Pulse - Get Real-Time News Insights

In today’s fast-paced world, we consume news in fragments — often just a headline or a single article. Pulse aims to bridge this gap by offering a smart, contextualised view of global events. Users can search or interact with recent happenings and receive real-time, LLM-powered insights beyond isolated articles. What makes Pulse unique is its local-first architecture, powered by RSS feeds, making it extendable to any domain where RSS is applicable — finance, health, or niche blogs. At its core is a fast incremental query engine (Feldora) that powers lightning-fast retrieval. Relevant articles are embedded and stored in Zilliz (Milvus), enabling semantic search. We built this as a prototype full-stack application using: • Rust server for performance-critical backend, • Python server for LLM orchestration (using Novita AI), • Next.js UI for intuitive interaction, • Traefik for reverse proxy and service routing, • TRAE for intelligent AI-assisted developer support across backend and frontend. Most importantly, we leveraged Trae's AI coding assistance to iterate rapidly across the stack — from auto-generating backend boilerplate, optimizing queries, to scaffolding frontend components. Trae helped us move faster and focus on logic rather than wiring.