pict-rs/README.md

pict-rs

a simple image hosting service

Navigation

1. Links
2. Usage
   1. Running
      1. Commandline
      2. Docker
      3. Bare Metal
         1. Distro Package
         2. Binary Download
         3. Compile from Source
   2. Api
3. Administration
   1. 0.3 to 0.4 Migration Guide
   2. Filesystem to Object Storage Migration
      1. Troubleshooting
4. Development
   1. Nix Development
      1. With direnv and nix-direnv
      2. With just Nix
   2. Docker Development
      1. With Arch
      2. With Alpine
5. Contributing
6. FAQ
   1. Is pict-rs stateless?
   2. Can I use a different database?
   3. How can I submit changes?
   4. I want to configure with $format
   5. How do I donate?
7. License

Links

• Find the code on gitea
• Join the discussion on matrix
• Hit me up on mastodon

Usage

Running

Commandline

$ pict-rs -h
A simple image hosting service

Usage: pict-rs [OPTIONS] <COMMAND>

Commands:
  run            Runs the pict-rs web server
  migrate-store  Migrates from one provided media store to another
  help           Print this message or the help of the given subcommand(s)

Options:
  -c, --config-file <CONFIG_FILE>
          Path to the pict-rs configuration file
      --old-db-path <OLD_DB_PATH>
          Path to the old pict-rs sled database
      --log-format <LOG_FORMAT>
          Format of logs printed to stdout [possible values: compact, json, normal, pretty]
      --log-targets <LOG_TARGETS>
          Log levels to print to stdout, respects RUST_LOG formatting
      --console-address <CONSOLE_ADDRESS>
          Address and port to expose tokio-console metrics
      --console-buffer-capacity <CONSOLE_BUFFER_CAPACITY>
          Capacity of the console-subscriber Event Buffer
      --opentelemetry-url <OPENTELEMETRY_URL>
          URL to send OpenTelemetry metrics
      --opentelemetry-service-name <OPENTELEMETRY_SERVICE_NAME>
          Service Name to use for OpenTelemetry
      --opentelemetry-targets <OPENTELEMETRY_TARGETS>
          Log levels to use for OpenTelemetry, respects RUST_LOG formatting
      --save-to <SAVE_TO>
          File to save the current configuration for reproducible runs
  -h, --help
          Print help information
  -V, --version
          Print version information

$ pict-rs run -h
Runs the pict-rs web server

Usage: pict-rs run [OPTIONS] [COMMAND]

Commands:
  filesystem      Run pict-rs with filesystem storage
  object-storage  Run pict-rs with object storage
  help            Print this message or the help of the given subcommand(s)

Options:
  -a, --address <ADDRESS>
          The address and port to bind the pict-rs web server
      --api-key <API_KEY>
          The API KEY required to access restricted routes
      --worker-id <WORKER_ID>
          ID of this pict-rs node. Doesn't do much yet
      --media-preprocess-steps <MEDIA_PREPROCESS_STEPS>
          Optional pre-processing steps for uploaded media
      --media-skip-validate-imports <MEDIA_SKIP_VALIDATE_IMPORTS>
          Whether to validate media on the "import" endpoint [possible values: true, false]
      --media-max-width <MEDIA_MAX_WIDTH>
          The maximum width, in pixels, for uploaded media
      --media-max-height <MEDIA_MAX_HEIGHT>
          The maximum height, in pixels, for uploaded media
      --media-max-area <MEDIA_MAX_AREA>
          The maximum area, in pixels, for uploaded media
      --media-max-file-size <MEDIA_MAX_FILE_SIZE>
          The maximum size, in megabytes, for uploaded media
      --media-max-frame-count <MEDIA_MAX_FRAME_COUNT>
          The maximum number of frames allowed for uploaded GIF and MP4s
      --media-enable-silent-video <MEDIA_ENABLE_SILENT_VIDEO>
          Whether to enable GIF and silent video uploads [possible values: true, false]
      --media-enable-full-video <MEDIA_ENABLE_FULL_VIDEO>
          Whether to enable full video uploads [possible values: true, false]
      --media-video-codec <MEDIA_VIDEO_CODEC>
          Enforce a specific video codec for uploaded videos [possible values: h264, h265, av1, vp8, vp9]
      --media-audio-codec <MEDIA_AUDIO_CODEC>
          Enforce a specific audio codec for uploaded videos [possible values: aac, opus, vorbis]
      --media-filters <MEDIA_FILTERS>
          Which media filters should be enabled on the `process` endpoint
      --media-format <MEDIA_FORMAT>
          Enforce uploaded media is transcoded to the provided format [possible values: avif, jpeg, jxl, png, webp]
  -h, --help
          Print help information (use `--help` for more detail)

