shaz13 / katana Goto Github PK

View Code? Open in Web Editor NEW

109.0 5.0 22.0 561 KB

Katana project is a FastAPI template for ASAP 🚀 ML API deployment

Python 95.58% Shell 1.94% Dockerfile 1.83% Procfile 0.64%

fastapi-template docker machine-learning api-rest uvicorn

katana's Introduction

Introduction 🌻

Katana project is a template for ASAP 🚀 ML application deployment

Features 🎉

FastAPI inbuilt
Swagger UI and uvicorn integration
Docker ready configuration
Integrated GitHub actions
Production ready code 🚀

Set-up Instructions 🔧

We recommend using flask default serving for development and uvicorn server for production

We included following setup instructions;

Local development
Docker supported deployment

Local Development 👨🏻‍💻

Clone this repo with [email protected]:shaz13/katana.git
Set up environment using python3 -m venv .env
Activate envrionment using

# Linux / Mac / Unix
$ source .env/bin/activate

# Windows
$ \.env\Scripts\activate

Install requirements using pip install -r requirements.txt
For debugging run from root - python main.py
Deploy using Procfile or bash scripts/launch.sh
Your API is being served at localhost:9000

Docker Setup ⛴

Clone this repo with [email protected]:shaz13/katana.git
Install docker in your system
Run docker-compose up
Your local port is mapped and being served at localhost:9000

Contributors 😎

Mohammad Shahebaz - @shaz13
Aditya Soni - @AdityaSoni19031997

License 👩🏻‍💼

MIT License

katana's People

Contributors

Stargazers

Watchers

katana's Issues

Add api versioning and blueprint

Current app base url has no versioning according to swagger design standards. The PR should address it and have current endpoints work with /api/v1 route

Fix logging and set apt level of logging

Logging at gunicorn.access is not working. All goes into gunicorn.error

Add docker compatibility

Feature Request : Custom Artifacts Upload for Scikit/XGB Inference

Lets me upload joblib/pkl file (FEATURE REQUEST) for scikit/xgb trained models.
I am allowed to edit a payload json.
Hit on Execute to perform inference on my joblib model with the pred_input.json as input and printout status(200 Success) and confidence/other metrics.

The above can be done using GCS where I have to create a bucket and use their AI Platform for prediction phase. However that is not beginner-friendly and requires a lot of preliminary GCP setups and API enabling overheads.

Please look into this @shaz13 @shaz13-socgen @AdityaSoni19031997