How it works - Type in the chat; each message is sent to /api/receptionist which forwards the conversation plus visitor details to your n8n webhook. - The UI supports text/plain or JSON replies from n8n. If JSON, return with reply/message/text fields. - A “Test webhook” button sends a simple probe to verify connectivity. - Optional fallback: If an OpenAI API key is configured, the server will use the AI SDK’s generateText with openai('gpt-4o') to produce a response if the webhook is unreachable. [^2] Optional setup for AI fallback - Add OPENAI_API_KEY as a server environment variable to enable the fallback using the AI SDK. [^2] Deploy and install - Click Deploy to host this on Vercel, or Download Code to install locally. - No environment variables are required unless you want the AI fallback. Notes for your n8n workflow - Expect a POST with JSON containing: - sessionId: unique visitor session id - latestMessage: most recent user message - messages: full conversation history [{ role, content }] - visitor: { name, company, contact, purpose } - meta: { test, userAgent, receivedAt, clientTimestamp, referrer, ip } - Respond with either: - text/plain body for the chatbot to display, or - application/json: { "reply": "your message" } (reply/message/text keys supported) Reference - We use the AI SDK’s generateText with openai('gpt-4o') for the optional fallback path. [^2]

To configure the generation, complete these steps: