The Usage page provides a comprehensive overview of the Token consumption and dialogue usage on the OAgent platform, serving as a data foundation for cost assessment, budget formulation, and consumption trend analysis. Click Usage in the left navigation bar to access this page.
When to use
Understand the overall Token consumption scale and trend of the platform to assess the cost level of model usage.
Compare the composition ratios of input and output Tokens to understand the consumption structure, serving as a reference for model selection and budget formulation.
Identify peak Token consumption periods and pay attention to abnormal fluctuations in usage.
Procedure
Click Usage in the left navigation bar to go to the Usage page.
View the top four cards: Total Tokens, Input Tokens, Output Tokens, and Dialog Count, to grasp the consumption and usage scale within the selected time range. (By default, data from the last 7 days is displayed. You can adjust the time range at the top of the page.)
CardDescriptionTotal Tokens The total number of Tokens consumed within the selected time range, including both input and output Tokens. Input Tokens The number of Tokens sent to the model within the selected time range. Dialogue context and tool return results are sent back to the model with each request, so input Tokens are usually significantly higher than output Tokens. Output Tokens The number of Tokens generated by the model within the selected time range. Dialog Count The number of sessions returned within the selected time range (up to 1000). Observe the Token Consumption Trend line chart, which shows how Token consumption changes over time. This helps you observe the overall trend, compare consumption levels across different periods, and identify consumption peaks.
What to do next
To view the Token consumption and models used in a single dialogue, see View the tool invocation process.
To reduce Token consumption by starting or stopping tools to free up irrelevant context, see Tools.
For model evaluation, refer to the testing feature on the model configuration page. See Manage model services.