The Problem: AI Amnesia
We’ve all been there: You're 12 chats deep into a complex project. You’ve explained your architecture, your database schema, and that one obscure bug you finally fixed at 2 AM. Then you open a new chat, and the AI starts from zero. You spend the next 20 minutes re-explaining everything.
I got tired of this "context bankruptcy," so I built SYNQ.
What is SYNQ?
SYNQ is an open-source Chrome extension + local backend that gives your AI assistant (Claude, ChatGPT, or Gemini) persistent memory. It captures your conversations, distills them into a semantic knowledge graph, and automatically injects relevant context into your next prompt using a zero-loss RAG pipeline.
The Tech Stack
To keep everything private and fast, I went with a fully local-first architecture:
Backend: Node.js & Express 5.
Extension: TypeScript & Chrome Manifest V3.
Vector DB: ChromaDB for semantic search.
Graph DB: Neo4j for entity relationship mapping.
Local Embeddings: Ollama (nomic-embed-text).
LLM Processing: Groq LLaMA 3.1 (for high-speed graph extraction).
Key Features I’m Proud Of
Zero-Loss Chunker: Unlike most RAG apps that use lossy LLM summarization, SYNQ uses a sliding window chunker. This ensures that even small personal details—like the name of your cat—are preserved.
Auto-Connect: It intercepts your prompts in real-time. Before the AI sees your message, SYNQ silently prepends the top-3 most relevant historical "memories".
Knowledge Graph: I used D3.js to build a visualization of your project history. You can see how entities like Database, User Auth, and Stripe connect over time.
Privacy Scrubbing: It automatically redacts API keys and JWTs before anything is processed.
Why Open Source?
I’m currently an engineering student preparing for placements. Building this helped me manage my own project workflows, but I realized that "Context Sovereignty"—the idea that you should own your chat history, not the LLM provider—is something we all need.
Check it out on GitHub:
⭐ https://github.com/Eshaan-Nair/Synq
I’d love to hear your thoughts on the pipeline or any suggestions for new features!
This article was originally published by DEV Community and written by Eshaan.
Read original article on DEV Community