AURA — Adaptive Unified Retrieval Assistant

AURA: A Voice-Powered Personal Memory & Knowledge Retrieval System

"Your second brain — remember everything, retrieve anything."

---

Overview

AURA (Adaptive Unified Retrieval Assistant) is an AI-powered desktop assistant designed to act as a personal memory system.

Unlike traditional chatbots that forget conversations, AURA captures, organizes, and retrieves information from your personal knowledge base using natural language, voice commands, and AI-powered semantic understanding.

AURA combines:

• AI Memory Management
• Voice Interaction
• Knowledge Graphs
• Local Database Storage
• Intelligent Retrieval
• Interactive HUD Interface

into a single futuristic desktop application inspired by J.A.R.V.I.S from Iron Man.

---

Key Features

Voice-Powered Memory Capture

Speak naturally.

AURA can:

• Listen to voice input
• Convert speech to text
• Understand context
• Store memories automatically

Examples:

I have my Data Engineering internship presentation next Monday.

Remember that I need to submit my project report before Friday.

My Groq API key issue was fixed today.

---

Intelligent Memory Retrieval

Ask questions naturally.

Examples:

Tell me about my internship.

What was I working on last week?

Show memories related to SQL.

What project did I discuss yesterday?

AURA searches stored memories and generates contextual answers using AI.

---

Knowledge Graph Generation

Every memory is analyzed and converted into structured relationships.

Example:

You worked on API refactoring during internship.

Extracted as:

You
 └── worked_on
      └── API Refactoring

These relationships are visualized in the Cognitive Personality Graph.

---

AI Memory Tagging

Each memory is automatically classified:

{
  "topic": "Internship",
  "category": "Work",
  "emotion": "Focused",
  "importance": 8,
  "tags": [
    "internship",
    "api",
    "backend"
  ]
}

---

Cognitive Personality Graph

AURA continuously builds a knowledge graph from your memories.

Tracks:

• Projects
• People
• Ideas
• Technologies
• Goals
• Relationships

Interactive graph supports:

• Dragging nodes
• Real-time layout physics
• Dynamic relationship visualization

---

Conversation Memory

AURA remembers recent conversations.

Example:

User:
Tell me about my internship.

AURA:
You worked on API validation and Flask integration.

User:
Tell me more.

AURA:
You also discussed scheduling, debugging, and deployment.

---

AI Assistant Personality

AURA uses a loyal assistant personality.

Responses are designed to feel like:

Systems online, Boss.

I found 5 relevant memories related to your internship.

Inspired by:

• J.A.R.V.I.S.
• Friday
• Personal Digital Twin systems

---

Text-to-Speech Responses

AURA speaks responses aloud using:

edge-tts

Features:

• Natural sounding voice
• Real-time playback
• Mute support
• Audio visualization

---

Real-Time Audio Monitoring

AURA monitors:

Microphone Input

Displays:

MIC IN

with live amplitude meter.

System Output

Displays:

SYS OUT

with live speech playback levels.

---

Architecture

User
 │
 ▼
Voice Input
 │
 ▼
Speech Recognition
 │
 ▼
Groq Whisper STT
 │
 ▼
AURA Brain
 │
 ▼
Groq LLM
 │
 ├── Memory Tagging
 │
 ├── Relationship Extraction
 │
 └── Query Resolution
 │
 ▼
SQLite Database
 │
 ├── Memories
 ├── Tags
 ├── Retrieval Logs
 └── Knowledge Triples
 │
 ▼
PyQt6 HUD Interface

---

Technology Stack

Backend

• Python 3.12
• Flask
• Flask-CORS
• SQLite

---

AI Layer

• Groq API
• Llama 3.3 70B Versatile
• Whisper Large

---

Voice Layer

• SpeechRecognition
• PyAudio
• edge-tts
• pygame

---

Interface

• PyQt6
• QPainter
• Custom HUD Components

---

Visualization

• Force Directed Graph
• Real-Time Physics Simulation

---

Database Design

Memories

Stores:

memory_id
content
timestamp
topic
category
emotion
importance

---

Tags

Stores:

tag_id
tag_name

---

Memory Tags

Many-to-many relationship:

memory_id
tag_id

---

Retrieval Logs

Stores:

query
timestamp
results

Used for analytics.

---

Knowledge Triples

Stores:

source
relation
target
memory_id
timestamp

Example:

You
 └── learned
      └── Flask

---

Project Structure

AURA/
│
├── assets/
│   ├── logo.png
│   └── logo.ico
│
├── ai_brain.py
├── config.py
├── database.py
├── main.py
├── server.py
├── seed_memories.py
├── ui.py
├── voice.py
│
├── requirements.txt
├── .env.example
├── README.md
│
└── aura_memory.db

---

Installation

Clone repository:

git clone https://github.com/yourusername/AURA.git
cd AURA

Create virtual environment:

python -m venv venv

Activate:

Windows

venv\Scripts\activate

Linux

source venv/bin/activate

Install dependencies:

pip install -r requirements.txt

---

Environment Variables

Create:

.env

Example:

GROQ_API_KEY=YOUR_API_KEY

AURA_INPUT_DEVICE=9
AURA_SAMPLE_RATE=48000

AURA_SILENCE_THRESHOLD=0.0015
AURA_WAKE_CHUNK_SECONDS=1.4
AURA_WAKE_COOLDOWN_SECONDS=1.2

---

Running AURA

python main.py

---

Build Desktop Application

pyinstaller --onedir --windowed --name AURA main.py

With icon:

pyinstaller --onedir --windowed --icon=assets/logo.ico --name AURA main.py

---

Use Cases

Students

Remember:

• Assignments
• Notes
• Exams
• Projects

---

Developers

Track:

• Bugs
• Decisions
• Architecture
• Research

---

Researchers

Maintain:

• Knowledge Base
• Citations
• Findings
• Experiments

---

Personal Productivity

Store:

• Goals
• Ideas
• Meetings
• Reflections

---

Why AURA?

Most assistants answer questions.

AURA remembers.

Most note-taking apps store information.

AURA understands relationships.

Most chatbots forget.

AURA builds a living knowledge graph of your life, projects, and ideas.

---

Author

Lakshya Bhandari

B.Tech CSE (Data Science) SKIT, Jaipur

AURA — Adaptive Unified Retrieval Assistant

"Your memories. Organized. Searchable. Alive." 🚀