Monitoring PaaS usage

8.4.5 Usage Reports

The Usage Reports section of the admin panel gives you detailed monitoring of your PaaS platform's resource consumption. This feature lets you analyze usage, optimize performance, and effectively manage operational costs.

Section overview

Usage Reports give you a complete picture of platform consumption through:

• Real-time monitoring of usage by user and Agent;

• Detailed analysis of API and AI model usage;

• Period management for time-based analysis;

• Data export for advanced external analysis.

Report structure

The section is organized into two main tabs:

• Usage by Agent: analysis of consumption per individual Agent and user;

• Usage by key: monitoring of API key consumption.

Navigation controls

Reference period

You can customize the time period shown in the reports:

• Time navigation: use the arrows to move between previous and next periods;

• Current month: click to quickly view this month's data;

• Custom period: select specific intervals (e.g. 01/07/2025 - 01/08/2025).

Additional features

• Export CSV: download the data in CSV format for external analysis;

• Data refresh: reports update automatically with the latest usage data.

Usage by Agent tab

Table columns

The usage-by-Agent table shows the following metrics:

• Users > Agent: the owner of the Agent;

• Prompt Tokens: number of tokens used in requests;

• Completion Tokens: number of tokens generated in responses;

• Total Tokens: total sum of tokens used;

• Valid chats: number of completed conversations;

• Chats with Deep Thinking: sessions that used advanced memory;

• AI responses: number of responses generated by the AI;

• AI responses (Uncounted): responses generated with custom models;

• AI responses with DT: responses with Deep Thinking;

• AI responses with DT (Uncounted): DT responses with custom models;

• Imported documents: number of documents processed;

• Imported documents (Uncounted): documents processed with custom models;

• Average tokens per message: average token consumption per interaction.

Example data

example.user

• Prompt Tokens: 14,516;

• Completion Tokens: 1,780;

• Total Tokens: 16,296;

• Valid chats: 6;

• Chats with Deep Thinking: 2;

• AI responses: 5;

• AI responses (Uncounted): 2;

• AI responses with DT: 2;

• Imported documents: 0 p;

• Average tokens per message: 1,810.67.

dev.user

• Prompt Tokens: 2,199,180;

• Completion Tokens: 30,361;

• Total Tokens: 2,229,541;

• Valid chats: 20;

• Chats with Deep Thinking: 1;

• AI responses: 35;

• AI responses (Uncounted): 0;

• AI responses with DT: 1.

Usage by key tab

Table columns

The usage-by-API-key table shows:

• Key: masked identifier of the API key used;

• Prompt Tokens: tokens used in requests;

• Completion Tokens: tokens generated in responses;

• Total Tokens: total token consumption for the key.

Example data

Provider-ModelA-Config

• Configuration: Provider-Q&A-Processing

• Key: 7640860******o5I

• Model: chat-model-version

• Prompt Tokens: 128,530;

• Completion Tokens: 6,148;

• Total Tokens: 134,678.

Provider-ModelB-Config

• Configuration: Provider-ImportExport-Processing

• Key: 7640860******o5I

• Model: processing-model-version

• Prompt Tokens: [data available in the report].

Monitored metrics

Key indicators

1. Administrator accounts (tenant only)

• Number of active admins;

• Maximum allowed limit.

2. Conversational Agents

• Number of active Agents per user and tenant;

• Maximum allowed limit;

• "Show Agent list" button to view the full list.

3. Valid chats

• Number of completed conversations per user and tenant;

• Applicable limit when configured.

4. AI-generated responses

• Total number of responses generated per user and tenant;

• Maximum allowed limit;

• Uncounted AI responses: responses generated with custom models.

5. Imported documents

• Number of pages imported per user and tenant;

• Maximum allowed limit;

• Uncounted pages: pages processed with custom models.

Using custom models

Uncounted consumption

When you use custom language models with your own API keys:

• Generated responses: not counted against the standard limit;

• Imported pages: not counted against the standard limit;

• Separate display: this consumption appears as "uncounted".

💡 Important: Agent and valid chat limits per user and tenant are not affected by using your own API keys.

Interpreting the data

Usage analysis

For each metric, the following is shown:

• Consumption value: actual usage relative to platform limits;

• Maximum limit: the threshold configured for each item (when present);

• Uncounted consumption: usage with custom models.

Performance indicators

• Average tokens per message: conversation efficiency;

• Prompt/Completion ratio: balance between requests and responses;

• Deep Thinking usage: adoption of advanced features.

Optimizing consumption

Efficiency strategies

• Regularly monitor usage patterns to identify anomalies;

• Analyze average tokens per message to optimize efficiency;

• Balance usage between standard and custom models;

• Export data periodically for advanced analysis.

Cost management

• Use custom models to reduce counted consumption;

• Optimize conversations to reduce token usage;

• Monitor usage of premium features;

• Plan growth based on consumption trends.

Best practices

• Regularly export data to CSV for detailed analysis;

• Compare consumption across different periods to identify trends;

• Identify users with unusual consumption for targeted optimization;

• Document usage patterns for future planning;

• Configure custom alerts for critical consumption thresholds.

The Usage Reports section gives you the tools you need for efficient, informed management of your PaaS platform's resources — letting you optimize costs and performance based on accurate, up-to-date data.