What are we doing in a nutshell?
- Our AI-driven hotel intelligence system seamlessly integrates real-time web scraping, semantic search, and LLM-powered contextual understanding to deliver precise, dynamic insights. Using Puppeteer, we extract hotel reviews, transform them into high-dimensional OpenAI embeddings, and store them in Pinecone for lightning-fast vector retrieval. When a user queries a hotel, we enrich the prompt with hotel-specific metadata, retrieve the top 5 semantically relevant insights, and leverage GPT to generate highly contextualized responses. Powered by Node.js, Express, MongoDB, Pinecone, and React.js, this end-to-end pipeline ensures real-time, AI-enhanced decision-making for users.
This is a service that communicates with client directly.
Why seperate codebase?
- We are using pupeteer (headless browser) for srcaping the website, thus we need to deploy this as a seperate endpoint/ service.
Request comes in
-
- Hotel page url (to scrape & store data in db)
- It calls server 2 that scrapes & stores data to db & pinecone
-
- AI ask endpoint, which is used by client to search from vector db & answer based on the context.