AIFit Try-On Service

A FastAPI microservice that uses Google's Gemini API to generate virtual try-on images. Takes a user mirror selfie and an outfit reference image, then returns an edited image with the outfit applied.

Setup

1. Install dependencies (preferably in a virtual environment):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt
   

2. Set up environment variables:

   • Create a .env file in the project root (same directory as main.py)
   • Add your Gemini API key with NO spaces around the = sign:

     GEMINI_API_KEY=YOUR_REAL_KEY_HERE
     

   • Important:
     • No quotes around the value
     • No spaces: GEMINI_API_KEY=key ✅ (correct)
     • Wrong: GEMINI_API_KEY = key ❌ (has spaces)
     • Wrong: GEMINI_API_KEY="key" ❌ (has quotes)
   • The .env file should be in the same directory as docker-compose.yml

Running the Service

Option 1: Direct Python

Start the FastAPI server:

uvicorn main:app --reload

Option 2: Docker Compose

Build and run with Docker:

docker-compose up --build

The API will be available at:

• API: http://localhost:8000
• Interactive docs: http://localhost:8000/docs
• Alternative docs: http://localhost:8000/redoc
• Health check: http://localhost:8000/health

API Endpoints

GET /health

Health check endpoint for monitoring and Docker healthchecks.

Response:

{"status": "ok"}

POST /api/try-on/

Generate a try-on image by combining a user photo with an outfit reference.

Parameters:

• user_photo (file): User mirror/body photo (JPEG or PNG)
• outfit_photo (file): Collection/outfit photo (JPEG or PNG)

Response:

• Returns the generated image as image/png

Example using curl:

curl -X POST "http://localhost:8000/api/try-on/" \
  -F "user_photo=@user_selfie.jpg" \
  -F "outfit_photo=@outfit_reference.jpg" \
  --output result.png

Testing in Postman:

1. Health Check:

   • Method: GET
   • URL: http://localhost:8000/health
   • No body or headers needed
   • Expected response: {"status": "ok"}

2. Try-On Endpoint:

   • Method: POST
   • URL: http://localhost:8000/api/try-on/
   • Body tab → Select form-data
   • Add two fields:
     • user_photo (type: File) → Select your user mirror/body photo
     • outfit_photo (type: File) → Select your collection/outfit photo
   • Click Send
   • Response will be an image (PNG format)
   • To save: Click "Save Response" → "Save to a file"

Import Postman Collection: You can import postman_collection.json into Postman for pre-configured requests.

Notes

• The service uses gemini-2.5-flash-image model
• CORS is enabled for all origins (tighten in production)
• Images are validated before processing
• The service handles errors gracefully with appropriate HTTP status codes
• Docker healthcheck is configured to use the /health endpoint
• For production, remove --reload from the Dockerfile CMD and remove volume mounts from docker-compose.yml