Arabic Sign Language

FastAPI Application Structure

2 min read

FastAPI Application Structure

api fastapi architecture

The backend of the Arabic Sign Language Recognition system is built using FastAPI, a modern, high-performance web framework for building APIs with Python based on standard Python type hints.

Overview

The application is structured to serve both a RESTful API and a WebSocket interface for real-time communication. It is designed to be lightweight, asynchronous, and easily deployable via Docker.

Call Graph

graph TD
    Client[Web Client]
    App[FastAPI App]
    Lifespan[Lifespan Manager]
    Model[ONNX Model]
    Static[Static Files]
    WS[WebSocket Router]

    Client -->|HTTP GET /| App
    Client -->|WS Connect| WS
    App -->|Startup| Lifespan
    Lifespan -->|Load| Model
    App -->|Serve| Static
    App -->|Include| WS

Key Components

1. Application Lifecycle (main.py)

The application uses an asynchronous lifespan context manager to handle startup and shutdown events.

  • Startup:
    • Loads the customized ONNX Runtime inference session.
    • Initializes the model from the path specified by ONNX_CHECKPOINT_FILENAME.
    • Stores the loaded model in app.state.onnx_model for global access.
  • Shutdown:
    • Cleans up the model resources.
    • Ensures graceful termination.

2. Middleware

  • CORS (Cross-Origin Resource Sharing):
    • configured using CORSMiddleware.
    • Allowed origins are controlled via the DOMAIN_NAME environment variable.
    • Permits all methods and headers to facilitate flexible frontend interaction.

3. Routing

The application defines several types of routes:

  • Static Files:
    • Serves the frontend application (HTML, CSS, JS) from the static/ directory.
    • Mounted at /static and also serves index.html at the root /.
  • WebSocket:
    • A dedicated router is included for ws_live_signs to handle real-time inference.
  • Utility:
    • Endpoints for favicon.ico and Chrome DevTools configuration.

Graph View

Backlinks