Operating Modes

Overview

chans supports two operating modes, configured per-agent:

Enhanced Mode (Default)

In Enhanced mode, chans handles the complete voice AI pipeline:

User speaks → STT → LLM → TTS → User hears response

Features

• System Prompt -- Configure the agent's personality and instructions
• RAG -- Upload documents for knowledge-grounded responses with hybrid search
• Conversation Memory -- Semantic search over past conversations for context
• MCP Tools -- Connect external data sources and APIs

When to Use

• You want to get started quickly without setting up your own LLM
• You need built-in conversation memory across sessions
• You want document-grounded responses via RAG
• You want to use MCP tools for external integrations

Passthrough Mode

In Passthrough mode, you bring your own AI/LLM:

User speaks → STT → Your Webhook → TTS → User hears response

chans handles voice (STT + TTS) while you control all AI logic.

How It Works

1. User speaks into your app
2. chans transcribes speech to text
3. chans sends transcript to your webhook
4. Your webhook returns the response text
5. chans converts response to speech

When to Use

• You have an existing LLM/AI system
• You need custom business logic for responses
• You want full control over the AI behavior
• You're integrating with a specific AI provider or model

Configuration

Set up passthrough mode in your agent settings:

1. Set agent mode to "passthrough"
2. Configure your webhook URL
3. Optionally add an API key for authentication

See Webhook API for the complete request/response format.

Choosing a Mode

Most users start with Enhanced mode to validate their use case, then migrate to Passthrough when they need more control.