PyGPT - pygpt.net

Release: 2.9.1 (2026-10-06)
Official Website: https://pygpt.net
GitHub: https://github.com/szczyglis-dev/py-gpt
Add-ons: https://github.com/szczyglis-dev/py-gpt-addons
PyPI: https://pypi.org/project/pygpt-net
AppImage(s): https://github.com/szczyglis-dev/py-gpt/releases
Snap Store: https://snapcraft.io/pygpt
Microsoft Store: https://apps.microsoft.com/detail/XP99R4MX3X65VQ
Donate: https://www.buymeacoffee.com/szczyglis

Contents

Introduction
- Overview
- Features
How to Install
- Prebuilt binaries
- Microsoft Store
- AppImage
- Snap Store
- PyPI (pip)
- Running from source
  - Poetry
- Troubleshooting
- Other requirements
- Debugging and logging
Quick Start
Modes
- Chat
- Agents
- Realtime + audio
- Research
- Completion
- Image and video generation
- Computer use
- Experts
- Autonomous mode
- Custom agents
Canvas
- Annotations
- Live interactive rendering
- Using a Painter drawing as model input
- Example
- Websites and local HTML server
- Enabling Canvas
Agent Skills
- Overview
- Managing Skills
- Importing from GitHub
- Explore catalog
- Runtime behavior
- Working directory and scripts
- Resources and binary files
- Security and permissions
- Creating a simple Skill
Context and memory
- Short and long-term memory
- Settings layout
- Advanced context handling (experimental)
- Handling multiple contexts
- Projects and project data workdirs
  - Project attachment Library
- Clearing history
- Context storage
- Automatic extra-context summary
Files and Attachments
- Uploading attachments
- Downloading files
Indexing and RAG
- Using RAG in Chat
- Indexing files for RAG
- Querying single files
- Index types
- Supported data
- File indexing
- Context indexing
  - Global context indexes
- Isolated project indexes
  - Project lifecycle
- Using the current project index
- Attachments and temporary RAG
- Vector stores
- Embeddings
- Data loaders
- Clear and truncate
- Performance and indexed-file tracking
- Token and usage notes
Presets
- What is preset?
- Example usage
Profiles
- Importing and exporting profiles
Models
- Built-in models
- Adding a custom model
- Custom providers (OpenAI-compatible)
- Per-model API base and API key
- How to use local or other models
Plugins
- Overview
- Creating Your Own Plugins
- API calls
- Audio input
- Audio output
- Autonomous mode
- Canvas (inline)
  - Backends and runtime rules
  - Options
  - Canvas tools
  - Annotations
  - Local preview server
  - Security and practical notes
- Bitbucket
- Chat history (inline)
- Crontab / Task scheduler
- Custom commands
- Experts (inline)
- Extra system prompt
- Facebook
- Filesystem, Python and OS
- GitHub
- Google (Gmail, Drive, Calendar, Contacts, YT, Keep, Docs, Maps, Colab)
- Image generation (inline)
- Jev / System One (inline)
- Mailer
- MCP
- Memory (inline)
- Mouse and keyboard
- OpenStreetMap
- Python execution
- RAG (inline)
- Real time
- Serial port / USB
- Server (SSH/FTP)
- Slack
- System commands
- Telegram
- Tuya (IoT)
- TwelveLabs
- Vision (inline)
- Voice control (inline)
- Web search
- Wikipedia
- Wolfram Alpha
- X/Twitter
MCP Connectors
- Overview
- Managing Connectors
- Importing existing MCP configurations
- Explore catalog
- Runtime behavior
- Security
Tools
- Notepad
- Painter
- Calendar
- Indexer
- Media Player
- Image Viewer
- Text Editor
- Transcribe Audio/Video Files
- OpenAI / Google Vector Stores
- Python/OS
  - Managing sandbox environments
- Terminal
- HTML/JS Canvas
- Translator
- Canvas
- Agent Workflow
- Custom agent builder
Functions, commands and tools
Token usage
- Input tokens
- Total tokens
Accessibility
Configuration
- Settings
  - General
  - API Keys
    - OpenAI
    - Google
    - Anthropic
    - HuggingFace
    - DeepSeek
    - xAI
    - Azure OpenAI
    - Perplexity
    - Mistral AI
    - VoyageAI
    - OpenRouter
    - Forge
    - Eden AI
    - Jev
  - Layout
    - General
    - Code syntax
  - Files and attachments
    - General
  - Chats
    - List
    - Render
    - Annotations
    - Options
  - Context
    - General
    - Tools
    - Auto-strip
    - Advanced handling
  - Remote tools
    - OpenAI
    - Google
    - Anthropic
    - xAI
  - Models
  - Prompts
  - Images and video
    - Image
    - Video
  - Vision and camera
    - Camera
  - Audio
    - Devices
    - Options
    - Cache
  - Indexes / RAG
    - General
    - Vector Store
    - Chat
    - Embeddings
    - File indexing
    - Context indexing
    - Data loaders
    - Clear and truncate
  - Agents and experts
    - Agents
    - Custom agents
    - Autonomous
    - Options
  - Accessibility
  - Security
    - General
    - Computer use
    - Linux
    - Windows
    - macOS
  - Personalize
