# LUCI *Logical Unit for Communicational Interactivity*

> [![Docs Link](https://camo.githubusercontent.com/1befdf0564cbd07feb821f495807479244aa25bb17ee9ce27eeaa98471e4ab72/68747470733a2f2f62616467656e2e6e65742f62616467652f646f63732f6769746875625f77696b693f69636f6e3d676974687562)](https://github.com/brunolcarli/Luci/wiki) LUCI is a conversational tamabotchi for discord.

## Developers ### Cloning and executing ![Apple badge](https://camo.githubusercontent.com/cc08722b89ec9207a64c37fce804239cada13454114db48dcac7625bdacd15b1/68747470733a2f2f62616467656e2e6e65742f62616467652f4f532f4f53582f3a636f6c6f723f69636f6e3d6170706c65) #### Local machine Clone this project and engage a **python3** virtual environment the way as you like and install the requirements through Makefile. Example using virtualevwrapper: ``` $ git clone https://github.com/brunolcarli/Luci.git $ cd luci/ $ mkvirtualenv Luci $ (Luci) make install ``` Create a `.env` file and add the following content: ``` TOKEN= LISA_URL= BACKEND_URL=None SETTINGS_MODULE=development REDIS_HOST= REDIS_PORT= ``` You should configure your own environment cloning both lisa and bot api: - `LISA_URl` is required since the bot requests text offense and sentiment to the service API. - `BACKEND_URL` is the long range memory storage (AKA database) which is a separated service. - `REDIS_HOST` and `REDIS_PORT` is the short term memory, and it depends on a redis-server. Note: If `SETTINGS_MODULE=production` the bot will run on a flask server instance. Execute the bot: ``` $ (Luci) make run ``` ### Training Models The bot is trained over a serie of .`json` datasets on `core/training/json/intentions/`. For each intention directory there are located a serie of `dataset_x.json` where `x` is a sequence number of the train dataset. You can create a new one containing the text and training target label on the following shape: ```json [ {"text": "An example text.", "intention": 1}, {"text": "Another example text.", "intention": 1}, {"text": "Way othe example text.", "intention": 2}, {"text": "keep going.", "intention": 2} ] ``` The training data **must** respect the intention archicture proposed on the docs, otherwise it may cause the model to behave unproperly. Train data through Makefile: ``` $ (Luci) make train ``` You can view some models score and the number o sample data used for each intention through a cvross validation test: ``` $ (Luci) make no_free_lunch ``` ### Test the bot The goal is give the models enough data to pass the test when executing: ``` $ (Luci) make test ``` ## Docker

Set up the `.env` file as described above and build the container with docker-compose: ``` $ docker-compose build ``` Run with: ``` $ docker-compose up ``` Or run in background ``` $ docker-compose up -d ```