This release marks an important milestone in heptabot development, as all of our major goals have now been reached. The project was split into two branches, cpu with support for lightweight PyTorch model that runs only on CPU and gpu-tpu with more sophisticated T5 v. 1.0 TensorFlow models capable of producing better results. We also release three versions, tiny, medium and xxl, which differ by the checkpoint size of T5 text generation model employed. For each of these versions, we provide a notebook so that they could be conveniently used in Google Colab or Kaggle Kernels. This release also features updated installation procedures (we switched to Docker images as they are more stable and more convenient to work with), enhanced diff computing between the original and corrected texts (the error spans now align better), a function to convert heptabot's output to .ann format used by brat and some bug fixes for better stability.