{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/9d307bc4f47040668cc92c50b9711800\" 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/9d307bc4f47040668cc92c50b9711800-37b14da704fd9fb8.gif","duration":154.84,"title":"Privacy-First Document Q&amp;A on Cloudflare","description":"Hi everyone, I built Walltrack for small law firms and healthcare clinics that need AI documentation search but cannot send confidential files to OpenAI or other third party providers. I explain why the default RAG stack is not enough and that self hosted setups require heavy GPU and can cost about $6 million for enterprise platforms. Walltrack runs entirely on Cloudflare, with no data leaving the Cloudflare network, using products like Workers, Workers AI, Vectorize, R2, D1, AI Gateway, KV, and Secrets. I show uploading a PDF, chunking and embedding it, then asking a question and getting answers from Llama 3.1. We also log everything for compliance and auditing, and there is no action requested from viewers."}