Skip to main content
Control how the AI agent behaves across your organization. These settings are available to admins only. Settings are grouped into collapsible sections. Open Agent capabilities to find Artifact resources.

​
Artifact resources

Artifact resources is off by default. Enable it to let artifacts use persistent records, private uploads and approved AI operations with streamed responses. The setting enables the capability; it does not grant viewers access to every resource. Existing sharing and resource permissions still apply. Disabling it blocks resource access without deleting stored records or files. The Allow File Upload setting below concerns files attached to a conversation. Uploads inside an artifact also require the app’s file resource and its permissions. See Build apps with artifacts for examples and updating installed skills if an existing organization still has older app-building guidance.

​
Demo data generation

Demo data generation is off by default and appears under Training Mode in Agent capabilities. When it is on, admins who can manage connections can ask the agent in a Training session to design and generate a realistic mock dataset for any domain, then create suggested agents on it. The rows are generated by the organization’s small default model. See Create a demo dataset.

​
Sub-agents

Sub-agents is on by default. When it is on, the agent can delegate independent sub-questions (for example one hypothesis each in a root cause analysis) to bounded background sub-agents that investigate on their own and report back as typed records in one turn. When it is off, the agent never delegates. See Sub-agents.
The Sub-agents section of AI Settings
The section holds the posture, the limits and the model:
  • Sub-agent posture (Default: proactive): proactive lets the agent delegate on its own when a request splits into three or more independent sub-questions. explicit makes it delegate only when the user or an instruction asks for it.
  • Sub-agent max steps (Default: 40): Planner steps one sub-agent may take before it is forced to finish with a partial record. 5 to 200.
  • Sub-agent deadline (minutes) (Default: 15): Wall-clock limit for a wave of sub-agents. When it fires, still-running sub-agents are stopped and recorded as timed out, and the agent answers from the records that exist. 2 to 60.
  • Concurrent sub-agents (Default: 3): How many sub-agents of one report may run at the same time. The rest wait inside their wave. 1 to 6.
  • Max sub-agents per wave (Default: 5): How many sub-agents the agent may launch from one planning step. Extra ones are refused with a note so the agent can plan a second wave. 1 to 10.
  • Sub-agent model (Default: auto): The model sub-agents plan on. auto uses the small default model for investigation and the regular default for validation; small or default force one; a model id from the LLM page pins it.
  • Allow sub-agent model override (Default: on): Let the agent pick a different model for one sub-agent when the user asked for it or an instruction names one. The pick is validated against the models the asking user may use.
  • Sub-agents may create queries (Default: off): Let sub-agents create tracked queries whose results the agent can chart and cite. Off, sub-agents are research-only: they describe, inspect, read and take notes.

​
Other AI settings

  • Allow LLM to See Data: Enable the AI to view sample data from tables for better context and more accurate analysis.
    If false, the system will use metadata only
  • Allow File Upload: Let users upload spreadsheets (xls/csv) and PDFs. The AI will parse and incorporate the data into analysis.
    When file upload is allowed, the LLM will see the data of the file (for indexing) - this cannot be avoided.
  • Allow Users to Edit and Execute LLM Generated Code: Users can view, edit, and run the SQL or Python code generated by the AI. Useful for advanced users who want more control.
  • Enable LLM Judge: The AI evaluates query quality for accuracy, efficiency, and best practices. Provides feedback to improve results over time.
  • Autogenerate Instructions: Automatically suggest saving user clarifications as instructions. For example: “Active users should exclude test accounts” becomes a suggested instruction.
    Suggested instructions require admin approval before becoming active.
  • Validate Code: Automatically validate AI-generated code for syntax errors, security issues, and schema correctness before execution.
  • Limit Analysis Steps (Default: 6): Controls how many sequential steps (queries, visualizations, calculations) the AI can take to answer a question.
  • Limit Code Retries (Default: 3): Limits how many times the AI can retry code generation after failures to prevent infinite loops.
  • Top K schema: Set how many schema tables are expanded and included in the initial system prompt.
  • Top K metadata resources: Set how many metadata resources (dbt, markdown, etc.) are expanded and shown in the initial system prompt.