1# Backend
 2
 3This project provides a unified AI backend for image generation, inpainting, background removal, and captioning. It supports two modes of operation:
 4
 5## 1. Local Server (`index.py`)
 6
 7Runs on your local machine.
 8
 9> [!IMPORTANT]
10> Requires Cuda-enabled NVIDIA GPU  
11> Might require additional intervention for PyTorch installation
12
13To run backend locally, run the following commands:
14
15```shell
16pip install -r requirements.txt
17python ./index.py
18```
19
20> [!NOTE]
21> Tested on Python 3.11, compatibility with other versions is not guaranteed
22
23## 2. Cloud Server (`modal_app.py`)
24
25Deploys to [Modal.com](https://modal.com) for serverless GPU inference.
26
27To deploy, run the following commands:
28
29```shell
30pip install modal
31modal setup
32modal deploy modal_app.py
33```
34
35# Setup
36
37> [!NOTE]
38> For [Nix](https://nixos.org/) users, a [Flake](./flake.nix) for the development shell is provided
39
40## Models
41
42Download all the models from [here](https://drive.google.com/drive/folders/143A296pgkTiUGqffE4xG8ADPPEGOY-W2) and place them inside this directory
43
44## Environment Variables
45
46Ensure that a [`.env`](./.env) file is present in this directory containing the following contents:
47
48```dotenv
49FAL_KEY=<Fal.ai API Key>
50SHARED_SECRET_KEY=<Base64 Security Key>
51```
52
53> [!CAUTION]
54> You must update the root [`.env`](../.env) file for the app to recognize the deployed backend