VIRTUAL ASSISTANT GUIDE

OVERVIEW

Gwen the Virtual Assistant

Welcome to the repository for Gwen, a female virtual assistant agent who is the link between your voice activated prompts and your prefered compute platform. Through Python, Gwen is provided a platform to control and manage your desktop environment in hopes to perform tasks that you might otherwise find remidial.

INSTALLATION GUIDE

We are currently working on building an installation for the project such that anyone can download and install Gwen with minimal experience with Python Programming.

Installation Guide coming Soon

DEVELOPMENT GUIDE

Backend API

Curious on how it works? Gwen's job is to translate your commands into backend API calls in the form of JSON strings that mirror the following format:

{"target": "ClassName.FunctionTarget", ...}

She also includes a set of keyword arguments for the specific function.

Much of Gwen's backend implements AI technologies through the world-wide web, however we are currently developing many different useful AI models for speeding up the data pipeline. For more information regarding current model development, please reach out to:

Jonathan Koch: johnnykoch02@gmail.com

Backend Classes and Functions

Here's a look at the current featured backend classes and their respective function calls:

• Class Gwen:

  • clear_context(): Clears all running contexts and switches back to passive mode.
  • collect_keyword_data(int num_samples): Collects new user Keyword activations for the model to use and train.
  • output_speech(string text): Uses the Configured TTS Engine to speak the given text.

• Class YouTube:

  • play(string search_query, string channel_name = None): Searches and plays the top search result for the provided query and channel if present.

• Class Spotify:

  • play(string song, string artist): Plays a Song with a specified artist.
  • pause(): Pauses the music of the Spotify Bot.

• Class Netflix:

  • watch(string query): Plays a Show/Viewing for a specific target query.

Usage Examples

Here's some examples of how Gwen translates some user commands into backend API calls:

1. "Open up Spotify and play Radioactive by Imagine Dragons":

   {"target": "Spotify.play", "song": "Radioactive", "artist": "Imagine Dragons"}
   

2. "Play 3Blue1Brown's video series about deep learning":

   {"target": "YouTube.play", "query": "Deep Learning", "channel_name": "3Blue1Brown"}
   

3. "Search YouTube for Funny Cat Videos":

   {"target": "YouTube.play", "query": "Funny Cat Videos"}
   

4. "Stop", "Turn this off", "Close the current application":

   {"target": "Gwen.clear_context"}
   

5. "What is Euler's Theorem?":

   {"target": "Gwen.output_speech", "text": "Euler's theorem, also known as Euler's identity, is a mathematical equation discovered by Swiss mathematician, physicist, and astronomer Leonhard Euler. It states that e^(i * pi) + 1 = 0, where e is the base of the natural logarithm (known as Euler's number), i is the imaginary unit, and pi is a ratio of a circle's circumference to its diameter."}
   

Future Development

We have a list of features we want to add and improvements we're planning to make. We invite you to suggest more:

• LLM Server: Run a LLM for some provided weights path, receive GET REQUESTS and respond appropriately.
• LLM Trainer: Uses supervised learning to train input weights for some transformer.
• Gwen Repo: Expand backend capability, fix the Keyword model, potentially move the backend of this to Tinygrad, reduce API call delay, and more!
• Add a Gwen Backend: Prompt Processor for the user, can be trained using SL and/or RLHF via LoRA.

Contribute

We welcome all contributions! If you have suggestions or want to report bugs, please don't hesitate to open an issue or submit a pull request.