{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/a4b0c6e042534c79be4d2ed06ff662b5\" frameborder=\"0\" width=\"1728\" height=\"1296\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1296,"width":1728,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1296,"thumbnail_width":1728,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/a4b0c6e042534c79be4d2ed06ff662b5-bbc9f4f5e91afc03.gif","duration":179.305,"title":"ScholarsIQ200 – A Multimodal Chatbot with Retrieval‑Augmented Generation (RAG) and Tool‑Calling Capabilities","description":"I built ScholarsIQ200, a serverless, LLM‑powered chatbot written in TypeScript and Next.js and deployed on Vercel. It uses an architecture that combines Upstash Vector for retrieval‑augmented generation (RAG) with the Grok API for the LLM, enabling answers from private documents and live web browsing.\n\nAt one point it abruptly stopped working because the vision model had been deprecated, and I fixed the issue with a recent commit. The system is fine‑tuned for the tech‑industry audience, includes LaTeX formatting for chemical equations, supports voice input and PDF export, and allows exporting the full chat history."}