Aurora-Vault

Developed by Md Tareq Shah Alam

Aurora-Vault is a lightweight, production-ready RAG (Retrieval-Augmented Generation) engine that enables instant semantic search without any setup or preprocessing.

Overview

Aurora-Vault is designed to remove the complexity of building RAG systems.

Unlike traditional pipelines that require:

• Data preprocessing
• Embedding generation
• Index building

Aurora-Vault provides a prebuilt knowledge index, allowing developers to use RAG instantly.

Key Idea

Install → Load → Use

No setup. No build. No waiting.

Features

• Instant setup (no index building)
• Lightweight package (no large files inside)
• Automatic index download on first run
• Fast semantic search
• Precomputed embeddings for efficiency
• Plug-and-play API
• Production-ready architecture

Installation

pip install aurora-vault

Quick Start

1. Import

import aurora_vault

2. Initialize

vault = aurora_vault.load()

First-Time Setup (Automatic)

On the first run, Aurora-Vault will:

• Create a rag/ directory in your project
• Download a prebuilt index
• Initialize the retrieval system

You will see:

Aurora Vault Setup
Downloading core index (~1GB)...

Subsequent Runs

• No download
• No setup
• Instant loading

Data Storage

Aurora-Vault creates:

your_project/
 └── rag/
      └── index.pkl

• index.pkl → prebuilt vector index
• Stored locally for reuse
• No repeated downloads

Custom Storage Path

You can control where data is stored:

vault = aurora_vault.load(path="D:/my_data")

How It Works

Aurora-Vault follows a simplified RAG pipeline:

User Query
   ↓
Query Embedding (on demand)
   ↓
Vector Similarity Search
   ↓
Retrieve Relevant Context
   ↓
(Optional) LLM Processing
   ↓
Final Output

Why Aurora-Vault

Traditional RAG systems are:

• Slow to initialize
• Complex to manage
• Resource heavy

Aurora-Vault solves this by:

• Removing build-time overhead
• Using precomputed embeddings
• Providing instant usability

Use Cases

• AI Chatbots
• Knowledge Retrieval Systems
• SaaS AI Platforms
• Document Search
• Internal Assistants
• Automation Systems

Important Notes

• First run requires internet (for index download)
• Download happens only once
• Do not delete rag/index.pkl unless reset is needed

Roadmap

• FAISS-based ultra-fast search
• Advanced embedding strategies
• Incremental updates
• Multi-dataset support
• Monitoring tools

Contributing

Contributions are welcome.

Aurora-Vault aims to make RAG systems simple, fast, and accessible.

License

MIT License