# Text Agent Settings Overview {% embed url="" %} Text agents in Dapta can be customized using several settings. Each setting controls a specific part of how your agent looks, behaves, or responds to users. This guide explains **each setting in simple terms**, what it controls, and how it affects your agent after it has been created. > **Important concept to understand first**\ > Once an agent is created, the **prompt becomes the source of truth**.\ > Some settings only help generate the initial prompt and do not automatically change agent behavior afterward. *** ### Language

#### What this setting controls The Language setting controls the **language of the chat widget interface**, such as: * Button labels * Placeholder text * System UI text shown to users It does **not** control the language your agent speaks. *** #### How agent language actually works When a text agent is created, the language it uses to communicate is defined **inside the agent prompt**. After the prompt exists: * Changing the Language setting **will not** change how the agent responds * The agent will continue using whatever language rules are written in the prompt This works the same way as the **Agent Purpose** setting. *** #### How to change the agent’s language To make an agent: * Speak a different language, or * Support multiple languages You must update the **agent prompt**. You can do this by: * Editing the prompt manually, or * Using **Improve with AI** to rewrite the prompt with the new language instructions *** #### Multilingual agents Agents can be multilingual if this behavior is defined in the prompt. For example: * Respond in English when the user writes in English * Respond in Spanish when the user writes in Spanish This cannot be enabled using the Language setting alone. *** #### Common mistake Changing the Language setting and expecting the agent to automatically start responding in that language. If the prompt is not updated, the agent’s behavior will not change. *** ### Identity Name (Agent Name)

#### What this setting controls The Identity Name is the **name of your agent**, which defines how it identifies itself during conversations. *** #### How it works * When you create an agent, you are prompted to enter its name * This name is stored **inside the agent prompt**, not as a separate system field *** #### Editing the agent name after creation Changing the Identity Name setting alone does not update the agent. To rename an agent after it has been created, you must: * Edit the name directly inside the prompt, or * Use **Improve with AI** to update the name across the entire prompt *** #### Best practices * Use human-friendly names * Match the name to the agent’s role (for example, support, sales, scheduling) * Keep naming consistent across agents *** ### Agent Purpose

#### What this setting controls Agent Purpose helps you **generate an initial prompt** based on a common use case. Available purposes include: * Ecommerce Sales Agent * Set Up My Own Assistant * Qualify Inbound Leads * Support Sales Inquiries in Ecommerce * Customer Service and Support * Website Q\&A Assistant * Schedule Meetings * Confirmations * Reminders *** #### Important behavior to understand Once the agent prompt is created: * Changing the Agent Purpose **does not automatically update the prompt** * The agent will continue behaving according to the original prompt *** #### How to change an agent’s role If you want your agent to perform a different task: * Use **Improve with AI** to rewrite the prompt for the new purpose, or * Manually edit the prompt to reflect the new behavior *** #### Common mistake Changing the Agent Purpose dropdown and expecting the agent’s behavior to change without updating the prompt. *** ### Model

#### What this setting controls The Model setting determines **which AI model powers your agent**. Different models can vary in: * Speed * Cost * Reasoning quality * Response accuracy *** #### How to choose a model There is no single best model for all use cases. The best approach is to: * Test different models using real conversations * Evaluate performance based on your specific needs * Upgrade only if you notice limitations *** #### Best practice Start with a default model, test real interactions, and switch models only if needed. *** ### Temperature (Creativity Level)

#### What this setting controls Temperature controls **how creative or strict** your agent is when generating responses. You can think of it as a creativity dial. *** #### Low temperature * More predictable responses * Follows instructions very closely * Best for structured tasks such as: * Customer support * Lead qualification * Data collection *** #### High temperature * More flexible and creative responses * More conversational tone * Best for: * Brainstorming * Open-ended conversations * More natural dialogue *** #### Best practice * Use **lower temperature** when accuracy and consistency matter * Use **higher temperature** when creativity and tone are more important *** ### Key Takeaways * The **prompt controls behavior**, not the settings alone * Language and purpose changes must be made inside the prompt * Identity Name lives inside the prompt * Temperature controls creativity, not intelligence * Model choice should be based on real testing