Witsy is a desktop AI assistant designed for users who want to leverage various Large Language Models (LLMs) and AI services with their own API keys or local Ollama instances. It offers a unified interface for chat, text-to-image, text-to-video, and document interaction, aiming to boost productivity across applications.

How It Works

Witsy acts as a universal client, abstracting away the complexities of interacting with numerous AI providers through a consistent API layer. It supports a wide array of LLM providers (OpenAI, Anthropic, Google, MistralAI, Groq, etc.) and multimodal services (text-to-image, text-to-video). Key features include "Prompt Anywhere" for in-app text generation, "AI Commands" for context-aware text manipulation, and Retrieval-Augmented Generation (RAG) for chatting with local documents.

Quick Start & Requirements

• Install: Download binaries from witsyai.com or the releases page, or build from source (npm install, npm start).
• Prerequisites: API keys for services like OpenAI, Anthropic, Google, MistralAI, Groq, Tavily (for internet search), and ElevenLabs (for text-to-speech). Ollama must be installed and configured for local model usage.

Highlighted Details

• Supports chat completion with vision models and text-to-image/video generation from multiple providers.
• Features "Prompt Anywhere" and "AI Commands" for seamless integration with any desktop application.
• Enables RAG for local document interaction and offers Speech-to-Text (via Whisper) and Text-to-Speech capabilities.
• Includes an extensible plugin system for features like Python code execution and internet search.

Maintenance & Community

The project is actively maintained by nbonamy, with a GitHub Actions workflow for testing and coverage reporting. Sponsorships are accepted via GitHub Sponsors.

Licensing & Compatibility

The README does not explicitly state a license. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

Auto-update functionality is noted as not working on Windows. Some features, like automatic expert selection based on the foreground app, are Mac-specific. The project is still marked with a "TODO" for proper database storage.