A deep learning engine for your own personal virtual assistant

Animius is an open source software library for creating deep-learning-powered virtual assistants. It provides an intuitive workflow that extracts data from existing media (such as anime and TV shows) and trains on them to provide a personalized AI. The flexible architecture enables you to add custom functionality to your virtual assistant.

Animius also ships with a high-level API animius.Console that allows users without programming experience to use Animius.


Install the current release from PyPi:

pip install animius

Then, install Tensorflow (recommended version 1.14. Animius does not support TF 2.0). We recommend using the GPU package (tensorlfow-gpu) if you are going to train your own virtual assistant. Read more on Tensorflow installation here.

See Installing Animius for detailed instructions and Docker installation guide.

Getting Started

For more information, check out our quick start guides.

Here’s a quick overview of how to create an intent-ner model with Animius:

# create a new model config for intent ner and set default values
myIntentModelConfig = am.ModelConfig(cls='IntentNER')

# we can set & view the values of the config before creating the model 
myIntentModelConfig.hyperparameters['learning_rate'] = 0.01

# load a pretrained word embedding from disk
myGloveEmbedding = am.WordEmbedding()
myGloveEmbedding.create_embedding('\\resources\\glove.twitter.27B.50d.txt', vocab_size=40000)

# create a data object for your training set
myIntentData = am.IntentNERData()
myIntentData.set_intent_folder('resources\\IntentNER Data')

# create a model
myIntentModel = am.IntentNER.IntentNERModel()

# build a TF graph based on the config and data
myIntentModel.build_graph(myModelConfig, myData)
# start tensorflow

# train!

After training the intent-ner and chatbot models, we can create a waifu - a virtual assistant. The code is a bit long, so I will skip a lot of it. Feel free to check out our demo and our quick start guide for more information.

Anyways, you can simply interact with a working waifu like this:


myWaifu.predict('What is the time in New York right now?')

myWaifu.predict('Play me some music')

For more information

Also, feel free to check out our other projects surrounding Animius:


Apache License 2.0

Xiaoyang Liu
Xiaoyang Liu
Incoming Freshman at UCLA
Software Engineer at Wortharead