Institutional AI Assistant
A 24/7 conversational assistant for the Learning Planet Institute. RAG architecture powered by the site's content.
Architecture
The system is built on a Retrieval-Augmented Generation (RAG) architecture. Website content is crawled, indexed in a vector database and retrieved dynamically to generate contextual responses.
Crawl → Retrieval → Generation
How It Works
1. Crawling: Automated extraction of website content (pages, news, educational resources).
2. Generation: The LLM synthesises a response framed by the institutional system prompt.
System Prompt
The system prompt defines the assistant’s tone, values and boundaries. Answers must remain factual and grounded exclusively in indexed content. Complex requests are redirected to a human contact, and every response must respect the institute’s editorial voice. This keeps the assistant aligned with the institution’s identity.
Impact
The assistant provides 24/7 access for international visitors, reduces pressure on the support team and improves access to information. It handles recurring questions and directs specific requests to the appropriate members of staff.
My Role
Creative Tech Lead
Prompt engineering
Components
Chatbase crawling pipeline
Retrieval-Augmented Generation
Institutional system prompt
Web integration
Context
Learning Planet Institute
Paris, France