https://github.com/simonw/s3-image-proxy

A tiny proxy for serving and resizing images fetched from a private S3 bucket
https://github.com/simonw/s3-image-proxy

pillow starlette vercel

Last synced: 2 months ago
JSON representation

A tiny proxy for serving and resizing images fetched from a private S3 bucket

• Host: GitHub
• URL: https://github.com/simonw/s3-image-proxy
• Owner: simonw
• License: apache-2.0
• Created: 2020-04-28T14:56:00.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-05-21T20:21:28.000Z (over 4 years ago)
• Last Synced: 2024-10-07T11:54:11.302Z (3 months ago)
• Topics: pillow, starlette, vercel
• Language: Python
• Homepage: https://s3-image-proxy.now.sh
• Size: 15.6 KB
• Stars: 36
• Watchers: 3
• Forks: 5
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# s3-image-proxy

Tiny [Starlette](https://www.starlette.io/) application for retrieving image files from a private S3 bucket, resizing them based on querystring parameters and serving them with cache headers so the resized images can be cached by a CDN.

See [Using SQL to find my best photo of a pelican according to Apple Photos](https://simonwillison.net/2020/May/21/apple-photos-sqlite/) for background information on this project.

## Demo

You can see a demo of this proxy in action serving images [on this page](https://dogsheep-photos.dogsheep.net/public/photos_with_apple_metadata).

## Configuration

The following environment variables are required:

- `S3_BUCKET`

- `S3_AWS_ACCESS_KEY_ID`

- `S3_AWS_SECRET_ACCESS_KEY`

Here [are some notes](https://github.com/dogsheep/dogsheep-photos/issues/4) on creating an S3 bucket with the right credentials.

If you want to be able to access the original image files, set a `ORIGINAL_TOKEN` with a secret token that you plan to use to access the originals.

## Deployment

You can deploy this tool directly to [Vercel](https://vercel.com/). You'll need to set the necessary environment variables.

Vercel provides a CDN, so resized images should be served very quickly on subsequent requests to the same image.

## Local development

For local development you will need to install an additional dependency: uvicorn.

    pip install -r requirements.txt

    pip install uvicorn

You can then run the server like this:

    S3_AWS_ACCESS_KEY_ID="xxx" \

    S3_AWS_SECRET_ACCESS_KEY="yyy" \

    S3_BUCKET="your-bucket" \

    ORIGINAL_TOKEN="your-secret-token" \

    uvicorn index:app

## Usage

Once up and running, you can access image files stored in the S3 bucket like so:

    http://localhost:8000/i/name-of-file.jpeg

To resize the image, pass ?w= or ?h= arguments:

    http://localhost:8000/i/name-of-file.jpeg?w=400

    http://localhost:8000/i/name-of-file.jpeg?h=400

Use `?bw=1` to convert the image to black and white.

If you are serving JPEGs you can control the quality using `?q=` - e.g. `?q=25` for a lower quality (but faster loading) image.

To access the original image file, use this URL:

    http://localhost:8000/o/name-of-file.jpeg?token=YOUR_SECRET_TOKEN