Assistants

Overview

An Assistant is the logical agent runtime deployed to your visitors. While a project manages infrastructure and keys, the assistant defines how the agent thinks, speaks, and behaves.

Anatomy of an Assistant

An assistant comprises four synchronized components:

1. Visual Entity: The active Avatar (model file, styling, sizing, and position).

1. System Persona: Custom instructions shaping tone, vocabulary, greeting cadence, and boundaries.

1. Grounding Scope: The linked Knowledge Sources queried on each message.

1. Execution Policy: Supported client-side commands (e.g., page navigation, highlighting buttons, triggering support modals).

Response Policies

Assistants operate under configurable response styles:

• Helpful & Concise (Default): Answers queries in 1–3 clear sentences with bullet points and documentation links.

• Technical & Detailed: Includes full code snippets, parameter specifications, and debugging steps.

• Conversational & Sales: Focuses on feature discovery, product comparisons, and scheduling calls.

Behavioral Controls

• Hallucination Protection: When a question cannot be answered from the indexed knowledge base, the assistant gracefully replies: "I don't have enough verified information to answer that question, but you can contact our support team at /contact."

• Language Localization: Automatically detects user browser language (over 40 supported languages) and translates answers on the fly.