Personalized Fashion Designer

An AI fashion assistant that turns user style preferences into original clothing patterns and fashion sketches. Visuals are generated with Arrexel/pattern-diffusion. The system exposes an API (FastAPI) and a lightweight UI (Streamlit).

• User preferences: colors, patterns, garment types, occasions, trends, optional reference images
• Image generation: unique patterns and fashion sketches with variations
• Iteration: refine results with feedback and maintain session context
• API-first design, modular model adapters, and simple UI for quick testing

Here are some examples of fashion designs generated by the AI model:

Example 1: Fashion design with unique pattern and style

Example 2: Fashion design showcasing different color combinations

Example 3: Fashion design demonstrating pattern variations

• Python 3.10+
• A Hugging Face token with Inference API access

# From the project root
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Copy and edit environment
cp .env.example .env
# Then edit .env with your Hugging Face token and optional overrides

uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload

Open: http://localhost:8000/docs

streamlit run ui/app.py

The UI will point to the API defined in your .env.

Set these in .env:

• HF_TOKEN: Your Hugging Face token (required)
• IMAGE_MODEL_ID: Defaults to Arrexel/pattern-diffusion
• API_HOST: Defaults to http://localhost:8000
• API_PORT: Defaults to 8000
• UI_API_BASE: UI API base URL; defaults to http://localhost:8000

• app/main.py: FastAPI app and endpoints
• app/schemas/*: Pydantic request/response models
• app/services/*: Prompt builder and session management (SQLite)
• app/models/*: Model adapters for image generation and text advice (Hugging Face Inference API)
• ui/app.py: Streamlit client for interactive exploration

• The code uses the Hugging Face Inference API for image generation to avoid heavy local GPU requirements. Swap adapters if you prefer running models locally (Diffusers).
• The session store is a simple SQLite DB located at data/app.db.
• Image responses are Base64-encoded in JSON; the UI decodes and displays them.

MIT