deeppavlov-chatbot

Requirements

• Python (>=3.6)
• Node (>=v8)
• Linux or Windows (installation for Windows requires git and Visual Studio 2015/2017 with C++ build tools installed!)

Installation

0. Clone this repository or download the latest version.

1. Create and activate a virtual environment:

   • Linux

   cd nlu
   python -m venv env
   source ./env/bin/activate

   • Windows

   cd nlu
   python -m venv env
   .\env\Scripts\activate.bat

2. Install the Python requirements:

   pip install -r requirements.txt
   

3. Download and install spacy en_core_web_sm:

   python -m spacy download en_core_web_sm
   

4. Download and install DeepPavlov datasets for intent classification:

   python -m deeppavlov download model_config.json
   python -m deeppavlov download intents.json
   

5. Now install the frontend dependencies:

   cd ../chatbot
   npm install
   

Running the chatbot

1. Launch DeepPavlov REST API:

   cd nlu
   DP_SETTINGS_PATH=./deeppavlov-settings python -m deeppavlov riseapi model_config.json -p 4949

   The swagger api docs should be available at http://localhost:4949/.

2. Launch the frontend server:

   cd chatbot
   PORT=3000 npm start

   You should be able to access the web chat interface at http://localhost:3000/.

Training the models

This repository comes with pre-trained models for our chatbot. If you decide to add, remove, or modify the intents and/or the entities, you will have to re-train the models.

Intent model

python -m deeppavlov train intents.json

Entities model

python -m deeppavlov train ner.json

Docker

docker build -t ehrhart/deeppavlov-restaurants/chatbot . docker run -p 3000:3000 -d ehrhart/deeppavlov-restaurants/chatbot