embedding-demo-api

REST API for review sentiment and natural image classification web app.

Live Deployment

setup

• install dependencies running pip install -r requirements.txt
• get basilica API key and input into basilica scripts
• navigate to training directory
• run both basilica scripts
  • python basilica_scripty.py
  • python basilica_image_script.py
• run build model script
  • python build_model.py both
• run export FLASK_APP=review_api.py
• setup db
  • flask db init
  • flask db migrate -m "setting up db"
  • flask db upgrade
• run app:
  • flask run

routes

"/" - [GET] - simple ping with 'hello' response

"/predict/sentiment" - [POST] - get product review sentiment prediction

ex request json:

{
  "review": "Text of the review"
}

ex response json:

{
  "class": "positive"/"negative",
  "confidence": 1-0,
  "review": "text of the review"
}

"/predict/image" - [POST] - get natural image class prediction

request body has image named as "file".

ex response json:

{
  "class": "airplane"/"person"/"car"/"cat"/"dog"/"fruit"/"flower"/"motorbike",
  "confidence": 1-0
}

"/accuracy/sentiment" - [POST] - route to track whether prediction was correct or incorrect

ex request json:

{
  "answer": "correct"/"incorrect",
  "prediction": "positive"/"negative"
}

ex response json:

{
  "total": int,
  "true_positives": int,
  "true_negatives": int,
  "false_positives": int,
  "false_negatives": int
}