Generative AI Controller tables
Generative AI Controller use dedicated tables to log AI activities and track Now Assist usage across Now Assist capabilities.
Generative AI Log [sys_generative_ai_log] table
Logs generative AI prompts, responses, and edited responses for debugging LLM calls. HR-related log entries are restricted to HR administrator to protect sensitive information.
| Column | Data type | Description |
|---|---|---|
| Additional Data | String | Contains concatenated values for actions performed on the generated LLM response, such as copied or edited. This field is populated through the feedback API. |
| Caller | String | Contains information about the calling application that generated the log entry. |
| Completed At | Date/Time | Timestamp when the generative AI execution completed. |
| Conversation | Reference | Reference to the conversation this log is part of. |
| Created | Date/Time | Date and time when the log entry was created. |
| Created By | String | Contains user information that initiated the execution or flow. |
| Definition | Reference | Reference to the capability definition responsible for generating the log entry. |
| Derived Scope | Reference | Associated scope of the application that generated the log entry. |
| Domain | Domain ID | Domain associated with the execution that generated the log entry. |
| Domain Path | Domain Path | Path associated with the domain from where the capability is executed. When domain separation is enabled, this column is automatically populated by the platform. |
| Edited Response | String | Captures the edited version of the response when a user modifies it. |
| Error | String | Error message, if the request encountered an error. |
| Error Code | String | Error code associated with the error message. |
| External | True/False | Error code associated with the error message. |
| Feedback | Choice | Feedback on the LLM response. Possible values: Accepted, Rejected, Ignored, SlightlyPositive, SlightlyNegative, LikelyPositive. |
| Feedback Timestamp | Date/Time | Timestamp when the feedback was submitted. |
| Gen AI Usage Log | Reference | Reference to the Gen AI Usage Log. |
| Metadata Document | Document ID | Document ID for metadata used for the requested execution, if only a single document is used. |
| Metadata Documents | Glide List | Document IDs for metadata used for the requested execution, if multiple documents are used. |
| Metadata Table | Table Name | Table name for the metadata. |
| Model Name | String | Name of the model used for processing the generative AI request, such as gpt_large, claude_small, or gemini_small. |
| Model Version | String | Version of the ServiceNow LLM model used. |
| Output Metadata | String | Metadata received from the model or LLM, including tokens, processing time, and response time. |
| Prompt | String | Final prompt passed to the model. |
| Prompt Config | Reference | Reference to the prompt configuration, which contains settings such as model and temperature used in the request to the LLM. |
| Prompt Token Count | Integer | Number of tokens used in the prompt. |
| Response | String | The response generated by the LLM. |
| Response Token Count | Integer | Number of tokens in the response. |
| Skill Config Id | Reference | Reference to the skill configuration. |
| Source | String | Source feature responsible for the execution, such as ChatSummary, Now Assist in AI Search, or WWNA: Refine. |
| Started At | Date/Time | Timestamp when the generative AI execution started. |
| Status | String | Status of the AI log. Possible values: Success, Error. |
| Target Language | String | Target language for the AI log. |
| Time Taken | Integer | Time taken for the AI log process. |
| Untranslated Prompt | String | The untranslated version of the prompt. |
| Updated | Date/Time | Date and time when the log entry was last updated. |
| Updated By | String | User who made updates to the log. |
| Updates | Integer | Number of updates made to the log record. |
Gen AI Usage Log [sys_gen_ai_usage_log] table
Tracks each use of a Now Assist capability triggered from Workflow Studio, Virtual Agent Designer topics, or scripts such as business rules and UI actions.
| Column | Data Type | Description |
|---|---|---|
| Assists | Integer | Number of assists used or consumed. |
| Created | Date/Time | Date and time when the log entry was created. |
| Created By | String | Contains user information that initiated the execution or flow. |
| Document | Document ID | Document ID for metadata used for the requested execution. |
| Document Table | String | Table name for the document table. |
| Domain | Domain ID | Domain associated with the execution that generated the log entry. |
| Domain Path | Domain Path | Path associated with the domain from where the capability is executed. |
| Execution Type | Choice | Type of execution. Possible values: Test, Evaluation, Auto Prompt Tuning. |
| Feature | Reference | Source feature for the assist usage. |
| Line Item Name | String | License line item name associated with the assist usage. |
| License Name | String | License name associated with the assist usage. |
| Plugin Name | String | Plugin name that the feature responsible for the assist usage is part of. |
| Skill Config Id | Reference | ID of the skill responsible for the assist usage. |
| Status | Choice | Status of the usage log. Possible values: Completed, Pending. |
| Strategy | Choice | Strategy used for pricing the assist usage. |
| Trial Assists | Integer | Number of assists counted as trial assists. |
| User | Reference | User associated with the assist usage. |