JARVIS is an AI-powered personal assistant designed to bridge the gap between intelligence and execution. Unlike conventional assistants that stop at conversation, JARVIS extends its capabilities into your local operating system - opening applications, managing system controls, navigating the web, and responding visually to the world around it.

At its core, JARVIS is a modular, locally-aware AI system that combines cloud-level reasoning with on-device control. It listens, understands, decides, and does things, bringing the concept of a truly functional AI assistant closer to reality.

---

---

---

Tech Stack

Category	Technologies
Programming Languages
AI / LLM
Vision & Speech
Memory / Vector DB
Web Search
Backend Framework
Frontend Framework
Local Automation
Tools
Deployment / Runtime

---

🏗️ System Architecture

JARVIS operates on a modular architecture designed for speed, privacy, and extensibility.
The system is divided into three core components: the AI Brain Layer, Vision Module, and Local Agent.

graph TD
    User([👤 User Input]) --> Brain
    Brain[🧠 AI Brain Layer] -->|Text Query| Chat[💬 Response]
    Brain -->|Visual Data| Vision[👁️ Vision Module]
    Brain -->|System Command| Agent[⚙️ Local Agent]
    
    Vision --Description--> Brain
    Agent --Action--> OS[💻 System / OS]

Loading

🧩 Core Modules

JARVIS is built on three primary components that work together to understand, analyze, and execute user commands.

---

🧠 1. AI Brain Layer

The intelligence core that thinks, decides, and orchestrates actions.

Powered by: LangChain (Groq, OpenAI, Gemini, Mistral)

The AI Brain acts as the central decision engine. It interprets user input, understands intent, and determines the appropriate course of action.

Responsibilities:

1. Natural Language Understanding: Interprets complex and conversational queries.
2. Decision Engine: Chooses whether to respond, analyze visuals, or execute a command.
3. Task Routing: Directs requests to the Vision Module or Local Agent.

---

👁️ 2. Vision Module

The perception system that allows JARVIS to “see” and understand images.

Powered by: Multimodal LLMs (Google Gemini Vision / OpenAI GPT-4V)

This module converts visual input into meaningful descriptions and insights that the AI Brain can process and reason about.

Capabilities:

1. Image-to-Text Conversion: Generates accurate image descriptions.
2. Visual Question Answering (VQA): Answers questions based on image content.
3. Scene Analysis: Identifies objects, context, and relationships.

---

⚙️ 3. Local Agent

The action layer that interacts directly with the operating system.

Runs on: Local Host (Low latency & secure)

The Local Agent transforms decisions into real system actions, enabling JARVIS to control applications and system settings.

Functions:

1. Application Control: Open, close, and manage programs.
2. System Adjustments: Modify volume, brightness, and power settings.
3. Web Automation: Launch websites and perform browser tasks.

---

📂 Project Structure

JARVIS/
├── backend/                        # Main Server & Logic
│   ├── brain/                      # AI Intelligence Modules
│   │   ├── app.py                  # Core Brain application logic
│   │   ├── database.py             # Database operations
│   │   ├── llm_services.py         # Connects to LLM (LangChain)
│   │   ├── memory_manager.py       # Handles chat history & context
│   │   ├── memory_services.py      # Memory storage and retrieval
│   │   └── web_search.py           # Web search integration
│   ├── routers/                    # API Endpoints
│   ├── chroma_db/                  # Vector Database for Long-term memory
│   ├── main.py                     # FastAPI Entry Point (Run this to start)
│   ├── auth.py                     # User Authentication & Security
│   ├── requirements.txt            # Python Dependencies
│   └── users.db                    # User database
│
├── frontend/                       # User Interface (React + Vite)
│   ├── src/
│   │   ├── components/
│   │   │   ├── ChatInterface.tsx   # Main chat window
│   │   │   ├── Login.tsx           # Authentication screen
│   │   │   └── Sidebar.tsx         # Chat history navigation
│   │   ├── api.ts                  # Connection to Backend
│   │   ├── App.tsx                 # Main Application Layout
│   │   └── main.tsx                # Frontend Entry Point
│   ├── package.json
│   └── vite.config.ts
│
├── local_agent/                    # OS Control Source Code
│   ├── agent.py                    # Websocket client for OS commands
│   └── os_controller.py            # Logic to open apps/control system
│
├── Jarvis_Agent.exe                # Compiled Local Agent executable
├── load_test_report.md             # Load test report
└── README.md                       # Project Documentation

🚀 How to Use JARVIS

Jarvis can be accessed by the link. Follow the steps below to access the Local Agent for local device task on Windows.

1️⃣ Download the Agent file

Download the Jarvis_Agent.exe file.

2️⃣ Run the Jarvis_Agent.exe file

This can done by clicking on the downloaded exe file or using command terminal

.\Jarvis_Agent.exe  #or the location of the downloaded file

This should open a pop-up to enter your login credentials. Use the same login credentials which you used to log in on the website.

---

🚀 How to Run JARVIS locally

Follow the steps below to run the project locally and use all its features.

1️⃣ Clone the Repository

git clone https://github.com/your-username/JARVIS.git
cd JARVIS

2️⃣ Move to backend directory

cd backend

3️⃣ Create & Activate Virtual Environment

python -m venv venv
venv\Scripts\activate

4️⃣ Install Dependencies

pip install -r requirements.txt

5️⃣ Set Environment Variables

Create a .env file in the root directory and add:

GROQ_API_KEY=your_groq_api_key
SERPER_API_KEY=your_serper_api_key

6️⃣ Run the Backend Server

python main.py

7️⃣ Start frontend

Start a new terminal and move to frontend

cd frontend

Install dependencies and run

npm install
npm run dev

---

🤝 Contributors

• Avanish Salunke
• Pratyush Rao
• Purvasha Singh
• Kavish Nasta