cost Pattern: The bar chart shows higher costs at specific times (09:46 AM and 09:50 AM), which correlates with model usage.

Performance Metrics

• Latency: The average of 3 ms/req demonstrates excellent platform responsiveness.
• Latency Pattern: The line chart shows variation in response times throughout the period, with a peak in the middle that gradually decreases toward the end of the timeframe.

Using the Dashboard Effectively

Time Range Selection Best Practices

• Use 24H for daily operations and recent performance analysis
• Use 7D for identifying weekly patterns and trends
• Use 1M or 3M for long-term analysis and month-over-month comparisons
• Use All for historical data assessment and platform changes over time

Monitoring Tips

• Regular Check-ins: Review the dashboard at consistent intervals to establish baseline metrics and identify anomalies.
• Status Filtering: Use the status filter dropdown to isolate successful requests from failures when troubleshooting.
• Cost Efficiency: Compare the average cost per request against token usage to optimize prompt engineering.
• Model Selection: Review the Top Models table to ensure appropriate model selection for different tasks.
• Performance Tracking: Monitor the latency graph for unexpected spikes that may indicate system issues.

Data Analysis Strategies

• Token Optimization: Analyze the ratio between prompt and completion tokens to identify opportunities for input optimization.
• Cost Management: Track cost patterns to identify high-usage periods and potential areas for cost reduction.
• Usage Patterns: Correlate request volume with time periods to understand peak usage times.
• Model Efficiency: Compare model usage with cost metrics to determine the most cost-effective models for specific tasks.

Troubleshooting Common Issues

Issue	Possible Cause	Solution
No data displayed	Time range too narrow or no activity	Expand time range or check for system issues
Missing metrics	Recent system changes or data processing delay	Refresh dashboard or wait for processing completion
Unexpected cost spikes	High-volume usage or premium model selection	Review model usage table and optimize prompt design
Latency increases	System load, network issues, or model performance	Check system logs and infrastructure monitoring
Request count discrepancies	Filtering issues or data processing delays	Clear filters and refresh the dashboard