Atlas AI is a proof of concept for a completely open source home automation assistant with features aimed at facilitating the process of adding more features and commands. The concept is to incorporate the Rasa framework to an easy learning experience for the user and the assistant.
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.
NOTE: This program has been tested with python 3.6 only
Before starting you'll need to have a few libraries installed. First open console in the main program folder and then copy the line below and run it.
It is recomended to setup a vm for this python version, first install miniconda. Then setup a python vm.
conda create -n env-name python=3.6
Then to start using your vm.
conda activate env-name
Finally install the required python packages.
Before starting you'll need to have a few libraries installed. NOTE: If youre using a virtualenv make sure to activate it beforehand.
cd bin/
bash requirements.sh
NOTE: Some actions require enviroment variables to be setup (for the APIs) In linux input the following command
sudo -H gedit /etc/enviroment
Then inside the file write
ENV_VARIABLE_NAME="variableValue"
Save and then log out of the account so the variables are started.
Before training you'll need to setup the datasets and understand how they work. First you need to train the model to understand the specific intent, head over to /data/nlu.md A great tutorial for the nlu.md file can be found here
Once this is done you'll need to work on the /data/stories.md and /configs/domain.yml Information on stories and domain can be found in their respective links.
Later if you have any custom actions defined they'll be edited on the actions.py file. Information on actions can be found here.
Once all of this is set and youre ready to begin training just run the main script with the --train command.
bash bin/run.py --train
If you just want to test the latest NLU training data with some written examples found in /bin/modelTrain.py
bash bin/run.py --train --use-python
Before testing, download and compile duckling. (This is used for time, date and distance recognition)
After compilation in the duckling directory open an example file.
stack exec duckling-example-exe
Now run the actions local server and dialogue program
bash bin/run.sh
If you want to train while you test run is like this.
bash bin/run.sh --interactive
- Guy S Garcia - Initial work - FloppyDisck
See also the list of contributors who participated in this project.
This project is licensed under the MIT License - see the LICENSE.md file for details
- This project would not be even close to possible without the Rasa framework and Spacy