Wingman AI

What it does

Imagine you could talk to AI characters anywhere in whatever language you like and they’ll respond. Just picture you could control not only their personality and voice, but also teach them how to carry out commands for you and draw on external knowledge in your games.
Wingman AI is a local client-server app that runs on Windows and MacOS. The backend is open source and written in Python, while the frontend is a closed source Web App. The UI is easy to use, so both AI fans and gamers can use different AI providers to make their Wingmen their own.
A Wingman is a combination of personality prompts, a text-to-speech, speech-to-text and LLM provider configuration, command definitions and optional skills.
Wingman AI Core handles the synthesis of the user’s speech, then sends it to Gemini, processes the response and finally transforms it back into speech. It uses AI function calling to execute commands like keypresses or macros that users can record in the frontend and run without having to learn exact activation phrases.
Skills are basically attachable custom Python scripts/extensions that allow a Wingman to run any code which can be used to connect to external APIs, SDKs, AI vision etc.
Wingman AI isn’t limited to Gemini but we integrated it as first-class provider as soon as we got access to it. We found it to be an excellent model to handle complex system prompts and long conversations and are also impressed by its summarization capabilities.