Monitoring

The entry point for monitoring is as follows:

Statistics

Agent Monitoring

The Data Center provides comprehensive statistics and analysis on the usage of intelligent assistants and user conversation content, helping users better understand the effectiveness and operational status of AI, and optimize management and decision-making.

The Data Center supports data statistics by assistant dimension, system API dimension, and time dimension.

Assistant Analysis

This section focuses on quantitative analysis of the usage of each AI assistant, including:

• Number of Users: The number of unique users who have interacted with the assistant.

• Conversation Count: The total number of interactions between users and the assistant, measuring usage activity.

• Token Consumption: Measures the resource consumption of model calls during assistant operation.

• Assistant Usage Ranking: Ranks all assistants by usage frequency to identify high-frequency assistants.

• Token Consumption Ranking: Statistics on Token usage by each assistant, assisting in resource optimization and cost control.

Conversation Analytics

This section delves into the content of conversations between users and assistants, gaining insights into user needs and interaction patterns, including:

• High-frequency words: Extracts keywords from user conversations to identify hot topics and demand trends.

• Conversation Logs: View historical conversation details for quality review, issue troubleshooting, and content auditing.

💡 Tip: Conversation count statistics and assistant usage rankings are updated once per day; other information such as conversation logs is updated in real time.

Notification Management

Notification management is used to count and view various notification messages received by all users in the system. Through this feature, administrators can track the delivery status of notifications, understand whether users have read them, and whether the notification content has been effectively communicated.

For example: When a user provides feedback on an assistant's response during a conversation, the system will send a feedback notification to that user. Such notifications will be recorded and displayed in notification management.

Feature Description

On the notification management page, you can view the following information:

• Recipient: The account or username of the user who received the notification.

• Status: Whether the notification has been read (read / unread).

• Send time: The specific time the system sent the notification.

• Detailed Content: Detailed information of the notification, supporting viewing of original prompts, feedback results, etc.

• Category: The category to which the notification belongs, such as feedback notification, system notification, task reminder, etc.

💡 Tip: Notifications can be filtered by time dimension.

Billing Details

In the AI Central system, you can view billing details for Token usage. The line chart at the top of the interface displays the changes in Token-related values at different time points, providing an intuitive view of usage trends. The table below records in detail the Token consumption and corresponding costs of various models within a specific time period, and lists related usage types and other information, making it easy to clearly understand the billing situation for each Token usage.

Information Provided on the Billing Details Page

On the billing details page, you can view the following information:

• Model Group: Displays the group category to which the model belongs.

• Model Set: Indicates the name of the set to which the model belongs.

• Model: Shows the name of the specific model used.

• Channel: Identifies the channel through which the model operates.

• API Key Name: Displays the name of the API key used to access the model.

• Input Tokens: Records the number of tokens input into the model.

• Input Tokens: Shows the number of tokens input when the cache is hit.

• Cached Input Tokens: The number of input tokens that were served from a cache, reducing cost and latency.

• Output Tokens: Displays the number of tokens consumed by the model's output results.

• Token cost: Lists the cost incurred by using tokens.

• Time: Records the specific time when the model operation occurred.

• Action: Provides executable options such as viewing log details.

Purpose of Billing Details

• Cost transparency: Accurately track the cost incurred by each model call, achieving cost visibility, management, and control.

• Usage trend analysis: Identify resource consumption patterns through trend charts, providing data support for resource procurement and budget planning.

• Issue troubleshooting and optimization: Analyze the reasons for high-cost requests using detailed logs, and optimize prompts or business processes to control costs.

• Multi-dimensional allocation: Allocate costs to different projects or departments through dimensions such as model, channel, API Key, etc.