Skip to content

README.md

Latest commit

 

History

History
122 lines (83 loc) · 4.03 KB

README.md

File metadata and controls

122 lines (83 loc) · 4.03 KB
Raw

ElCorrecto

ElCorrecto is a resume-to-role analysis application. Upload a CV as a PDF, paste a job description, and receive an AI-powered comparison showing the match score, matched skills, missing skills, and suggested improvements.

The application uses a React and TypeScript frontend with Ant Design and Tailwind CSS. Its Azure Functions backend uploads the resume to Azure Blob Storage, extracts text with Azure Document Intelligence, and sends the extracted resume and job description to Azure Foundry for analysis.

Project structure

api/       Azure Functions API and Azure service integrations
frontend/  React, TypeScript, Vite, Ant Design, and Tailwind UI
bruno/     Bruno collection for testing the API

Application architecture

flowchart LR
	User[User] --> Frontend[React frontend\nAzure Static Web Apps]
	Frontend -->|Axios multipart/form-data| Function[Azure Function\nAnalyzeResume]
	Function -->|Store PDF| Blob[Azure Blob Storage]
	Function -->|Extract text| Document[Azure Document Intelligence]
	Function -->|Resume text + job description| Foundry[Azure AI Foundry\nGPT deployment]
	Foundry -->|Structured JSON analysis| Function
	Function -->|Analysis response| Frontend
Loading

The application is split into two deployable layers:

  • Frontend: A Vite-powered React and TypeScript single-page application hosted on Azure Static Web Apps. It handles PDF selection, job-description input, loading and error states, and result presentation. The Azure Function URL is supplied through VITE_AZURE_FUNCTION_URL.
  • API: A Node.js Azure Functions v4 HTTP API hosted in the analyze-resume Function App. It validates the multipart request, uploads the PDF, extracts resume text, calls Azure AI Foundry, and returns typed JSON.
  • Storage and AI services: Blob Storage retains the uploaded PDF, Document Intelligence performs OCR and text extraction, and Azure AI Foundry evaluates the resume against the job description.

The browser communicates only with the HTTP Function endpoint. Azure credentials remain server-side in Function App environment variables; the frontend exposes only its public API URL.

How it works

  1. The user drops a PDF resume into the frontend.
  2. The user pastes the target job description.
  3. The frontend sends both fields as multipart/form-data with Axios.
  4. The Azure Function stores the PDF in Blob Storage.
  5. Document Intelligence extracts the resume text.
  6. Azure Foundry compares the resume with the job description.
  7. The frontend displays the structured analysis.

Setup

Install dependencies in both workspaces:

npm install --prefix api
npm install --prefix frontend

Copy the API settings template and add your Azure credentials:

cp api/local.settings.example.json api/local.settings.json

The frontend Function URL is configured in frontend/.env. Use frontend/.env.example as a template when setting up another environment. Never commit real secrets or local settings files.

Required Azure services:

  • Azure Storage Account with a resumes Blob container
  • Azure Document Intelligence resource
  • Azure AI Foundry model deployment
  • Azure Function App

Run locally

Start the Azure Functions API from the project root:

npm start

The API runs at http://localhost:7071/api/analyze-resume.

Start the frontend in a second terminal:

npm run start:frontend

Open http://localhost:5173/ in your browser.

Build and checks

Build both applications:

npm run build

Run frontend checks:

npm --prefix frontend run lint
npm --prefix frontend run format:check

Format the frontend with Prettier:

npm --prefix frontend run format

API request

The API expects a POST request with multipart/form-data:

  • resume: PDF file
  • jobDescription: job description text

The deployed Function endpoint is:

https://analyze-resume-ebgeb5bkaubjchhu.westus3-01.azurewebsites.net/api/analyze-resume

A Bruno request is available in bruno. The API-specific setup notes are in api/README.md.