AquilaX-NL-JSON-Start-Scan

Overview

AquilaX-NL-JSON-Start-Scan is a model built using Hugging Face's T5-small to convert natural language queries about vulnerabilities into JSON queries for MongoDB.

Model Information

Model

  • Name: AquilaX-NL-JSON-Start-Scan
  • Architecture: T5-small
  • Framework: Hugging Face Transformers

Description

The AquilaX-NL-JSON-Start-Scan model is designed to interpret natural language queries related to vulnerabilities in code and convert them into JSON queries that can be executed on a MongoDB database. This facilitates automated scanning and analysis of code repositories for security issues. The model leverages the capabilities of the T5-small architecture, which is well-suited for natural language understanding and generation tasks.

Getting Started

Usage

Below we share some code snippets on how to get quickly started with running the model. First make sure to pip install -U transformers[torch], then copy the snippet from the section.

Requirements

pip install transformers[torch]

Inference Code

import json
import requests

def convert_to_json(answer):
    """
    Convert a string representation of a dictionary to a JSON object.

    This function takes a string representation of a dictionary, cleans it by removing
    specific unwanted tokens and correcting boolean representations, and then converts
    it into a JSON object.

    Parameters:
    answer (str): The input string representing a dictionary.

    Returns:
    dict: The JSON object converted from the input string.
    """
    answer = answer.replace("<pad>", "").replace("</s>", "")
    answer = answer.strip("'")
    answer = answer.replace("false", "False").replace("true", "True")
    answer_dict = eval(answer)
    answer_json = json.dumps(answer_dict)
    json_data = json.loads(answer_json)
    return json_data

def valid_url(url):
    """
    Validate the given URL against a list of supported platforms.

    This function checks if the provided URL belongs to one of the supported
    platforms for scanning. If the URL is valid, it returns True. Otherwise,
    it returns a message indicating that the URL is not supported and lists the
    available scanners.

    Parameters:
    url (str): The URL to be validated.

    Returns:
    bool or dict: Returns True if the URL is valid, otherwise returns a
                  dictionary with a message indicating the URL is not supported
                  and lists the available scanners.
    """
    valid_list = [
        "github.com", "bitbucket.org", "sourceforge.net", "aws.amazon.com",
        "dev.azure.com", "gitea.com", "gogs.io", "phabricator.com",
        "gitkraken.com", "beanstalkapp.com", "gitlab.com"
    ]
    platform = url.split("//")[1].split("/")[0]

    if platform in valid_list:
        return True

    return {
        'message': 'Provide a valid URL for scanning. Currently, we support PII_Scanner, SAST_Scanner, Sac_Scanner (Open_Source_Security), IaC_Scanner, Container_Scanner'
    }
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch
import time

tokenizer = AutoTokenizer.from_pretrained("AquilaX-AI/NL-JSON-Start-Scan")
model = AutoModelForSeq2SeqLM.from_pretrained("AquilaX-AI/NL-JSON-Start-Scan")

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# Change YOUR_QUERY eg: can this https://github.com/mr-vicky-01/educational-assitant on every week using pii and sast scan
query = "Translate the following text to JSON: " + "YOUR_QUERY".lower()
query = query.replace(",", "")

start = time.time()

inputs = tokenizer(query, return_tensors="pt")
model.to(device)
inputs = inputs.to(device)
outputs = model.generate(**inputs, max_length=256)
answer = tokenizer.decode(outputs[0])
try:
  json_data = convert_to_json(answer)
except:
  json_data = {'message': 'We encountered an issue with your query. Please use the Personalized Scan option for accurate results.'}

to_return = json_data.copy()
try:
    valid = valid_url(json_data["repo"])
    if valid != True:
        to_return = valid
    else:
      url = re.findall(r'https?://\S+', query)
      to_return['repo'] = url

except:
    pass

end = time.time()
print(to_return)
print(f"Time taken: {end - start}")

License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

Authors

Acknowledgments

  • Hugging Face for the Transformers library.
Downloads last month
13
Safetensors
Model size
60.5M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.