Analysis Dashboard¶

In Agent Monitoring > Analysis Dashboard, you can view multiple metrics for selected applications in chart form, gaining a comprehensive understanding of application performance and cost distribution. The top of the page supports filtering by application.

Note

When there is no data within the selected time range, the chart area will display a "No data" status. You can adjust the viewing range using the time widget in the upper right corner.

LLM Monitoring Application¶

After selecting an LLM monitoring application, the Analysis Dashboard displays the following content.

Core Metrics Overview¶

The top of the dashboard shows key metric statistics for the selected time range:

Request Count: Total number of requests.
Span Count: Total number of associated Span calls.
Request Error Rate: Percentage of failed requests out of total requests.
Total Token Consumption: Total number of tokens consumed by all requests.
Average Response Time: Average processing time for all requests.

Trend Analysis Charts¶

Through multi-dimensional trend charts, you can track changes in application performance and usage trends:

Request Trend Chart: View the distribution of request volume across different time periods to identify usage peaks.
Request Error Rate: Shows the trend of error rate changes, helping to quickly detect abnormal fluctuations.
Request Response Time: Shows Avg, P75, P90, P99 percentile latencies to locate long-tail latency issues.
Token Consumption Trend Chart: Shows the time distribution of token usage, aiding in cost forecasting.
Average Time to First Token (TTFT): Measures first-token latency, reflecting model response speed.

Model and User Analysis¶

Analyze resource usage and performance from the model and user dimensions:

Request Proportion by Model: A donut chart shows the call distribution of different models (e.g., qwen3:1.7b).
Token Usage Ranking by Model: Compares resource consumption across different models.
Request Response Analysis by Model (Top 20): A table shows detailed performance metrics for each model, such as request count, average/maximum latency, percentile latencies, etc.
User Request Ranking (Top 10): Identifies high-frequency users.
User Token Consumption Ranking (Top 10): Locates high-cost users, assisting in cost allocation.

Agent Monitoring Application¶

After selecting an Agent monitoring application, the Analysis Dashboard displays content in four modules: Request, Model, Skill, and Tool.

Request Module¶

Average Request Duration: Latency fluctuations for different Agents over a time range.
Total Request Trend: Trend of total requests for different Agents over a time range. Clicking on a point or selecting a time range allows you to jump to the corresponding Agent details with the time context.

Model Module¶

Model Usage Ranking: View the call popularity of each model.
Average Request Duration by Model: Compare the response latency trends of different models to quickly identify performance differences.
Model Token Consumption Ranking: View the token consumption ranking of models, helping to clarify costs.

Skill Module¶

Skill Usage Ranking: View the usage popularity of Skills, facilitating optimization of call strategies.
Average Request Duration by Skill: Compare the execution fluctuations of different Skills.

Tool Module¶

Tool Call Ranking: View the ranking of call counts for each Tool.
Average Tool Call Duration: Statistics on the average execution time for each Tool.
Maximum Tool Call Duration: Statistics on the maximum execution time for each Tool.
Minimum Tool Call Duration: Statistics on the minimum execution time for each Tool.