Try running help commands for more runtime configuration options

$ pict-rs run filesystem -h
$ pict-rs run object-storage -h
$ pict-rs run filesystem sled -h
$ pict-rs run object-storage sled -h

See pict-rs.toml for more configuration

Example:

Run with the default configuration

$ ./pict-rs run

Running on all interfaces, port 8080, storing data in /opt/data

$ ./pict-rs \
    run -a 0.0.0.0:8080 \
    filesystem -p /opt/data/files \
    sled -p /opt/data/sled-repo

Running locally, port 9000, storing data in data/, and converting all uploads to PNG

$ ./pict-rs \
    run \
        -a 127.0.0.1:9000 \
        --media-format png \
    filesystem -p data/files \
    sled -p data/sled-repo

Running locally, port 8080, storing data in data/, and only allowing the thumbnail and identity filters

$ ./pict-rs \
    run \
        -a 127.0.0.1:8080 \
        --media-filters thumbnail \
        --media-filters identity \
    filesystem -p data/files \
    sled -p data/sled-repo

Running from a configuration file

$ ./pict-rs -c ./pict-rs.toml run

Migrating to object storage from filesystem storage. For more detailed info, see Filesystem to Object Storage Migration

$ ./pict-rs \
    migrate-store \
    filesystem -p data/files \
    object-storage \
        -a ACCESS_KEY \
        -b BUCKET_NAME \
        -r REGION \
        -s SECRET_KEY

Dumping configuration overrides to a toml file

$ ./pict-rs --save-to pict-rs.toml \
    run \
    object-storage \
        -a ACCESS_KEY \
        -b pict-rs \
        -r us-east-1 \
        -s SECRET_KEY \
    sled -p data/sled-repo

Docker

Run the following commands:

# Create a folder for the files (anywhere works)
$ mkdir ./pict-rs
$ cd ./pict-rs
$ mkdir -p volumes/pictrs
$ sudo chown -R 991:991 volumes/pictrs
$ wget https://git.asonix.dog/asonix/pict-rs/raw/branch/main/docker/prod/docker-compose.yml
$ sudo docker-compose up -d

Note

• pict-rs makes use of the system's temporary folder. This is generally /tmp on linux
• pict-rs makes use of an imagemagick security policy at /usr/lib/ImageMagick-$VERSION/config-Q16HDRI/policy.xml

Bare Metal

There are a few options for acquiring pict-rs to run outside of docker.

1. Packaged via your distro of choice
2. Binary download from the releases page
3. Compiled from source

If running outside of docker, the recommended configuration method is via the pict-rs.toml file. When running pict-rs, the file can be passed to the binary as a commandline argument.

$ pict-rs -c /path/to/pict-rs.toml run

Distro Package

If getting pict-rs from your distro, please make sure it's a recent version (meaning 0.3.x stable, or 0.4.0-rc.x). If it is older, consider using an alternative option for installing pict-rs. I am currently aware of pict-rs packaged in the AUR and nixpkgs, but there may be other distros that package it as well.

Binary Download

pict-rs provides precompiled binaries that should work on any linux system for x86_64, aarch64, and armv7h on the releases page. If downloading a binary, make sure that you have the following dependencies installed:

• imagemagick 7
• ffmpeg 5 or 6
• exiftool 12 (sometimes called perl-image-exiftool)

These binaries are called by pict-rs to process uploaded media, so they must be in the $PATH available to pict-rs.

Compile from Source

pict-rs can be compiled from source using a recent version of the rust compiler. I do development on 1.69 and produce releases on 1.70. pict-rs also requires the protoc protobuf compiler to be present at build-time in order to enable use of tokio-console.

Like the Binary Download option, imagemagick, ffmpeg, and exiftool must be installed for pict-rs to run properly.

API

pict-rs offers the following endpoints:

• POST /image for uploading an image. Uploaded content must be valid multipart/form-data with an image array located within the images[] key

  This endpoint returns the following JSON structure on success with a 201 Created status

  {
      "files": [
          {
              "delete_token": "JFvFhqJA98",
              "file": "lkWZDRvugm.jpg",
              "details": {
                  "width": 800,
                  "height": 800,
                  "content_type": "image/jpeg",
                  "created_at": "2022-04-08T18:33:42.957791698Z"
              }
          },
          {
              "delete_token": "kAYy9nk2WK",
              "file": "8qFS0QooAn.jpg",
              "details": {
                  "width": 400,
                  "height": 400,
                  "content_type": "image/jpeg",
                  "created_at": "2022-04-08T18:33:42.957791698Z"
              }
          },
          {
              "delete_token": "OxRpM3sf0Y",
              "file": "1hJaYfGE01.jpg",
              "details": {
                  "width": 400,
                  "height": 400,
                  "content_type": "image/jpeg",
                  "created_at": "2022-04-08T18:33:42.957791698Z"
              }
          }
      ],
      "msg": "ok"
  }
  

