API Definition

This project is a mail classifier for an hypotetical energy company. It reads the mails from the users and distributs them in different classes depending on the content using a zero-shot classifier, in this case the facebook bart as it is a small and fast which can work really good with the complexity of the task.

It reads the data from a SQL file in a Docker file with the bart model.

I made this project to practice the Docker initialization, joining SQL databases and LLM models making it callable and scalable.

Endpoint: /classify-email

Method: POST

Parameters (sent in the body as JSON text):

{ 
  "client_id": integer, 
  "fecha_envio": string, 
  "email_body": string 
} 

Response

If the client has no outstanding debts

Response code: 200

Response body (in JSON format):

{ 
  "success": boolean,
  "prediction": string 
} 

If the client has outstanding debts

Response code: 200

Response body (in JSON format):

{ 
  "success": false,
  "reason": "The client has outstanding debts"
} 

Description: The /classify-email endpoint accepts requests with the POST method. It expects three parameters in the request body: client_id, which must be an integer; fecha_envio, which represents the date and time of sending in datetime format (format YYYY-MM-dd hh:mm:ss); and email_body, which is a string containing the body of the email to classify. If the request is successfully processed, the endpoint will return a status code 200 and a JSON object with a field named prediction, which will contain a string with the prediction, along with a field success, a boolean with the value true.

If the client is listed as having outstanding debts, the endpoint will return a status code 200 and a JSON object with a success field set to false and a reason field with the message "The client has outstanding debts".