Your personal, fully local AI memory that never forgets.
Team: AI MINDS
Constraint: Fully local, open-source LLM < 4B params — zero proprietary APIs
We constantly capture information across images, text, audio, documents, and web pages. Over time, this creates a massive personal information space that's nearly impossible to organize or search. Important knowledge is lost — not because it's unavailable, but because it can't be retrieved when needed.
Synapsis is a zero-touch cognitive assistant that automatically ingests your personal data across multiple modalities, understands its meaning, and builds a persistent, structured memory you can query with natural language.
🚀 No uploads. No tagging. No manual work. Just use your computer — Synapsis learns in the background.
| Feature | Description |
|---|---|
| 📂 Automatic Multimodal Ingestion | Watches your directories for text, PDFs, images (OCR), audio, and documents — zero manual uploads |
| 🧩 Semantic Understanding | Interprets content meaning and purpose regardless of format |
| 🧠 Persistent Memory | Structured knowledge graph (vector + graph DB) that survives restarts and grows over time |
| 💬 Natural Language Q&A | Ask questions and get grounded answers with full source citations |
| ✅ Self-Verification | Reasons about relevance and verifies its own answers before presenting them |
| 🔗 Proactive Insights | Automatically surfaces connections, contradictions, and actionable items across your data |
| 🔒 Fully Air-Gapped | Zero internet required — everything runs locally on your machine |
| Layer | Technology | Purpose |
|---|---|---|
| 🤖 LLM | -000?style=flat-square&logo=ollama){kind=icon} |
Local language model for reasoning & generation |
| 📐 Embeddings |  |
Semantic embeddings (384-dim, fully local) |
| 🗄️ Vector DB |  |
Vector similarity search (on-disk persistence) |
| 🕸️ Graph Store |  |
Entity & relationship graph persistence |
| ⚡ Backend |  |
REST API, ingestion pipeline, reasoning engine |
| 🎨 Frontend |   |
Chat UI, 3D knowledge graph, timeline views |
| 📦 Deployment |  |
One-command local deployment |
| Rule | Status |
|---|---|
| No proprietary APIs | ✅ Zero external API calls |
| LLM < 4B parameters | ✅ Phi-4-mini = 3.8B |
| Fully local & air-gapped | ✅ Localhost only (127.0.0.1) |
| Continuous operation | ✅ Background file watcher + persistent store |
| Persistent memory | ✅ Qdrant + SQLite, survives restarts |
| Open-source model | ✅ MIT licensed |
# 1. Clone the repo
git clone https://github.com/Lacoste-INSAT/AI_Minds.git
cd AI_Minds
# 2. Start everything with Docker
docker-compose up --build
# 3. Open the UI
# → http://localhost:3000| Member | GitHub |
|---|---|
| Akram Zaabi | @AkramZaabi |
| Makki Aloulou | @MakkiAloulou |
| Rami Troudi | @Rami-Troudi |
| Yassine Kolsi | @yassinekolsi |
| Youssef Rekik | @youssefrekik1 |
Built with ❤️ during MSB AI Hackathon