  - Custom providers
  - Updates
  - Debug
- Package Manager
- JSON files
- Manual configuration
  - Project data workdirs
- Setting the Working Directory Using Command Line Arguments
- Translations / locale
- Data Loaders
- Vector stores
Updates
- Updating PyGPT
Debugging and Logging
- Logging
- Debug Menu
- Logger / Console
- Access to database
- Debugging compiled builds
- Import tracing
Extending PyGPT
- Overview
- External Add-ons
- Legacy custom launcher registration
  - Registering components programmatically
- Plugins and GUI Tools are different
- Custom themes and styles
  - User themes
- Adding a custom model
- Adding a custom plugin
- Adding a custom GUI Tool
- Adding a custom LLM wrapper
- Adding a custom vector store
- Adding a custom data loader
- Adding audio providers
  - Audio input
  - Audio output
- Adding a web search provider
- Adding a custom agent
- Source code as API reference
Add-ons API
- What is an Add-on?
  - Storage scope and migration
- Quick start: create, test and package an Add-on
  - Minimal directory layout
- Manifest reference
  - Required fields
  - Integrity field: sha256
  - Runtime-only fields
  - Python package namespace and imports
- External dependencies and Package Manager
- Packaging as ZIP
- Local testing
  - Installer-based test
  - Manual-copy test
  - Debugging
- Add-on-owned translations
  - Directory and file naming
  - Automatic domain assignment
  - Fallback and overrides
  - Plugin Add-ons
  - GUI Tool Add-ons
  - LLM provider Add-ons
  - Audio-input, audio-output and Web provider Add-ons
  - Loader Add-ons
  - Other runtime Add-ons
  - Custom launcher objects
- Runtime lifecycle
- Publishing on GitHub and in the public registry
  - Standalone repository
  - Monorepository
  - Public py-gpt-addons registry
- Application API available to Python Add-ons
  - Documented API versus internal object tree
  - Common window access examples
- Plugin Add-ons
  - Base class
  - What a plugin Add-on is for
  - Minimum useful plugin
  - Core methods you will use most often
  - Returning command results
  - Important attributes
  - Additional plugin lifecycle conventions
  - Plugin option schema
  - Command/tool example
  - Event API
  - Complete generic Event reference
  - Typical text-message ordering
  - Full BasePlugin method reference
- GUI Tool Add-ons
  - What a GUI Tool Add-on is for
  - Minimum implementation
  - Creating a tab-capable tool
  - Presentation policy: tabs, dialogs and menu actions
  - Tab bar titles and context menus
  - Left toolbar entries
  - Selecting an independent runtime
  - Lifecycle hooks in practice
  - Full BaseTool method reference
- LLM Add-ons
  - What an LLM Add-on provides
  - Provider identity and capabilities
  - setup(): provider-owned configuration
  - llama(): the main LlamaIndex wrapper
  - Additional capability hooks
  - get_models(): model discovery/import
  - Minimal provider skeleton
  - Full BaseLLM method reference
- Vector-store Add-ons
  - What a vector-store Add-on is for
  - Required identity and state
  - Core persistence methods
  - Minimal directory-backed pattern
  - Full BaseStore method reference
- Data-loader Add-ons
  - What a loader Add-on is for
  - Required metadata and get()
  - Web/external loaders
  - Configurable reader constructor arguments
  - Attachments and runtime compatibility
  - Full BaseLoader method reference
- Audio-input Add-ons
  - What an audio-input Add-on is for
  - Minimum implementation
  - transcribe(path) contract
  - Configuration checks
  - Full audio-input provider method reference
- Audio-output Add-ons
  - What an audio-output Add-on is for
  - Minimum implementation
  - speech(text) and playback ownership
  - Configuration checks
  - Full audio-output provider method reference
- Web-search Add-ons
  - What a web-search Add-on is for
  - Provider lifecycle and settings
  - search() contract
  - is_configured(cmds)
  - Full web-provider method reference
- Agent Add-ons
  - What an Agent Add-on is for
  - Required identity
  - get_agent(): common workflow-provider contract
  - Options and prompts
  - Model selection inside agent roles
  - Direct run() providers
  - Full BaseAgent method reference
- Theme Add-ons
  - What a Theme Add-on is for
  - Package layout
  - Naming and compatibility
- Locale Add-ons
  - What a Locale Add-on is for
  - Package layout
  - Runtime behavior
  - Files preview providers (v2.8.38+)
- Using a custom launcher instead of an installed Add-on
- Compatibility and API-version guidance
- Runnable examples
Security
- Runtime Security settings
- Reporting a Vulnerability
- External Libraries
Credits
- Links
- Special thanks
- Third-party libraries