The purpose of this module is to manage the costs of AI consumption for both. quarries and the embeddings
The analysis BL/ ingestor BL components will be the one updating both tables
Tables
Table : AI_USAGE_HISTORY
Name | Type | Comments |
|---|---|---|
RequestID | The unique identifier of the request | |
RequestType | Summary, Tasks, Query, Embedding, Topics | |
UserUPN | The user principal name of the requester. Can be used to get the account from BG USERS | |
RequestPromptTokenCount | The number of tokens in the request prompt | |
RequestCompletionTokenCount | The number of tokens in the request completion | |
RequestCostUSD | The cost of the request in US dollars. Calculated by cost per token in the Gateway_AI_pricing table in the gateway DB | |
RequestDate | The date of the request | |
RequestSourceType | Chat | |
RequestSourceId | ID of the RequestSourceType | |
Project | Any call to “AI Service Hub” with a Project name that is new will be added to this table |
Table: AI_USAGE_COST
Name | Type | Comments |
|---|---|---|
UserUPN | The user principal name of the user | |
CurrentMonthUsageUSD | The amount of usage in US dollars for the current month | |
MonthlyUsageLimitUSD | The limit of usage in US dollars for the current month | |
LastUsageUpdateTime | The last time the usage was updated | |
SubscriptionStartDate | When user started using the subscription |
For every new users when someone asks a question, the analysis BL adds a record to this table with the MonthlyUsageLimitUSD taken from the general setting table
Dashboard UI
Under the menu item of “User settings,” the user will be able to see a new page named: “Usage”
The page will show at the top:
Total monthly costs is X$ from the limit of Y$. Show bar.
Also add detailed list displaying all the records from AI_USAGE_HISTORY with option to export to Excel.
In the Dashbaord, on every activity - need to check if user has passed the limit and display message if needed “It seems that you have run out of budget of AI tokens. Please contact support@agatsoftware.com “
Operation is - new content or asking questions.