Recall
AI Second Brain Ecosystem
A modular monolith with an asynchronous pipeline for deep knowledge extraction, featuring a vector-driven knowledge graph and remote content ingestion.
Next.js
Express
BullMQ
Cloudflare R2
TypeScript
Redis
Recall is a comprehensive second brain ecosystem designed to streamline knowledge capture and retrieval. It bridges the gap between raw information gathering and structured learning by employing an async pipeline.
How it's built
Built on a modular monolith architecture. Next.js handles the frontend with server-side rendering for SEO and initial load speed. The backend is powered by Node.js/Express, utilizing BullMQ for asynchronous task processing (e.g., website scraping, PDF parsing). Pinecone serves as the vector database for semantic search, and Cloudflare R2 provides cost-effective object storage.
What it does
- Semantic Vector Search using Pinecone
- Force-Graph visual knowledge representation
- Chrome Extension for 1-click web scraping
- Async background processing with BullMQ & Redis
Where it got hard
Managing latency during vector embeddings generation and ensuring the Force-Graph rendered smoothly with hundreds of nodes without freezing the UI.
Implementing BullMQ drastically improved my understanding of resilient background job processing and how to decouple heavy AI-processing tasks from the main request thread.
What's next
Integrating local LLM support to run entirely offline, and adding collaborative workspaces for teams.