This is AI module serving Columns intelligence need. Some of external services are directly used by api, such as GPT-3. But this module includes all the interfaces serving columns internally.
You can get started with this public API in production from here.
Flask is simple service we use build up a web service, install it from
pip install Flask
We use black for python code formatting. VS code will auto install it and configure it upon the prompt.
Please refer to src/model/dev.md for more details.
To produce training data, we use real transactions in https://www.fina.money. Samples are stored in src/data/cat_train.csv for training.
In real model training, we use about 1 million records.
As time goes, we will have more data to train to improve the model's accuracy, also using user feedback to automatically improve the model.
This repo is super simple, it is structured as
- data: store training data and trained model
- model: source code to train model and serve model
- service.py: flask service to serve model through REST API.