BLAB Chatbot - DEEPAGÉ

This Python module allows the integration of DEEPAGÉ with BLAB.

Currently, the Python versions supported by this module are {v | 3.8.1 ≤ v < 3.9}.

Installation

• Obtain the trained DEEPAGÉ model. It should be a directory containing files such as pytorch_model.bin and optimizer.pt. If you don't have access to the model, please contact the authors of DEEPAGÉ.

• Obtain the document containing the texts. Each line should have a title and a text, separated by a tab character. If you don't have access to the documents, please contact the authors of DEEPAGÉ.

• Install a version of Python between 3.8.1 (inclusive) and 3.9 (exclusive). Other versions are not supported.

• Make sure that distutils is installed (e.g. the package python3.8-distutils in some Linux distributions). It should be possible to execute python3.8 -c 'import distutils; print("OK")' without errors.

• Install Elasticsearch 7.10 - see instruction to install from a .tar.gz archive or as a deamon. Other versions may not be supported.

• Follow these installation instructions using settings_deepage_TEMPLATE.py as a template.

  If errors are shown, install the following packages on your system and try again:

  libblas3 liblapack3 liblapack-dev libblas-dev gfortran libatlas-base-dev
  

• Start Elasticsearch and wait a few seconds.

• Enter the src/ directory and create the index:

  poetry run ./run.py --config=name_of_your_config_file.py index

• Follow these instructions to execute the program. The addition to your controller settings can be:

  CHAT_INSTALLED_BOTS.update({
      "DEEPAGÉ": websocket_external_bot(url="http://localhost:25226"),
  })

Documentation for the code and its API is hosted on GitHub Pages alongside instructions on how to generate it locally.