HoloMemory

HoloMemory is an interactive playground for encoding, storing, and retrieving memories using Holographic Reduced Representations (HRRs). It demonstrates that vector-symbolic architectures can perform associative recall without embedding models or external ML APIs, purely through circular convolution and FFT.

• Frontend built with Next.js, React, TypeScript, and Tailwind CSS with Framer Motion animations
• Backend API in FastAPI (Python) handling HRR encoding/decoding with NumPy FFT
• Local-first SQLite storage for persisted memory traces
• Recall Duel mode comparing keyword search vs holographic retrieval side-by-side
• Interactive playground for encoding text into high-dimensional vectors and querying them

• No external ML APIs: all computation is local FFT-based convolution
• Chose HRRs over other VSA variants (MAP, BSC) for their clean frequency-domain math
• Split architecture (Next.js frontend + FastAPI backend) to keep NumPy computation server-side
• SQLite for persistence to keep the demo self-contained and deployable without a managed DB