Overview

Let's check Textgrader Backend Project Overview in 5 minutes.

Technologies

Textgrader API is a FastAPI project, so it uses FastAPI concepts that are avaiable with more details here. Textgrader was originally a AI project, in this page we will cover just what is more relevant for Textgrader API use case.

Along with FastAPI this project also uses:

1. Python (Python is a high-level, general-purpose programming language)
2. FastAPI (FastAPI is a modern, fast (high-performance), web framework for building APIs with Python)
3. TensorFlow (TensorFlow is a free and open-source software library for machine learning and artificial intelligence.)
4. Scikit-learn (Scikit-learn is a free software machine learning library for the Python programming language)

Running the Project Natively

Besides using Docker, it's also possible to run the backend project natively, for this you will need installed on your machine:

1. Python (Install 3.10 version)

With this installed you just need to run the commands bellow in backend folder:

python -m venv venv && . venv/bin/activate && pip install -U pip

pip install -U "jax[cpu]" tensorflow-cpu tensorflow_text tensorflow_hub scikit-learn nltk gensim pyarrow

pip install -U pandas fastapi "uvicorn[standard]" orjson openpyxl colorlog pyspellchecker coloredlogs

cd src && python requirements.py && python run.py task general_tasks task_correct_essays

python run.py task general_tasks task_pipeline_essays && uvicorn api:app --host 0.0.0.0 --reload"

File Tree Structure

/backend

├── src                                     | Backend source directory
│   ├── README.md                           | Textgrader AI README.md
│   ├── api.py                              | Backend API code
│   ├── dags                                | Textgrader AI folder
│   │   ├── config.py                       | Textgrader AI configuration file
│   │   ├── escondo                         | Textgrader AI folder
│   │   │   ├── general.py                  | Textgrader AI code
│   │   │   ├── runner.py                   | Textgrader AI code
│   │   │   └── simple.py                   | Textgrader AI code
│   │   ├── experiments                     | Textgrader AI folder
│   │   │   ├── experiments.py              | Textgrader AI code
│   │   │   └── feature_engineering.py      | Textgrader AI code
│   │   ├── feature_engineering             | Textgrader AI folder
│   │   │   ├── generate_features.py        | Textgrader AI code
│   │   │   ├── latent_semantic_indexing.py | Textgrader AI code
│   │   │   └── word_embeddings.py          | Textgrader AI code
│   │   ├── general_tasks.py                | Textgrader AI code
│   │   ├── logging.py                      | Textgrader AI code
│   │   ├── model_training                  | Textgrader AI folder
│   │   │   ├── doc_to_vec_trainer.py       | Textgrader AI code
│   │   │   └── train_model.py              | Textgrader AI code
│   │   ├── predict                         | Textgrader AI folder
│   │   │   ├── predict.py                  | Textgrader AI code
│   │   │   └── predict_from_text.py        | Backend code to grade essays on real time
│   │   ├── preprocessing                   | Textgrader AI folder
│   │   │   ├── general_preprocessing.py    | Textgrader AI code
│   │   │   ├── spell_correct.py            | Textgrader AI code
│   │   │   └── word_features.py            | Textgrader AI code
│   │   ├── run.py                          | Textgrader AI code
│   │   ├── runner.py                       | Textgrader AI code
│   │   └── utils.py                        | Textgrader AI code
│   ├── requirements.py                     | Backend setup script
│   └── run.py                              | Textgrader AI code
└── vercel.json                             | Backend vercel configuration file