{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/e25c8b6ad4804451a3d5d778a345d818\" frameborder=\"0\" width=\"1660\" height=\"1245\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1245,"width":1660,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1245,"thumbnail_width":1660,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/e25c8b6ad4804451a3d5d778a345d818-738f65c9bd244622.gif","duration":405.01,"title":"Building a Hindi Trending Tags App with AI","description":"This Loom presents a live trending tag system for a Hindi-speaking ShareChat-style app and explains its end-to-end workflow. On every app open, it fetches signals from Google Trends (India) and an India RSS/News API, then scores topics using recency, search volume, and news coverage with a confidence bonus for overlap across sources. A local Olama Gemini-style AI layer generates Hindi hashtags and descriptions, assigns a Heat score from 1 to 100, and produces 3 to 4 sentence summaries, outputting ranked JSON of 12 topics. It was deployed on Vercel with analytics tracking visitors, page views, bounce rate, device and browser data, and includes UI changes like hero images, Heat badges, full detail views, video-first feeds, bookmarks, and a language toggle. It also mentions planned future additions like a Reddit India signal source, personalization from taps and saves, and a create flow for reels."}