• POST /image/backgrounded Upload an image, like the /image endpoint, but don't wait to validate and process it. This endpoint returns the following JSON structure on success with a 202 Accepted status

  {
      "uploads": [
          {
              "upload_id": "c61422e1-9294-4f1f-977f-c696b7939467",
          },
          {
              "upload_id": "62cc707f-725c-44b6-908f-2bd8946c3c29"
          }
      ],
      "msg": "ok"
  }
  

• GET /image/download?url={url}&backgrounded=(true|false) Download an image from a remote server, returning the same JSON payload as the POST /image endpoint by default.

  if backgrounded is set to true, then the ingest processing will be queued for later and the response json will be the same as the POST /image/backgrounded endpoint.

• GET /image/backgrounded/claim?upload_id={uuid} Wait for a backgrounded upload to complete, claiming it's result Possible results:

  • 200 Ok (validation and ingest complete):

    {
        "files": [
            {
                "delete_token": "OxRpM3sf0Y",
                "file": "1hJaYfGE01.jpg",
                "details": {
                    "width": 400,
                    "height": 400,
                    "content_type": "image/jpeg",
                    "created_at": "2022-04-08T18:33:42.957791698Z"
                }
            }
        ],
        "msg": "ok"
    }
    

  • 422 Unprocessable Entity (validation or otherwise failure):

    {
        "msg": "Error message about what went wrong with upload"
    }
    

  • 204 No Content (Upload validation and ingest is not complete, and waiting timed out) In this case, trying again is fine

• GET /image/original/{file} for getting a full-resolution image. file here is the file key from the /image endpoint's JSON

• GET /image/details/original/{file} for getting the details of a full-resolution image. The returned JSON is structured like so:

  {
      "width": 800,
      "height": 537,
      "content_type": "image/webp",
      "created_at": "2022-04-08T18:33:42.957791698Z"
  }
  

• GET /image/process.{ext}?src={file}&... get a file with transformations applied. existing transformations include

  • identity=true: apply no changes

  • blur={float}: apply a gaussian blur to the file

  • thumbnail={int}: produce a thumbnail of the image fitting inside an {int} by {int} square using raw pixel sampling

  • resize={int}: produce a thumbnail of the image fitting inside an {int} by {int} square using a Lanczos2 filter. This is slower than sampling but looks a bit better in some cases

  • resize={filter}.(a){int}: produce a thumbnail of the image fitting inside an {int} by {int} square, or when (a) is present, produce a thumbnail whose area is smaller than {int}. {filter} is optional, and indicates what filter to use when resizing the image. Available filters are Lanczos, Lanczos2, LanczosSharp, Lanczos2Sharp, Mitchell, and RobidouxSharp.

    Examples:

    • resize=300: Produce an image fitting inside a 300x300 px square
    • reizie=.a10000: Produce an image whose area is at most 10000 px
    • resize=Mitchell.200: Produce an image fitting inside a 200x200 px square using the Mitchell filter
    • resize=RobidouxSharp.a40000: Produce an image whose area is at most 40000 px using the RobidouxSharp filter

  • crop={int-w}x{int-h}: produce a cropped version of the image with an {int-w} by {int-h} aspect ratio. The resulting crop will be centered on the image. Either the width or height of the image will remain full-size, depending on the image's aspect ratio and the requested aspect ratio. For example, a 1600x900 image cropped with a 1x1 aspect ratio will become 900x900. A 1600x1100 image cropped with a 16x9 aspect ratio will become 1600x900.

  Supported ext file extensions include avif, jpg, jxl, png, and webp

  An example of usage could be

  GET /image/process.jpg?src=asdf.png&thumbnail=256&blur=3.0
  

  which would create a 256x256px JPEG thumbnail and blur it

• GET /image/process_backgrounded.{ext}?src={file}&... queue transformations to be applied to a given file. This accepts the same arguments as the process.{ext} endpoint, but does not wait for the processing to complete.

• GET /image/details/process.{ext}?src={file}&... for getting the details of a processed image. The returned JSON is the same format as listed for the full-resolution details endpoint.

