All work

Shift Green Hub

AI chatbot
Community platform
Web design
Visit live site
Shift Green Hub website

AI-assisted discovery for a sustainability community

Shift Green Hub connects people with sustainability events, articles and community initiatives. SouthSwell built the community website and an AI chatbot that helps visitors find relevant events and reading through a conversation.

Try ShiftGreenBot.

How the assistant works

A visitor describes what they are interested in. The assistant interprets the request, searches the platform’s event and article content, and prepares a conversational reply with recommendation cards linking back to the relevant pages.

The implementation uses retrieval-augmented generation: it retrieves relevant platform content before generating the reply. Conversation history provides context for follow-up questions. The current searchable content types are events and blog articles; the portfolio does not claim that the assistant can make bookings, send communications or complete purchases.

What we delivered

The website design, content and event pages remain the foundation. The chatbot adds a conversational discovery experience alongside them, giving visitors another way to explore the platform. This is implemented AI work, distinct from the exploratory agents described in other portfolio projects.

Technology stack

  • Web experience: Next.js, React and TypeScript; a client-side chat interface with linked recommendation cards
  • Backend: NestJS and TypeScript, with a dedicated chat-search endpoint
  • AI: Azure OpenAI for interpreting requests and generating replies; GPT-4o is the repository’s default chat deployment
  • Semantic retrieval: Embeddings and MongoDB Atlas Vector Search over a unified content index; text-embedding-3-large is the repository’s default Azure embedding deployment
  • Data access: Mongoose models retrieve the matching event and article records
  • Local development: An alternative Ollama provider, defaulting to llama3.2:3b and nomic-embed-text

The service follows three stages: interpret the request, retrieve matching content, and generate a reply with grouped recommendations. These details describe the inspected repository implementation. Model deployments are configurable; production settings may differ from the defaults.

Business value

ShiftGreenBot gives visitors a direct route to relevant events and articles, without searching through multiple pages. It makes the platform’s content easier to discover and helps turn a visitor’s interest into a clear next step: read an article or explore an event.

“Think of a crowdfunding platform for sustainability, targeting a very focused audience.”

Shift Green Hub’s Founder

Client

Shift Green Hub

Services

AI chatbot, Community platform, Web design
0%AI implementation
0%Feature
0%Strategy
0%Design

Want results like this for your team?

Start your project