BozzoAI - Voice-Controlled Local AI Agent

A production-grade voice-controlled AI system that accepts microphone or uploaded audio, transcribes speech, detects intent, generates code with explanations, and visualizes the full pipeline in a modern React UI with premium animations.

✨ Features

• Landing Page: Premium hero section with BozzoAI branding, particle animations, and liquid glass effects
• Voice Control: Live microphone recording and audio file upload (.wav, .mp3, .webm)
• Speech-to-Text: Local speech-to-text using Whisper (faster-whisper preferred)
• AI Code Generation: Local intent detection and multi-language code generation with Ollama (Mistral/Llama)
• Smart Explanations: Generated code includes automatic explanations shown in chat
• Safe Execution: Output-only filesystem sandbox with path traversal prevention
• UI Features:
  • Premium landing page with sparkle animations
  • Liquid glass button effects
  • Blue (#3B82F6) accent color scheme
  • Smooth fade transitions and animations
  • Back navigation from chat to landing page
  • Real-time pipeline logs and status updates
  • Responsive 3-panel layout

🎯 Supported Intents

• write_code - Generate code in any language (Python, JavaScript, HTML, CSS, etc.)
• create_file - Create files with specified content
• general_chat - General conversational responses
• summarize_text - Text summarization

🏗️ Architecture

Audio Input → Speech-to-Text → Intent Detection → Code Generation → UI Output
                                    ↓
                          LLM-powered explanation

📁 Project Structure

project/
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── BozzoAILanding.tsx    (Premium landing page)
│   │   │   ├── ChatPanel.tsx
│   │   │   ├── InputCard.tsx
│   │   │   ├── Sidebar.tsx
│   │   │   ├── SystemStatus.tsx
│   │   │   └── ui/
│   │   │       ├── sparkles.tsx      (Particle animations)
│   │   │       └── ...
│   │   └── api/
│   └── package.json
│
├── backend/
│   ├── main.py
│   ├── intent.py                     (Enhanced language detection)
│   ├── tools.py                      (Code generation with explanations)
│   ├── stt.py
│   ├── security.py
│   ├── config.py
│   └── requirements.txt
│
├── output/
│   └── .gitkeep
│
└── README.md

🌐 API Endpoints

POST /process-audio

• Content type: multipart/form-data
• Field: file (audio blob)

Response:

{
  "transcription": "user's speech",
  "intent": "write_code",
  "action": "file_created",
  "output": "[OK] Generated Python code\n[FILE] Saved to: /output/spell_checker.py\n[TASK] Spell Checker\n[INFO] Creates a spell checker using NLTK...",
  "logs": [...]
}

POST /process-text

• Content type: multipart/form-data
• Field: text

Same response format as /process-audio

🔒 Security Features

• Path Traversal Prevention: All file paths validated and restricted to /output directory
• Filename Sanitization: Filenames converted to snake_case and validated
• Safe Code Generation: LLM-generated code is cleaned before execution
• Sandbox Execution: Output-only filesystem access model
• Input Validation: All user inputs validated and sanitized
• Request Verification: Audio and text inputs validated before processing

🚀 Quick Start

1. Clone and Configure

git clone <repo>
cd "vc controlled ai agent"
Copy-Item .env.example .env

2. Backend Setup

cd backend
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt

3. Frontend Setup

cd frontend
npm install
npm run dev

Frontend: http://localhost:5173

4. Start Backend

cd ..
uvicorn backend.main:app --host 0.0.0.0 --port 8000

5. Ollama Setup

# Install Ollama from https://ollama.ai
ollama pull mistral
# Start Ollama service
ollama serve

🤖 Model Configuration

STT Models

• Primary: faster-whisper (recommended for speed)
• Fallback: openai-whisper

LLM Models

• Recommended: Ollama local models
  • mistral (7B - balanced)
  • llama2 (7B - fast)
  • neural-chat (7B - optimized)

🎨 UI/UX Design

• Color Scheme: Dark theme (#050505 base, #3B82F6 blue accents)
• Effects:
  • Sparkle particle animations
  • Liquid glass button effects with backdrop blur
  • Smooth fade and slide transitions
  • Responsive animations
• Landing Page: Premium hero with centered BozzoAI title, particle effects, and CTA button

🔍 Security Check Results

To run security checks:

# Python security
pip install bandit
bandit -r backend/

# Node security
cd frontend
npm audit

📝 Recent Updates (v2.0)

• ✅ Premium landing page with BozzoAI branding
• ✅ Particle animation system for enhanced visuals
• ✅ Liquid glass button effects
• ✅ Enhanced code generation with automatic explanations
• ✅ Back navigation button for seamless UX
• ✅ Improved color scheme consistency (blue accents)
• ✅ HTML/CSS language detection for code generation

🧪 Testing

Run tests:

# Backend
cd backend
pytest

# Frontend
cd frontend
npm test

🎯 Example Usage

Input: "Generate a Python spell checker"

Process:

1. Audio transcribed: "Generate a Python spell checker"
2. Intent detected: write_code
3. Code generated in Python
4. Automatic explanation provided
5. File saved to /output/
6. Results displayed in UI with explanation

📦 Dependencies

Backend

• FastAPI
• Faster-Whisper (STT)
• Ollama (LLM)
• Uvicorn

Frontend

• React 18.3.1
• Vite 5.4
• TailwindCSS 3.4
• Framer Motion
• Lucide React Icons

🛣️ Roadmap

• User authentication
• Rate limiting
• Advanced code analysis
• Multi-language support
• Real-time collaboration
• Code execution environment

📄 License

MIT

👤 Author

Niraj Pachte

---

Last Updated: April 14, 2026