Rewind

Overview

Rewind solves the black-box problem of user bug reports: developers can't fix what they can't see. Instead of relying on vague user descriptions and fragmented logs, Rewind acts as a time machine. A single script tag captures full DOM mutations, network requests, console logs, and custom events, allowing developers to play back exact user sessions. The platform is architected as a highly scalable Turborepo monorepo. On the client side, a lightweight 12kB vanilla JS tracker built with rrweb captures interactions. This data is streamed via WebSockets to a Node.js/Express ingestion gateway, which buffers bursts using Redis and BullMQ. A background worker then persists events into a PostgreSQL database managed with Drizzle ORM. The frontend is a Next.js 15 App Router dashboard that provides an interactive replay player with synchronized network and console side-panels, conversion funnels, and AI-generated session summaries powered by pgvector. A core challenge was ensuring the system remained lean enough to self-host on a $6 VPS while still handling high-throughput ingestion. This was achieved through intelligent batching, zlib compression, and a decoupled worker queue. The entire stack is deployed via a single-command multi-stage Docker Compose setup, giving small teams enterprise-grade observability without the SaaS price tag.