GemDesk

Multi-file AI assistant powered by Gemini 3 Flash with 1M token context

GemDesk is a desktop application that enables deep analysis across multiple files simultaneously using Google's Gemini 3 Flash model. Upload up to 50 files, ask questions, generate reports, find contradictions, and visualize data - all with intelligent context caching and adaptive thinking modes.

✨ Key Features

🗂️ Multi-File Analysis

• 1M token context window - Analyze massive amounts of data simultaneously
• 50 file limit - Upload documents, spreadsheets, images, videos, and more
• 100+ file formats supported - PDFs, Office docs, images, videos, code files
• Smart conversion - DOCX → PDF with images, XLSX → CSV, and more
• Context caching - Files cached separately for faster responses

🤖 Advanced AI Capabilities

• Adaptive thinking modes - Adjust reasoning depth (minimal/low/medium/high)
• Specialized analysis presets via slash commands:
  • /report - Executive summaries and cohesive reports
  • /synthesize - Pattern identification and novel insights
  • /error-check - Contradiction and inconsistency detection
• Automatic chart generation - Just ask! "Plot sales over time" generates visualizations
• Function calling - Gemini intelligently uses tools when appropriate

📊 Visualization & Export

• Smart charting - Bar, line, pie, and scatter plots generated on demand
• Export conversations - Save chats as formatted PDFs
• Export charts - Save visualizations as high-quality PNGs
• Markdown rendering - Code highlighting, tables, and formatting

🎨 User Experience

• Dark/light themes - Toggle between modes
• Organized file shelf - Auto-categorized by type with collapsible folders
• Real-time token metering - Track context usage
• URL scraping - Add web pages and direct file downloads
• Thumbnail previews - Visual preview for images, PDFs, and videos

---

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• Gemini API key (Get one free)

Installation

1. Clone the repository

git clone https://github.com/openconstruct/gemdesk.git
cd gemdesk

2. Install dependencies

pip install -r requirements.txt

3. Set up API key

Create a .env file in the project root:

GEMINI_API_KEY=your_api_key_here

4. Run

python gem.py

The app will open in your default browser!

---

📖 Usage Guide

Basic Workflow

1. Upload files - Click "Add Files" or paste URLs
2. Ask questions - Type naturally or use slash commands
3. Get insights - Receive analysis with citations
4. Visualize data - Ask for charts when needed
5. Export results - Save conversations and charts

Slash Commands

Analysis Presets:

• /report - Generate executive summary with key findings
• /synthesize - Identify patterns and generate novel insights
• /error-check - Find contradictions across sources

Example:

/report focus on Q4 financial performance

Help:

• /help - Show all available commands

Charting

No buttons needed! Just ask naturally:

• "Plot the sales data over time"
• "Show me customer acquisition vs revenue"
• "Create a pie chart of market share"

Gemini will automatically analyze your files and generate appropriate visualizations.

Thinking Modes

Adjust reasoning depth via the dropdown:

• Minimal - Fast, basic reasoning
• Low - Light thinking
• Medium - Balanced (default for reports)
• High - Deep reasoning (default for synthesis/error-checking)

---

📁 Supported File Types

Documents

• PDF - Native support
• DOCX - Converted to PDF (preserves inline images)
• TXT, MD, RTF, HTML - Plain text formats
• ODT - OpenDocument Text

Spreadsheets

• XLSX - Converted to CSV (all sheets)
• ODS - OpenDocument Spreadsheet
• CSV - Native support

Presentations

• PPTX - Native support (Gemini 3)
• ODP - OpenDocument Presentation (converted to text)

Media

• Images - JPG, PNG, GIF, WEBP, HEIC, SVG
• Videos - MP4, MOV, AVI, WEBM, FLV
• Audio - MP3, WAV, FLAC, AAC, OGG

Code

80+ programming languages supported including:

• Python, JavaScript, TypeScript, Java, C/C++, Go, Rust
• HTML, CSS, PHP, Ruby, Swift, Kotlin, Scala
• And many more...

---

🏗️ Architecture

Context Caching Strategy

Files are cached separately from conversation history for optimal performance:

• All uploaded files → Single cache (1 hour TTL)
• Conversation history → Sent as text only
• Cache updates automatically when files added/removed

Benefits:

• Faster responses (files not resent with every message)
• Lower token costs
• Efficient multi-turn conversations

File Processing Pipeline

User Upload
    ↓
Type Detection
    ↓
Conversion (if needed)
    - DOCX → PDF (with images)
    - XLSX → CSV (all sheets)
    - ODT/ODP → Text
    ↓
Upload to Gemini
    ↓
Token Counting
    ↓
Add to Cache

Tool Integration

Gemini uses function calling to generate charts:

1. User requests visualization
2. Gemini analyzes data in files
3. Calls generate_chart tool with parameters
4. Chart rendered with matplotlib
5. Displayed in popup dialog

---

🛠️ Configuration

Environment Variables

• GEMINI_API_KEY - Your Gemini API key (required)

Model Configuration

Edit gem.py to customize:

MODEL_ID = "gemini-3-flash-preview"  # Model to use
MAX_CONTEXT_TOKENS = 1000000         # Context window size
MAX_FILES = 50                        # File upload limit

---

📊 Performance

Token Usage

• 1M token context - Analyze extensive documents
• Context meter - Real-time usage tracking
• Smart caching - Reduce redundant processing

File Limits

• Max files: 50 (configurable)
• Max file size: Limited by Gemini API
• Max context: 1M tokens total

---

🤝 Contributing

Contributions welcome! Areas for improvement:

• Additional file format support
• More chart types
• Enhanced error handling
• Performance optimizations
• UI/UX improvements

Development Setup

# Clone repository
git clone https://github.com/openconstruct/gemdesk.git
cd gemdesk

# Install dev dependencies
pip install -r requirements.txt

# Run with debug logging
python gem.py

---

🐛 Troubleshooting

Common Issues

"GEMINI_API_KEY not found"

• Create .env file with your API key
• Ensure file is in project root directory

"Module not found" errors

• Run: pip install --upgrade -r requirements.txt

File upload fails

• Check file size (must be under API limits)
• Verify file format is supported
• Check internet connection

Charts not generating

• Ensure matplotlib is installed: pip install matplotlib
• Verify chart request is clear (e.g., "plot X over time")

Context cache errors

• Files are automatically re-uploaded if cache expires
• Try removing and re-adding problematic files

---

📜 License

MIT License - See LICENSE for details

---

🙏 Acknowledgments

• Google Gemini - Powerful multimodal AI capabilities
• Flet - Beautiful Python UI framework
• matplotlib - Chart generation
• python-docx & reportlab - Document processing

---

🔗 Links

• Gemini API Documentation
• Flet Documentation
• Report Issues

---

📈 Roadmap

• Web deployment support
• Drag & drop file upload
• Google Search grounding integration
• Persistent chat history
• Multi-tab conversations
• Voice input support
• Collaborative features

---

Built for the Gemini 3 Hackathon 2026

Showcasing advanced multimodal analysis with intelligent context management and adaptive reasoning