Generative AI logs table

  • Versão de lançamento: Australia
  • Atualizado 12 de mar. de 2026
  • 1 min. de leitura

    • The Generative AI Logs [sys_generative_ai_log] table contains all execution logs for web search and can be used for debugging.

    To view the Generative AI logs table, admins can search for sys_generative_ai_log.list in the navigation filter. The following table describes some of the default columns that appear.

    Field Description
    Conversation Direct reference to the conversational record in the Conversations [sys_cs_conversation] table.
    Definition OneExtend capability definition.
    Time Taken Length of time taken to complete the OneExtend capability definition call. Time is recorded in milliseconds.
    Metadata document The record from the table that the LLM call was made for. The related table is listed in the Metadata table column.
    Prompt Inputs that are sent to the LLM AI provider.
    Response Outputs that are received from the LLM AI provider. Outputs can be direct (original) outputs or processed outputs.
    Error Description of an error.
    Error Code Error code number.

    The following table includes common issues that the Generative AI Logs table may help you find solutions for.

    Issue Possible cause Suggested solution
    Provider not supported error appears in the Error column. Using an AI provider that doesn't support web search, such as Azure. Switch the web search AI provider to Gemini either via Edit model providers or Configure AI search answers capability for web search.
    Empty or null response appears in the Error column. Missing or invalid API key. Verify that the API keys are correct. For more information on how to add API keys, see Configure AI search answers capability for web search.
    Timeout error appears in the Error column. Using a complex query or experiencing a slow response. Encourage end users to simplify the query.
    Incorrect answers appear in the Response column. Limiting query or contextual information was provided. Encourage end users to provide more context or refine the query.