🚀 v0.1.0 — Alpha
Your Privacy-First RAG System
100% local RAG with Knowledge Graph. Your documents never leave your machine. Hybrid search combines Vector embeddings, BM25, and Knowledge Graph for precise answers.
Why VaultMind?
Built for developers who care about privacy, performance, and control over their data.
🔒
100% Local
Data never leaves your machine. Zero cloud dependency. Works offline.
🧠
Hybrid RAG
Vector + BM25 + Knowledge Graph with RRF fusion and graph boosting.
⚡
GPU Accelerated
CUDA, Metal (Apple Silicon), ROCm via llama.cpp.
📝
Obsidian-native
wiki-links, YAML frontmatter, auto-linking concepts.
🔌
OpenAI-compatible API
/v1/chat/completions for Cursor, Continue, etc.
📦
Workspace Isolation
Separate knowledge bases per project.
🏗️ Architecture
Modular design for flexibility and performance.
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
│ ingestion │ → │ indexing │ → │ inference │
│ .py │ │ .py │ │ .py │
└─────────────┘ └──────────────┘ └─────────────┘
│ │ │
▼ ▼ ▼
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
│ Document │ │ FTS5 + │ │ llama.cpp │
│ Parser │ │ Vectors + KG │ │ GPU/CPU │
└─────────────┘ └──────────────┘ └─────────────┘
│
▼
┌─────────────────────┐
│ FastAPI Server │
│ /v1/chat/completions│
│ REST API + Web UI │
└─────────────────────┘