👁️ Glimpse

Glimpse is a lightweight CLI tool that periodically captures your screen and uses Vision Language Models (VLM) to understand what you're doing. It stores searchable context that helps you recall past activities.

• Features
• Quick Start
• Commands
• Configuration
• Privacy
• Development
• Roadmap
• Contributing
• License
• Acknowledgments

# Clone the repository
git clone https://github.com/MonadWorks/glimpse.git
cd glimpse

# Install with pip
pip install -e .

# Or install with optional dependencies
pip install -e ".[full]"

Set your OpenAI API key:

glimpse config set vlm.api_key sk-your-api-key-here

Or use an environment variable:

export OPENAI_API_KEY=sk-your-api-key-here

# Capture and analyze your screen
glimpse capture

# Search your history
glimpse search "python project"

# See today's activity summary
glimpse today

# View all configuration
glimpse config show

Capture a screenshot and analyze it with VLM.

# Basic capture
glimpse capture

# Capture without VLM analysis
glimpse capture --no-analyze

# Save screenshot to disk
glimpse capture --save

# Capture specific monitor
glimpse capture -m 2

Search your captured context history.

# Simple search
glimpse search "meeting notes"

# FTS5 syntax supported
glimpse search "python AND machine learning"

# Filter by date
glimpse search "code" --from 2025-01-01 --to 2025-01-31

# Limit results
glimpse search "project" -n 5

Show a summary of today's activities.

# Brief summary
glimpse today

# Detailed timeline
glimpse today --detail

Manage configuration.

# Show all config
glimpse config show

# Set a value
glimpse config set vlm.api_key sk-xxx
glimpse config set vlm.model gpt-4o
glimpse config set capture.interval 60

# Get a value
glimpse config get vlm.model

Show database statistics.

glimpse stats

List available monitors for capture.

glimpse monitors

Configuration is stored in ~/.glimpse/config.yaml:

# Capture settings
capture:
  interval: 30              # Seconds between captures (for daemon mode)
  monitors: [1]             # Which monitors to capture
  skip_similarity: 0.95     # Skip similar screenshots

# VLM settings
vlm:
  provider: openai          # openai / ollama
  model: gpt-4o-mini        # Model to use
  api_key: null             # Or use OPENAI_API_KEY env var

# Ollama settings (when provider=ollama)
ollama:
  host: http://localhost:11434
  model: llava:13b

# Storage settings
storage:
  path: ~/.glimpse          # Data directory
  save_screenshots: false   # Save original screenshots

# Privacy settings
privacy:
  retention_days: 30        # Auto-delete after N days
  excluded_apps:            # Skip these apps
    - "1Password"
    - "Keychain Access"

Glimpse is designed with privacy in mind:

• Local Storage — All data stays on your machine in ~/.glimpse/
• No Cloud Sync — Nothing is uploaded unless you explicitly use cloud VLM APIs
• Sensitive Detection — VLM is prompted to skip password fields and sensitive content
• App Exclusion — Configure apps to never capture
• Auto Cleanup — Old data is automatically deleted based on retention settings

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Run linter
ruff check .

# Type checking
mypy glimpse

Glimpse uses SQLite with FTS5 for efficient full-text search. Data is stored in:

~/.glimpse/
├── config.yaml      # Configuration
├── glimpse.db       # SQLite database
└── screenshots/     # Optional screenshot storage

• Daemon mode for continuous capture
• Vector/semantic search
• Smart change detection (skip similar screens)
• Todo extraction from activities
• Export to Markdown/JSON
• Web UI for browsing history

Contributions are welcome! Here's how you can help:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Please make sure to:

• Update tests as appropriate
• Follow the existing code style
• Update documentation for any new features

This project is licensed under the MIT License - see the LICENSE file for details.

Inspired by Rewind and MineContext, but designed to be lightweight, open-source, and privacy-focused.