A minimal Flutter web frontend for a GPT-2 FastAPI backend. The frontend communicates with the backend endpoint at http://localhost:8000/generate.

• Flutter SDK installed (with web and/or desktop support enabled) and on your PATH
• Python 3.10+ with virtual environment support
• Backend running (main.py at repository root)

From the repository root:

1. Create a virtual environment and install requirements:

python -m venv venv
.\venv\Scripts\Activate.ps1
pip install -r requirements.txt

2. Start the backend server:

uvicorn main:app --reload

From the frontend folder:

flutter pub get
flutter run -d chrome

Notes:

• The backend must be reachable at http://localhost:8000. If using a different host/port, update the API URL in lib/main.dart.
• If you encounter CORS issues, enable CORS in the FastAPI backend by installing fastapi[all] or fastapi + starlette and adding CORSMiddleware in main.py.

You can build this Flutter app as a native Windows desktop application, so it runs without a browser.

• Windows 10/11
• Flutter SDK with Windows desktop support (flutter doctor)
• Visual Studio with Desktop development with C++ workload
• Python 3.10+ for running the backend during development

cd frontend
flutter build windows --release

The release executable will be created under:

build\windows\runner\Release

A helper PowerShell script (frontend/scripts/start_backend_and_app.ps1) can:

• Start the backend (via backend.exe if available, or python -m uvicorn main:app)
• Wait until http://127.0.0.1:8000/ is available
• Launch the built frontend executable

Usage example (from frontend/scripts):

.\start_backend_and_app.ps1 -BackendScript "..\..\main.py" -AppReleaseFolder "..\build\windows\runner\Release"

• Option A (development): Use the launcher script with your dev Python/venv.
• Option B (single EXE): Use PyInstaller to create backend.exe. ⚠️ Note: PyTorch + Transformers result in very large executables and may require custom hooks.
• Option C (portable folder): Ship a portable Python runtime or venv alongside the backend files. ✅ Often the simplest approach for ML-heavy backends.

• Zip the release folder along with the backend (EXE or venv), or
• Use installer builders (e.g. Inno Setup) to create a proper installer.

• Torch/Transformers packaging: Backends with PyTorch are large. Prefer portable Python + venv or host the model remotely instead of single-file EXEs.
• CORS issues: Desktop builds default to http://127.0.0.1:8000. Ensure the backend allows this origin or update the base URL in app settings.

This project is licensed under the MIT License.