rest-api

Bug Detection Container

Docker image using flask to test the bug detection pipeline.

Quickstart

Docker Compose

docker compose up

Docker

docker build . --tag bug-repair-flask
docker run --it -p 8080:80 bug-repair-flask:latest

This will download all the dependencies from requirements.txt and the models from huggingface. Then it will run the Flask rest api on the address: localhost:8080.

Rest API

The application responds to the following routes:

• POST /api/inference The application will read the field "source_code" from the body of the request and will respond with a json object containing three fields "error_description", "token_class" and "source_code". The error description will contain the description of the error in natural language, the token class will contain an array for each character of the source code with 1 for buggy and 0 for correct, and the source code field will contain the repaired source code. The application can also be given as input the field "beam_size" also inside the body of the post request. This will choose a non greedy decoding method and will return instead a list will all possible results for the error description, token classes and new source code.

  For example the following curl request should return the error message, token classes and new source code.

  curl -X POST -m 200 -H "Content-Type: application/json" -d '{"source_code": "A = map(input().split())\nprint(A)"}' http://localhost:8080//api/inference

  And by using

  curl -X POST -m 200 -H "Content-Type: application/json" -d '{"source_code": "A = map(input().split())\nprint(A)", "beam_size": 5}' http://localhost:8080//api/inference

  one should expect to receive 5 examples for error descriptions, 5 examples for token classes and 25 (5 * 5) example for new source code examples.