___  _  _    _    ___    _
 |_ _|| \| |  /_\  | _ \  /_\
  | | | .` | / _ \ |   / / _ \
 |___||_|\_|/_/ \_\|_|_\/_/ \_\
-=It's Not A Random Acronym=-

Your own AI that lives on your desktop, controls your house, designs your parts, runs your browser, and prints your prototypes - all by voice.

Supported runtime: electron/main.js -> backend/server.py. The backend/core/ server refactor is archived reference code and is not the active app backend.

INARA is a modular AI agent platform built for real-world control. Not a chatbot. Not a wrapper around an API. A system with eyes, ears, hands, and opinions.

You can talk to it and get a response. Ask it to design a gear and it can create a 3D model, prepare it for printing, and send it to your printer. Tell it to dim the lights and it handles it. Ask it to look something up on Amazon and it opens a browser and searches for you.

What It Can Do

Feature	Description	Tech
🗣️ Real-Time Voice	Low-latency conversation with interrupt handling and wake word	Gemini Native Audio
🧊 Parametric CAD	Generate and iterate 3D models from natural language	build123d -> STL
🖨️ 3D Print Pipeline	Auto-slice and send to printers over your network	OrcaSlicer + Moonraker/OctoPrint
🖐️ Gesture Control	Minority Report-style window manipulation via hand tracking	MediaPipe
🌐 Web Agent	Autonomous browser - navigates, clicks, types, reads	Playwright + Chromium
🏠 Smart Home	Voice control for TP-Link Kasa + Home Assistant devices	python-kasa, HA REST API
👁️ Face Auth	Biometric login - local only, nothing leaves your machine	MediaPipe Face Landmarks
📁 Project Memory	Persistent context across sessions and conversations	File-based storage
⏰ Reminders	Voice-set reminders and recurring routines that fire on cue	File-backed store
🖥️ Desktop Control	Launch apps, take screenshots, search files, read clipboard	psutil, pyperclip
📷 Camera Vision	Describe scenes, detect presence, watch for conditions	Gemini Vision

🖐️ Gesture Control

INARA's Minority Report interface uses your webcam for hands-free window control:

Gesture	Action
✊ Closed Fist	Grab and drag a UI window
🤏 Pinch	Confirm / click
✋ Open Palm	Release

🔮 Coming Soon

Module	Description
📞 Phone Calls	Outbound/inbound call handling through voice
📅 Calendar Integration	Time-aware scheduling synced to external calendars
🔌 Matter / Thread	Next-gen smart home protocol support
🎤 Custom Wake Word	Train a personal wake phrase instead of a button press
🔊 Multi-Room Audio	Route voice and playback across rooms

---

🏗️ Architecture

Current supported runtime:

• Electron launches backend/server.py
• backend/server.py owns the active Socket.IO contract used by the React app
• backend/core/ is an archived experimental refactor and is not used by npm run dev

graph TB
    subgraph Frontend ["Frontend - Electron + React"]
        UI[React UI]
        THREE[Three.js 3D Viewer]
        GESTURE[MediaPipe Gestures]
        SOCKET_C[Socket.IO Client]
    end

    subgraph Backend ["Backend - Python + FastAPI"]
        SERVER[backend/server.py<br/>Canonical Socket.IO Runtime]
        LIVE[inara.py<br/>Gemini Live Voice Loop]
        CAD[cad_agent.py<br/>CAD Generation]
        WEB[web_agent.py<br/>Browser Automation]
        PRINTER[printer_agent.py<br/>3D Printing]
        KASA[kasa_agent.py<br/>Smart Home]
        AUTH[authenticator.py<br/>Face Auth]
        PROJECT[project_manager.py<br/>Project Context]
        SCHED[scheduler_agent.py<br/>Reminders & Routines]
        DESK[desktop_agent.py<br/>Desktop Control]
        VIS[vision_agent.py<br/>Camera Vision]
        DEV[device_agent.py<br/>Unified Devices]
    end

    UI --> SOCKET_C
    SOCKET_C <--> SERVER
    SERVER --> LIVE
    SERVER --> CAD
    SERVER --> WEB
    SERVER --> KASA
    SERVER --> PRINTER
    SERVER --> AUTH
    SERVER --> PROJECT
    SERVER --> SCHED
    SERVER --> DESK
    SERVER --> VIS
    SERVER --> DEV
    CAD -->|STL| THREE
    CAD -->|STL| PRINTER

Loading

The repository still contains an unfinished event-bus rewrite under backend/core/, but the supported runtime and active backend contract live in backend/server.py.

---

⚡ Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+

Setup

# Clone
git clone https://github.com/Herorif/inara.git && cd inara

# Python environment
python -m venv .venv

# Activate (pick your OS)
# Linux/macOS:
source .venv/bin/activate
# Windows:
.venv\Scripts\activate

# macOS only - required for PyAudio
# brew install portaudio

# Dependencies
pip install -r requirements.txt
playwright install chromium

# Frontend
npm install

# API keys
echo "GEMINI_API_KEY=your_key_here" > .env

🚀 Run

Single command:

npm run dev

Or split terminals (recommended - you'll want to see the logs):

# Terminal 1 - Backend
python backend/server.py

# Terminal 2 - Frontend
npm run dev

Manual backend shortcut:

npm run backend:dev

Make sure your venv is activated in any terminal that runs Python.

