PiRhoAI

📊 PiRhoAI - Privacy Focused AI Business Analytics Assistant

Transform your business data into actionable insights with AI-powered natural language Q&A! This is a minimum viable product for MSME analytics using Streamlit, LangChain, and Retrieval-Augmented Generation (RAG).

Live Demo: https://pirhoai.streamlit.app/

📸 Screenshots

Note: Add actual screenshots from https://pirhoai.streamlit.app/ to the screenshots/ directory

🚀 Features

Core Capabilities

• 📁 Data Integration: Upload local CSV/Excel files or import from Google Drive sharing links
• 🤖 AI-Powered Q&A: Natural language queries supported (English & हिंदी)
• 🔍 Smart Analysis: Retrieval-Augmented Generation for accurate, context-aware responses
• 📊 Data Preview: Real-time visual preview with auto-loaded sample business data
• ⚙️ Multi-Model Support: Switch between local models (Phi-4, Phi-3) or cloud API (Grok-4) for optimal performance
• 🔄 Auto Setup: Sample data loads automatically on first run for immediate exploration

🚀 Revolutionary Edge AI Technology

• 📱 Cloud-First with Local Fallback: API-based AI (Grok-4) for fast responses, local models when offline or hardware allows
• 🧠 Tiny LLMs for Full Analytics: Complete business intelligence using compact models (3B parameters) with CPU compatibility
• 🔒 Privacy & Performance Balance: API mode for sensitive data (no local storage), local models for offline privacy
• 🤖 AI Assistant with Agents: Intelligent workflows adapting to your business patterns
• ⚡ Smart Model Selection: Automatic CPU/GPU detection with appropriate quantization (8-bit GPU, float32 CPU)
• 💰 Flexible Costs: Free tier API usage, or zero-cost local processing

📋 Requirements Validation

Based on comprehensive market research and technical analysis:

🎯 Product Mission

Democratizing data analytics for MSMEs by enabling natural language queries without technical expertise.

🤖 Unique Selling Points

• 📱 Your Phone is the GPU: Transform your smartphone into a powerful analytics engine - no expensive hardware needed
• 🧠 Tiny LLMs, Massive Intelligence: Complete business analytics with efficient AI models (3B-7B parameters) that run entirely offline
• 🔐 Ultimate Privacy: 100% private analytics - your business data never leaves your device, fully compliant with privacy regulations
• 🚀 AI Agent Workflows: Intelligent custom assistants that adapt to your specific business needs and automate routine analytics tasks

💡 Value Proposition

• Data Sovereignty: Your business data stays completely private and secure
• Zero Infrastructure Costs: Leverage existing phone/desktop computing power
• Offline Functionality: Work without internet for maximum productivity
• Custom AI Workflows: AI agents learn your business patterns and provide tailored insights

🌍 Market Fit (Especially in Developing Markets)

• MSMEs drive 40-60% GDP in emerging economies (India, Brazil, Indonesia)
• 90% SMEs lag in analytics adoption; this bridges the gap
• Multi-language support for regional diversity
• Affordable free-tier storage integration (Google Drive)

🧠 Recommended LLM Stack

| Model | Parameters | Best For | Key Strengths | |——-|————|———-|—————| | Phi-4-mini | 3.8B | Primary | Complex reasoning, multilingual, edge deployments | | Phi-3 | 3.8B | NLP tasks | Coding/reasoning with 128K context | | IBM Granite 4.0 Tiny | 3B | Compliance | Hybrid efficiency, enterprise security |

⚖️ Competitive Analysis

• vs. Snowflake: Hours setup vs. weeks, $20-100/month vs. $200-500+ on 100GB
• vs. Zoho Analytics: LLM-first conversational AI, storage-first integration, disruptive pricing

🛠️ Setup

Local Development

1. Create virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

2. Install dependencies:

   pip install -r requirements.txt          # Full ML features
   # OR
   pip install -r requirements_basic.txt    # Basic functionality only
   

3. Set environment variables (optional, for API models):

   export XAI_API_KEY="your-xai-api-key-here"  # For Grok-4 API access (free tier available)
   # Local models work without API key, startup may take longer
   

4. Run the application:

   streamlit run main.py  # Default configuration with API-based AI for fast startup
   

📊 Data Preparation

The app supports:

• CSV Files: Standard comma-separated values
• Excel Files: .xlsx or .xls formats
• Google Drive: Paste sharing URL for cloud data import

🤖 AI Features (Optional)

For full AI capabilities, install ML dependencies. The app provides graceful fallbacks:

• If ML libs unavailable, data upload and preview still work
• AI responses show helpful messages about installation
• Basic pandas processing ensures core functionality

🚀 Deployment

Streamlit Cloud (Recommended)

1. Fork this repository
2. Connect to Streamlit Cloud
3. Set main module as main.py with Python 3.13

Vercel (Alternative)

# Requires vercel.json configuration (included)
vercel --prod

Docker (Advanced)

# Coming in future updates
# Supports both basic and ML configurations

📚 Key Files

• main.py: Entry point with UI configuration
• src/ui.py: Streamlit interface with tabs and styling
• src/core/rag.py: Retrieval-augmented generation logic
• src/core/llm.py: LLM integration with multiple model support
• src/data/connector.py: Google Drive and local file connectors
• requirements.txt: Full dependencies with ML packages
• requirements_basic.txt: Minimal setup for demo purposes

🎨 UI/UX Features

• Professional Design: Gradient headers and brown-themed color scheme
• Responsive Layout: Tabs for Analytics vs. Product Briefing
• Interactive Elements: Comparison tables with detailed specs
• Customer Journey Mapping: 5-step onboarding visualization
• Accessibility: Multi-language support and clear navigation

🔄 Customer Journey

1. Discovery: Find the solution via web/app stores
2. Data Connection: Upload files or connect cloud storage
3. Exploration: Ask natural questions in preferred language
4. Decision Making: Get AI insights to drive business actions
5. Growth: Scale with advanced features as business expands

📈 Performance Benchmarks

• Setup Time: Under 5 minutes
• Data Processing: Instantaneous for <1GB datasets
• AI Response Time:
  • API Mode (Grok-4): <3 seconds (requires internet)
  • Local Mode (Phi-3.5): <30 seconds first load, <5 seconds cached (CPU-compatible)
• Memory Usage: ~2GB max (local models), <1GB API mode
• Platform Compatibility: macOS, Windows, Linux with automatic CPU/GPU detection

🛡️ Privacy & Security

• Local data processing (no uploads to cloud)
• End-to-end encryption for file operations
• No user data stored beyond session
• Compliant with global privacy standards

🔮 Future Roadmap

• Advanced visualization charts and dashboards
• Voice input for hands-free operation
• Multi-file analysis and cross-data correlations
• Automated report generation and scheduling
• Integration with accounting software (QuickBooks, etc.)
• Predictive analytics for trend forecasting
• Team collaboration features

📞 Support & Community

• Documentation: Extensive in-app help and tooltips
• Community: Forums for user-shared templates
• Support: Email assistance for technical issues
• Learning: Video tutorials and sample datasets

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make improvements following the existing patterns
4. Submit a pull request

📄 License

This project is open-source under the MIT License. Feel free to use, modify, and distribute.

---

Product Vision: “Your phone becomes your business analyst” - Private, powerful AI analytics using your device’s processing power with Tiny LLMs for complete data intelligence. MSMEs get enterprise-grade insights with maximum privacy and zero infrastructure costs through intelligent AI agents and custom workflows.

Auto-updated from PiRhoAI development - https://pirhoai.streamlit.app/

This site is open source. Improve this page.