gpt.wedo.international — AI chat assistant · lead generation

GPT-powered consultant that explains WeDo’s AI services, answers a visitor’s questions and books a call with a rep through an in-chat form.

Role: Software architect and developer

Client: WeDo (in-house product)

Team: 2 developers

Period: 2023–2024

WeDo wanted to test whether an AI assistant could handle the first sales conversation. Visitors arriving with a rough idea needed someone to explain the agency’s AI services, the development process and how a project starts, and then capture contact details at the right moment. The tool had to be cheap to run, easy to iterate on and give the team a full view of every conversation.

The frontend is a Nuxt 3 app with Pinia, i18n and suggested questions, served from Cloudflare. Behind it, two Python microservices run in Docker. A WebSocket service keeps conversation history in Redis and calls the OpenAI Chat Completions API with a curated system prompt and function calling, which opens a contact form when the visitor wants to talk to a rep. A FastAPI admin service with JWT login lists conversations and leads. I designed the architecture and wrote the prompt.

Chat UI with suggested questions and a conversation restart

OpenAI function calling opens a lead form inside the chat

Conversation history in Redis, delivered over WebSockets

FastAPI admin with JWT login to review conversations and leads

Docker microservices, production compose and GitHub Actions

60 commits over six months; ran as a lead-generation experiment for WeDo’s AI services

Nuxt 3, Pinia, Python, WebSockets, FastAPI, OpenAI API, Redis, Docker, GitHub Actions