Guide › Flows and AI

🤖 The AI Assistant node

The richest node on the platform. Default settings: OpenAI, GPT-3.5 Turbo, history of 10 messages, 500 output tokens.

AI Configuration

• AI Provider and Model
• Credential Source — Auto (Company → System → Environment) by default
• Response Language — set it to your language, otherwise it answers in English: it is independent of the interface language
• Prompt Template — a starting prompt for common cases
• Assistant Mode (System Prompt) — the real instructions, with a Generate with AI button

Conversation

• Enable conversation history and History messages limit (10 by default) — more context means more coherent answers and a higher cost
• Max Output Tokens (500 by default) — limits the length of the answer and, with speech synthesis on, the length of the audio too

Session Takeover

• Stop Keyword — the word with which the user ends the AI session. When the feature is on, the assistant handles every following message until that word arrives.

Optional features

• Image Understanding — image reading, off by default
• Voice Processing — voice note transcription with OpenAI Whisper, and spoken replies with TTS
• Task Execution — enables function calling and advanced routing
• Google Calendar and Zoho Calendar — appointment and availability handling. The calendar behaviour is imposed by the system; the prompt only adjusts tone and extra instructions.
• ERP Automation — gives the assistant the ERP tools during the conversation

Knowledge Base — RAG Enhancement Enriches the answers with your own documents. It requires Pinecone credentials, to be configured after switching the toggle on. Without a Pinecone account the feature does not start: take that into account.