v1.0 live! Celebrating with 50% off -limited offer
-
Get Started

Workspace Chat Settings Page

The Chat Settings tab controls how the AI assistant behaves in this workspace. You can choose the AI model, chat mode, system prompts, temperature, history, and how the AI responds when it refuses to answer.

Key features

  • AI model selection — Choose which LLM powers this workspace
  • Chat mode — Toggle between Chat (conversational) and Query (factual)
  • System prompts — Customize the AI's base instructions
  • Temperature — Control response creativity
  • Chat history — Enable or disable conversation memory
  • Refusal response — Customize what the AI says when it declines to answer

AI model configuration

Model selection

Select the AI provider and model used for chat in this workspace. Options depend on your system configuration (e.g., OpenAI, Anthropic, Azure, Bedrock).

To change:

  1. Open the model dropdown or search menu.
  2. Pick a provider and, if available, a specific model.
  3. Click Update workspace when done.

System default

  1. — Use this to inherit the instance-wide LLM preference for the workspace.

Chat behavior

Chat mode

Two modes control how the AI responds:

| Mode | Behavior | |------|----------| | Chat | Conversational and contextual; uses document context and chat history | | Query | Direct and factual; uses document context only, no chat history recall |

When to use:

  1. Chat — Ongoing conversations, follow-ups, and discussions
  2. Query — Quick lookups and one-off factual questions

Chat history settings

Controls how many previous messages the AI remembers for context (1–45, default 20). Lower values reduce token usage; higher values give more context for long conversations.

Temperature

Controls randomness and creativity of responses (typically 0–2). Lower values produce more focused, deterministic answers; higher values allow more varied, creative output.

System prompts

Chat system prompt

The system prompt defines the AI's base instructions. It tells the model how to behave, what context to use, and how to format answers.

To change:

  1. Edit the System prompt textarea.
  2. Click Update workspace to save.

Tips:

  1. Be specific about tone, format, and constraints.
  2. Reference "the context above" or "the documents" so the model uses embedded documents correctly.

Query refusal response

When the AI cannot find relevant information in documents, it may refuse to answer. This setting lets you customize that refusal message.

Saving changes

All Chat Settings changes are saved together:

  1. Adjust any settings in the tab.
  2. Click Update workspace (shown when there are unsaved changes).
  3. Wait for the success toast.

Troubleshooting

  1. Model not appearing — Ensure the provider is configured in Instance Settings → LLM.
  2. Chat mode not taking effect — Changes apply to new messages; existing threads keep their prior behavior.
  3. Prompt changes not reflected — Save the form and start a new message or thread to test.

For more information, see:

  1. Workspace Settings Introduction — Overview of all settings tabs
  2. General Appearance — Suggested messages and workspace identity