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

# Chat Behavior

> Settings that shape how Ana reasons and responds.

Control how Ana reasons and responds — how thorough versus fast it is, and how many assumptions it makes versus clarifying questions it asks.

Chat behavior settings shape how Ana approaches a task, as opposed to Models (which AI model runs) or the tool gates that control what the agent can access. They are applied as instructions layered onto Ana's system prompt at the start of a conversation, so they influence the agent's reasoning and tone rather than granting or restricting access to any tool or data source.

These settings are provider-agnostic. They apply the same way whether a chat runs on Anthropic (Claude), OpenAI (GPT), Google (Gemini), or Moonshot (Kimi) models, and switching models mid-conversation does not change which behaviors are active.

Both settings are configured by administrators under **Settings → Capabilities.**

## Training Mode

Makes Ana ask clarifying questions and avoid assumptions.

When on, the agent is told it may lack full context, so it stays cautious about assumptions and asks clarifying questions before answering rather than guessing at intent. It's an org-wide system-prompt nudge — not a tool gate — so it changes how Ana reasons, never what it can access.

| Training Mode | Behavior                                                                                            |
| ------------- | --------------------------------------------------------------------------------------------------- |
| On            | Ana flags missing context, surfaces its assumptions, and asks clarifying questions before answering |
| Off           | Ana answers directly, inferring intent from the request without pausing to confirm                  |

Training Mode is useful when onboarding new users, when data models are unfamiliar, or for high-stakes analysis where a wrong assumption is costly. It trades a small amount of speed for fewer misdirected answers.

## Response Methodology

Adds a per-chat dropdown to pick how thorough versus fast and cautious replies are.

When enabled, a Methodology dropdown appears in the chat input, letting users choose per conversation how thorough versus fast the agent is and how many assumptions it makes versus clarifying questions it asks. When disabled, the dropdown is hidden and all chats run on the default methodology.

The available methodologies are:

| Methodology       | Behavior                                                                                             |
| ----------------- | ---------------------------------------------------------------------------------------------------- |
| Adaptive          | Ana judges per request how much depth and clarification each task needs                              |
| Prescriptive      | Ana moves quickly and decisively, making reasonable assumptions to deliver a direct answer           |
| Thorough          | Ana prioritizes depth and completeness over speed                                                    |
| Careful           | Ana minimizes assumptions and asks clarifying questions before acting                                |
| Ontology Building | Available with Ontology 3.0; Ana works with you to build and refine the semantic layer as it answers |

### Default Methodology

The methodology new chats start with.

***

## Frequently Asked Questions

Frequently Asked Questions

#### Do these settings depend on which model a chat uses?

No. They are provider-agnostic and apply consistently across Claude, GPT, Gemini, and Kimi. Switching models mid-conversation does not change which behaviors are active.

#### Where does the Ontology Building methodology come from?

It is only available to organizations on Ontology 3.0. Organizations on earlier ontology versions will not see it in the dropdown.

#### What happens if Response Methodology is disabled?

The Methodology dropdown is hidden from the chat input and every chat runs on the default methodology. Members cannot override it per chat until the setting is re-enabled.