• DELETE /image/delete/{delete_token}/{file} or GET /image/delete/{delete_token}/{file} to delete a file, where delete_token and file are from the /image endpoint's JSON

The following endpoints are protected by an API key via the X-Api-Token header, and are disabled unless the --api-key option is passed to the binary or the PICTRS__SERVER__API_KEY environment variable is set.

A secure API key can be generated by any password generator.

• POST /internal/import for uploading an image while preserving the filename as the first alias. The upload format and response format are the same as the POST /image endpoint.

• POST /internal/purge?alias={alias} Purge a file by it's alias. This removes all aliases and files associated with the query.

  This endpoint returns the following JSON

  {
      "msg": "ok",
      "aliases": ["asdf.png"]
  }
  

• GET /internal/aliases?alias={alias} Get the aliases for a file by it's alias

  This endpiont returns the same JSON as the purge endpoint

• DELETE /internal/variants Queue a cleanup for generated variants of uploaded images.

  If any of the cleaned variants are fetched again, they will be re-generated.

• GET /internal/identifier Get the image identifier (file path or object path) for a given alias

  On success, the returned json should look like this:

  {
      "msg": "ok",
      "identifier": "/path/to/object"
  }
  

Additionally, all endpoints support setting deadlines, after which the request will cease processing. To enable deadlines for your requests, you can set the X-Request-Deadline header to an i128 value representing the number of nanoseconds since the UNIX Epoch. A simple way to calculate this value is to use the time crate's OffsetDateTime::unix_timestamp_nanos method. For example,

// set deadline of 1ms
let deadline = time::OffsetDateTime::now_utc() + time::Duration::new(0, 1_000);

let request = client
    .get("http://pict-rs:8080/image/details/original/asdfghjkla.png")
    .insert_header(("X-Request-Deadline", deadline.unix_timestamp_nanos().to_string())))
    .send()
    .await;

Administration

0.3 to 0.4 Migration Guide

pict-rs will automatically migrate from the 0.3 db format to the 0.4 db format on the first launch of 0.4. If you are running the provided docker container without any custom configuration, there are no additional steps.

If you have any custom configuration for file paths, or you are running outside of docker, then there is some extra configuration that needs to be done.

If your previous PICTRS__PATH variable or path config was set, it needs to be translated to the new configuration format.

PICTRS_PATH has split into three separate config options:

• PICTRS__OLD_DB__PATH: This should be set to the same value that PICTRS__PATH was. It is used during the migration from 0.3 to 0.4
• PICTRS__REPO__PATH: This is the location of the 0.4 database. It should be set to a subdirectory of the previous PICTRS__PATH directory. I would recommend /previous/path/sled-repo
• PICTRS__STORE__PATH: This is the location of the files. It should be the files subdirectory of the previous PICTRS__PATH directory.

if you configured via the configuration file, these would be

[old_db]
path = "/previous/path"

[repo]
path = "/previous/path/sled-repo"

[store]
path = "/previous/path/files"

If the migration doesn't work due to a configuration error, the new sled-repo directory can be deleted and a new migration will be automatically triggered on the next launch.

Filesystem to Object Storage migration

After migrating from 0.3 to 0.4, it is possible to migrate to object storage. This can be useful if hosting in a cloud environment, since object storage is generally far cheaper than block storage.

The command will look something like this:

$ pict-rs \
    migrate-store \
    filesystem \
        -p /path/to/files \
    object-storage \
        -e https://object-storage-endpoint \
        -b bucket-name \
        -r region \
        -a access-key \
        -s secret-key \
    sled \
        -p /path/to/sled-repo

If you are running the docker container with default paths, it can be simplified to the following:

$ pict-rs \
    migrate-store \
    filesystem \
    object-storage \
        -e https://object-storage-endpoint \
        -b bucket-name \
        -r region \
        -a access-key \
        -s secret-key

This command must be run while pict-rs is offline.

After you've completed the migration, update your pict-rs configuration to use object storage. If you configure using environment variables, make sure the following are set:

• PICTRS__STORE__TYPE=object_storage
• PICTRS__STORE__ENDPOINT=https://object-storage-endpoint
• PICTRS__STORE__BUCKET_NAME=bucket-name
• PICTRS__STORE__REGION=region
• PICTRS__STORE__USE_PATH_STYLE=false (set to true if your object storage requires path style access)
• PICTRS__STORE__ACCESS_KEY=access-key
• PICTRS__STORE__SECRET_KEY=secret-key

If you use the configuration file, this would be

[store]
type = "object_storage"
endpoint = "https://object-storage-endpoint"
bucket_name = "bucket-name"
region = "region"
use_path_style = false # Set to true if your object storage requires path style access
access_key = "access-key"
secret_key = "secret-key"

Migration Troubleshooting

If you have enabled object storage without first migrating your existing files to object storage, these migrate commands may end up retrying file migrations indefinitely. In order to successfully resolve this multi-store problem the --skip-missing-files flag has been added to the migrate-store subcommand. This tells pict-rs not to retry migrating a file if that file returns some form of "not found" error.

