Skip to content

Architecture_Workflow.md

Latest commit

 

History

History
79 lines (62 loc) · 4.83 KB

Architecture_Workflow.md

File metadata and controls

79 lines (62 loc) · 4.83 KB

PlacementHub Architecture and Workflow

This document provides a detailed overview of the system architecture and the step-by-step workflow for the core processes within PlacementHub.

1. System Architecture Flow

The system is designed as a decoupled, event-driven web application. It relies on a React frontend and an Express/Node.js backend, heavily integrated with Azure PaaS (Platform as a Service) resources.

High-Level Architecture Diagram

graph TD
    Client[React Frontend] -->|HTTPS REST| API[Node.js / Express Backend]
    Client -->|WebSocket| WebPubSub[Azure Web PubSub]
    
    API -->|Read/Write| SQL[(Azure SQL Database)]
    API -->|Upload| Blob[Azure Blob Storage]
    API -->|Search| Search[Azure Cognitive Search]
    
    WebPubSub <-->|Messages| API
    API -->|Save History| Cosmos[(Azure Cosmos DB)]
    
    Blob -->|Blob Trigger Event| Func[Azure Functions]
    Func -->|Analyze PDF| AIDoc[Azure AI Document Intelligence]
    Func -->|Update DB| SQL
Loading

Component Architecture

  1. Frontend Layer (Client)

    • Built with React.js.
    • Communicates with the backend via standard HTTP/REST calls.
    • For real-time chat, it connects asynchronously via WebSocket to Azure Web PubSub.

  2. API Layer (Backend)

    • Built with Node.js & Express.
    • Handles authentication, request validation, business logic, and database operations.
    • Interfaces heavily with Azure SDKs to interact with various Azure resources.

  3. Data Layer

    • Azure SQL Database: The source of truth for structured data. Stores Students, Companies, Jobs, and Applications.
    • Azure Cosmos DB: Used exclusively for high-throughput saving and retrieving of chat message history.

  4. Event-Driven & AI Layer

    • Azure Functions: A serverless component that waits for new files to land in Blob Storage.
    • Azure AI Document Intelligence (Form Recognizer): An AI service called by the Azure Function to parse resume documents and pull out text and skills.

2. Core Workflows

A. AI Resume Upload and Processing Workflow

This workflow highlights the asynchronous event-driven capabilities of the platform.

  1. Upload Request: A student clicks "Upload Resume" in the React frontend.
  2. File Transfer: The frontend sends the file via a multi-part form HTTP POST request to the Node.js backend.
  3. Blob Storage: The backend takes the file buffer and securely uploads it directly to an Azure Blob Storage container (resumes).
  4. Trigger Fired: The moment the file lands in the container, it triggers a serverless Azure Function (ResumeProcessor).
  5. AI Extraction: The Azure Function takes the newly uploaded file and sends it to Azure AI Document Intelligence.
  6. Data Processing: AI Document Intelligence parses the document, pulling out contact info, education, and specific technical skills (e.g., "Python", "React", "SQL").
  7. Database Update: The Azure Function takes the parsed skills and updates the student's record in the Azure SQL Database.
  8. Completion: The student's profile now automatically reflects their skills without manual data entry.

B. Company Candidate Search Workflow

  1. Search Query: A company navigates to the search page and enters a required skill (e.g., "Node.js").
  2. Search API Call: The frontend calls the backend search endpoint (/api/companies/search?skills=Node.js).
  3. Cognitive Search: The backend formats the request and queries Azure Cognitive Search, which holds an optimized, searchable index of all student profiles and skills.
  4. Scoring: Cognitive Search instantly returns a list of candidate profiles ranked by relevance to the query.
  5. Display Results: The backend returns this list to the frontend, which renders the candidate profiles for the company to review.

C. Real-Time Chat Workflow

  1. Initialization: When a student or company opens the chat window, the frontend requests an access token from the backend.
  2. Token Generation: The backend uses the Azure SDK to generate a temporary Azure Web PubSub access token and returns it to the frontend.
  3. WebSocket Connection: The frontend uses the token to open a persistent WebSocket connection directly with the Azure Web PubSub service.
  4. Sending a Message:
    • User A types a message and hits send.
    • The message goes directly to the backend API (/api/chat/message).
  5. Persistence & Broadcast:
    • 1. Persistence: The backend immediately saves the message document into Azure Cosmos DB so the history is preserved for later.
    • 2. Broadcast: The backend uses the Web PubSub SDK to emit the message to the specific user's channel/hub.
  6. Receiving a Message: Azure Web PubSub pushes the incoming message down the open WebSocket connection to User B's frontend, instantly updating their UI without a page refresh.