> ## Documentation Index
> Fetch the complete documentation index at: https://docs.marsmind.co/llms.txt
> Use this file to discover all available pages before exploring further.

# Base Persona

> Set AI style, custom instructions, context, and derived prompts.

Set the assistant style, identity, and response principles.

Open: **My Assistant** → **Base Persona**.

<Frame>
  <img src="https://mintcdn.com/marsmind/mSdi0SY3kwN3OIoA/images/admin-guide/2026-10/en/persona.png?fit=max&auto=format&n=mSdi0SY3kwN3OIoA&q=85&s=6c8f72329a869b132e4934c368a8b60b" alt="Base Persona: AI style, custom instructions, context settings, and derived prompts" width="1440" height="1000" data-path="images/admin-guide/2026-10/en/persona.png" />
</Frame>

The screenshot uses demonstration data.

## Set style and custom instructions

<Steps>
  <Step title="Select a channel">
    Select the channel to configure under **Selected window** in the sidebar.
  </Step>

  <Step title="AI Style">
    Choose **Detailed and polite** or **Flexible and concise**. This choice saves separately and does not save your instruction draft.
  </Step>

  <Step title="Custom Instructions">
    Enter the assistant identity, service scope, and response principles. For example: act as an after-sales adviser, confirm the order and issue, follow the knowledge base, and ask for missing information.
  </Step>

  <Step title="Generate a candidate if needed">
    With text entered, select **AI Polish**. Review the comparison and select **Apply candidate** to replace the editor text, or **Keep current draft** to retain it.
  </Step>

  <Step title="Save Configuration">
    Select **Save Configuration** and wait for confirmation. To restore older text, open **History Records**, choose a version, apply it, then select **Save Configuration** again.
  </Step>
</Steps>

<Info>
  A successful save confirms the configuration was stored. Runtime application remains unconfirmed. Applying AI or history candidates only changes the draft.
</Info>

## Context Settings

<Steps>
  <Step title="Expand and select">
    Expand **Context Settings** and the relevant category. Check the conversation context or user information to include.
  </Step>

  <Step title="AI Suggest">
    Select **AI Suggest** for suggested selections. Review the checked items and reasons, then adjust them.
  </Step>

  <Step title="Save Configuration">
    Select this section's **Save Configuration**. Unknown or disabled items can only be unchecked; remove them before saving.
  </Step>
</Steps>

<Info>
  Saving the main instructions can automatically analyze and save context settings. Review this section afterward. Manual AI suggestions still require saving.
</Info>

## Derived Prompts

<Steps>
  <Step title="Re-derive">
    With main instructions entered, expand **Derived Prompts** and select **Re-derive**.
  </Step>

  <Step title="Review each candidate">
    Expand items marked **New** and review or edit their text. **Adopt** saves that item; **Discard** removes the candidate and keeps the stored configuration.
  </Step>

  <Step title="Edit an individual prompt">
    You can also edit an individual item and select **Save**. **Clear** clears that item so it uses the system default.
  </Step>
</Steps>

<Warning>
  Manual saving or clearing marks this item as manually locked. This page cannot unlock it; check the content before continuing.
</Warning>

<Info>
  Candidates are generated only when you select **Re-derive**. Generation does not save them. Manually locked items cannot be unlocked here. If lock status fails to load, select **Reload config**; writes remain blocked until it succeeds.
</Info>

## If settings fail to load

If settings fail to load, select **Retry**. Editing and saving remain unavailable until loading succeeds.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.