Property Enquiry Chatbot

A conversational agent for property enquiries that describes matching listings, asks the qualifying questions a salesperson would, and offers to book a viewing — holding context across the whole thread so a constraint mentioned late reshapes the earlier suggestions instead of resetting them.

Property buyers browse late and enquire once. If nobody answers, they are on a competitor's listing within minutes. A scripted form-filler does not help, because the useful part of the conversation is the follow-up: a buyer who says "I want property in Ahmedabad" three messages in expects everything after that to respect it. Losing that context makes the bot worse than a plain contact form.

We built a conversational agent that describes matching properties, asks qualifying questions, and offers to schedule a viewing — holding context across the whole thread so a later constraint reshapes the earlier suggestions rather than resetting them. Replies stream over a WebSocket so the answer appears as it is written, and conversations persist so a returning buyer is not starting again.

Key decisions — and why

• Context across turns, not per message

  In the demo the buyer changes the city mid-conversation and the agent re-scopes its listings to match. That single behaviour is what separates a useful assistant from an autoresponder.

• Always close on a next step

  Every branch of the conversation ends by offering a viewing. A chatbot that answers questions without asking for the appointment is a cost centre.

• Streamed replies over WebSocket

  Tokens appear as they generate. A reply that writes itself out reads as fast even when the full answer takes a few seconds.