---

✅ First Flight Checklist

Once it's running, try these:

1. 🗣️ Voice - Say "Hello INARA". She should respond.
2. 👁️ Face Auth - Look at the camera. If enabled, the lock screen should unlock.
3. 🧊 CAD - Open the CAD window and say "Create a cube". Watch it generate.
4. 🌐 Web - Open the Browser window and say "Go to Google".
5. 🏠 Smart Home - If you have Kasa devices, say "Turn on the lights".
6. 🖨️ Print - Generate a model, then say "Print it".
7. ⏰ Reminders - Say "Remind me to stretch in 5 minutes". It fires and announces.
8. 🖥️ Desktop - Open the Desktop window. Click the lightning bolt to pull system stats.
9. 📷 Vision - Open the Vision window. Type "person at door" and add a watch condition.

---

⚙️ Configuration

Settings live in backend/settings.json (auto-created on first run).

Key	Type	Description
face_auth_enabled	bool	Require face recognition before interaction
tool_permissions.generate_cad	bool	Require confirmation before CAD generation
tool_permissions.run_web_agent	bool	Require confirmation before browser automation
tool_permissions.write_file	bool	Require confirmation before writing files to disk
tool_permissions.launch_app	bool	Require confirmation before launching applications
tool_permissions.control_device	bool	Require confirmation before toggling smart devices
tool_permissions.watch_for	bool	Require confirmation before starting camera watch
printers	array	Saved printer configurations
kasa_devices	array	Saved smart home devices

🔑 API Keys

Create a .env file in the project root:

GEMINI_API_KEY=your_gemini_key
ANTHROPIC_API_KEY=your_claude_key
# Optional — only needed if using Home Assistant integration
HA_URL=http://homeassistant.local:8123
HA_TOKEN=your_long_lived_access_token

• Gemini key -> Google AI Studio
• Claude key -> Anthropic Console

---

🔧 Hardware Setup

🖨️ 3D Printers

Supports Klipper/Moonraker, OctoPrint, and PrusaLink. Printers are auto-discovered via mDNS on your local network, or can be added manually by IP.

Requires OrcaSlicer installed for slicing. INARA auto-detects the installation path and selects the right profile based on your printer model.

🏠 Smart Home

TP-Link Kasa devices are discovered automatically on your network. Control lights (on/off, brightness, color), plugs, and switches - by voice or through the UI.

🔐 Face Authentication

1. Take a clear photo of your face.
2. Save it as reference.jpg in the backend/ directory.
3. Toggle with face_auth_enabled in settings.

All processing is local. Nothing is uploaded. Nothing is stored externally.

---

📂 Project Structure

inara/
├── backend/
│   ├── core/                  # Event bus, reminder store, config, tool registry
│   ├── llm/                   # LLM abstraction (Gemini, Claude, router)
│   ├── agents/                # Agent modules (CAD, web, printer, kasa, scheduler, desktop, vision, device)
│   ├── desktop/               # System monitor, app registry, screen capture
│   ├── vision/                # Vision loop (continuous camera watch conditions)
│   ├── devices/               # Home Assistant bridge
│   ├── voice/                 # Voice pipeline (STT, TTS, VAD, audio I/O)
│   ├── inara.py               # Voice integration (Gemini Live API)
│   ├── server.py              # Canonical Socket.IO runtime
│   ├── printer_agent.py       # Printer discovery & slicing engine
│   ├── kasa_agent.py          # Kasa device control engine
│   ├── cad_agent.py           # CAD generation engine
│   ├── authenticator.py       # Face auth engine
│   └── project_manager.py     # Project context management
├── src/                       # React frontend
│   ├── App.jsx                # Main application shell
│   ├── store/                 # Zustand slices (chat, cad, kasa, reminders, desktop, vision, devices…)
│   └── components/            # UI components (windows, visualizer, tools panel…)
├── electron/                  # Electron main process
│   └── main.js                # Window & IPC setup
├── tests/                     # Test suite
├── .env                       # API keys (create this)
├── requirements.txt           # Python dependencies
├── package.json               # Node.js dependencies
└── README.md

---

🔒 Security

Aspect	Implementation
API Keys	Stored in .env, excluded from version control
Face Data	Processed locally, never transmitted
Tool Confirmations	Write/CAD/Web actions can require user approval
Project Data	Everything stays on your machine

Never share your .env file or reference.jpg. These contain credentials and biometric data.

---

🤝 Contributing

1. Fork the repo
2. Create a feature branch: git checkout -b feature/your-feature
3. Commit your changes
4. Open a pull request with a clear description

---

📄 License

Copyright 2026 Harif

This project is licensed under the MIT License. See LICENSE for details.

---

Built by Herorif
I love you 3000