Profile Editor - AI Chatbot

📋 Basic Information

Profile Name *

Description

💬 System Prompt

Main Instructions *

This is the primary instruction set for the AI.

📄 Resources

Add text or CSV documents that the chatbot can reference.

⚙️ Built-in Tools

Server-implemented tools. These are opt-in: their names and descriptions are shown to the AI only when enabled here, and they shape how the AI sees its role.

🔊 Voice (TTS)

The text-to-speech model and voice used when this profile is on a voice call. Takes effect on the next call.

Model

Voice

Preview text

Changing the voice plays this line so you can hear it.

🧠 Reasoning

How the model reasons before answering. Takes effect on the next message / call.

Enable thinking Native hidden reasoning (<think>). Improves grounding; shown in the context panel. Best for text chat. Silent: adds latency on voice. Ignored when "think out loud" is on. Think out loud Reason audibly inside the reply instead of a hidden phase. No dead-air pre-phase, so low-latency for voice. Overrides "enable thinking".

📞 Call Greeting

The opening line the assistant speaks when a voice call (SIP or WebRTC) connects. It's recorded as the first message of the transcript, so the assistant knows it already greeted.

Greet the caller on connect Off means the assistant stays silent until the caller speaks.

Greeting type

Static Speak a fixed line, every call. Fastest — no model call, so the caller hears the greeting the instant the call connects. LLM-driven Generate from a prompt at connect (thinking off). Adds latency — the caller waits on the model before hearing anything (capped ~6s, then falls back to the static line). Use only when a dynamic greeting is worth the slower connect.

Greeting

Plays this line in the profile's voice.

Greeting prompt

Generates a sample so you can see (and hear) what a caller would get.

🔧 Tools

Select tools that this profile can use during conversations.

📋 New Profile