hieuchaydi/MemoryFeed

MemoryFeed is a local-first social memory system for capturing, indexing and resurfacing personal social-media content. Intended users include privacy-conscious individuals, researchers, and integrators. Core components: browser extensions (Chrome/Firefox) capture posts after a 3s dwell; FastAPI backend with SQLite (WAL) + FTS5 for storage and full-text search; LanceDB for multilingual semantic vectors; React+Vite frontend for search, Active Feed and timeline; optional vision (Gemini) image captioning and Groq/Qwen text rewrite/summarization; optional C++ native acceleration. Key features: metadata (star/note/tags), export/import, Active Feed ranking (heat, decay, resurfacing gap, context), FTS + semantic search, explainable search debug output, diversity-aware reranking, integrity verify/repair, maintenance scheduling, MCP RAG tool for bounded retrieval, Docker/dev scripts, and a standalone memoryfeed-core Python library for integration. Privacy modes: offline/local-only, hybrid AI (selective cloud providers), and optional public deployment.

AI-related keywords found in this repo's description, topics, or README summary — grouped by category. Each badge links to the corresponding ranking detail page.