$ pict-rs
    migrate-store --skip-missing-files
    filesystem -p /path/to/files
    object-storage \
        -e https://object-storage-endpoint \
        -b bucket-name \
        -r region \
        -a access-key \
        -s secret-key
    sled \
        -p /path/to/sled-repo

Development

pict-rs has a few native dependencies that need to be installed in order for it to run properly. Currently these are as follows:

• imagemagick 7.1.1 (although 7.0 and 7.1.0 may also work)
• ffmpeg 6 (although 5 may also work)
• exiftool 12.62 (although 12.50 or newer may also work)

Additionally, pict-rs requires a protobuf compiler during the compilation step to support tokio-console, a runtime debug tool.

Installing these from your favorite package manager should be sufficient. Below are some fun ways to develop and test a pict-rs binary.

Nix Development

I personally use nix for development. The provided flake.nix file should be sufficient to create a development environment for pict-rs on any linux distribution, provided nix is installed.

With direnv and nix-direnv

With these tools, the pict-rs development environment can be automatically loaded when entering the pict-rs directory

Setup (only once):

$ echo 'use flake' > .envrc
$ direnv allow

Running:

$ cargo run -- -c dev.toml run

With just Nix

$ nix develop
$ cargo run -- -c dev.toml run

Docker Development

Previously, I have run pict-rs from inside containers that include the correct dependencies. The two options listed below are ones I have personally tried.

With Arch

This option doesn't take much configuration, just compile the binary and run it from inside the container

$ cargo build
$ sudo docker run --rm -it -p 8080:8080 -v "$(pwd):/mnt" archlinux:latest
> pacman -Syu imagemagick ffmepg perl-image-exiftool
> cp /mnt/docker/prod/root/usr/lib/ImageMagick-7.1.1/config-Q16HDRI/policy.xml /usr/lib/ImageMagick-7.1.1/config-Q16HDRI/
> PATH=$PATH:/usr/bin/vendor_perl /mnt/target/debug/pict-rs --log-targets debug run

With Alpine

This option requires cargo-zigbuild to be installed. Cargo Zigbuild is a tool that links rust binaries with Zig's linker, enabling easy cross-compiles to many targets. Zig has put a lot of effort into seamless cross-compiling, and it is nice to be able to take advantage of that work from rust.

$ cargo zigbuild --target=x86_64-unknown-linux-musl
$ sudo docker run --rm -it -p 8080:8080 -v "$(pwd):/mnt" alpine:3.18
> apk add imagemagick ffmpeg exiftool
> cp /mnt/docker/prod/root/usr/lib/ImageMagick-7.1.1/config-Q16HDRI/policy.xml /usr/lib/ImageMagick-7.1.1/config-Q16HDRI/
> /mnt/target/x86_64-unknown-linux-musl/debug/pict-rs --log-targets debug run

Contributing

Feel free to open issues for anything you find an issue with. Please note that any contributed code will be licensed under the AGPLv3.

FAQ

Question: Is pict-rs stateless

Answer: No. pict-rs relies on an embedded key-value store called sled to store metadata about uploaded media. This database maintains a set of files on the local disk and cannot be configured to use a network.

Question: Can I use a different database with pict-rs

Answer: No. Currently pict-rs only supports the embedded key-value store called sled. In the future, I would like to support Postgres and BonsaiDB, but I am currently not offering a timeline on support. If you care about this and are a rust developer, I would accept changes.

Question: How can I submit changes

Answer: If you would like to contribute to pict-rs, you can push your code to a public git host of your choice and let me know you did so via matrix or email. I can pull and merge your changes into this repository from there.

Alternatively, you are welcome to email me a patch that I can apply.

I will not be creating additional accounts on my forgejo server, sorry not sorry.

Question: I want to configure it with yaml instead of toml

Answer: That's not a question, but you can configure pict-rs with json, hjson, yaml, ini, or toml. Writing configs in other formats is left as an exercise to the reader.

Question: How do I donate to pict-rs

Answer: You don't. I get paid by having a job where I do other stuff. Don't give me money that I don't need.

License

Copyright © 2022 Riley Trautman

pict-rs is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

pict-rs is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. This file is part of pict-rs.

You should have received a copy of the GNU General Public License along with pict-rs. If not, see http://www.gnu.org/licenses/.