config.sample.yaml

# General repo configuration
repo:
  bindAddress: '127.0.0.1'
  port: 8000

  # Where to store the logs, relative to where the repo is started from. Logs will be automatically
  # rotated every day and held for 14 days. To disable the repo logging to files, set this to
  # "-" (including quotation marks).
  #
  # Note: to change the log directory you'll have to restart the repository. This setting cannot be
  # live reloaded.
  logDirectory: logs

  # Set to true to enable color coding in your logs. Note that this may cause escape sequences to
  # appear in logs which render them unreadable, which is why colors are disabled by default.
  logColors: false

  # Set to true to enable JSON logging for consumption by things like logstash. Note that this is
  # incompatible with the log color option and will always render without colors.
  jsonLogs: false

  # If true, the media repo will accept any X-Forwarded-For header without validation. In most cases
  # this option should be left as "false". Note that the media repo already expects an X-Forwarded-For
  # header, but validates it to ensure the IP being given makes sense.
  trustAnyForwardedAddress: false

  # If false, the media repo will not use the X-Forwarded-Host header commonly added by reverse proxies.
  # Typically this should remain as true, though in some circumstances it may need to be disabled.
  # See https://github.com/turt2live/matrix-media-repo/issues/202 for more information.
  useForwardedHost: true

# Options for dealing with federation
federation:
  # On a per-host basis, the number of consecutive failures in calling the host before the
  # media repo will back off. This defaults to 20 if not given. Note that 404 errors from
  # the remote server do not count towards this.
  backoffAt: 20

# The database configuration for the media repository
# Do NOT put your homeserver's existing database credentials here. Create a new database and
# user instead. Using the same server is fine, just not the same username and database.
database:
  # Currently only "postgres" is supported.
  postgres: "postgres://your_username:your_password@localhost/database_name?sslmode=require"

  # The database pooling options
  pool:
    # The maximum number of connects to hold open. More of these allow for more concurrent
    # processes to happen.
    maxConnections: 25

    # The maximum number of connects to leave idle. More of these reduces the time it takes
    # to serve requests in low-traffic scenarios.
    maxIdleConnections: 5

# The configuration for the homeservers this media repository is known to control. Servers
# not listed here will not be able to upload media.
homeservers:
  - name: example.org # This should match the server_name of your homeserver, and the Host header
                      # provided to the media repo.
    csApi: "https://example.org/" # The base URL to where the homeserver can actually be reached
    backoffAt: 10 # The number of consecutive failures in calling this homeserver before the
                  # media repository will start backing off. This defaults to 10 if not given.
    adminApiKind: "matrix" # The kind of admin API the homeserver supports. If set to "matrix",
                           # the media repo will use the Synapse-defined endpoints under the
                           # unstable client-server API. When this is "synapse", the new /_synapse
                           # endpoints will be used instead. Unknown values are treated as the
                           # default, "matrix".

# Options for controlling how access tokens work with the media repo. It is recommended that if
# you are going to use these options that the `/logout` and `/logout/all` client-server endpoints
# be proxied through this process. They will also be called on the homeserver, and the response
# sent straight through the client - they are simply used to invalidate the cache faster for
# a particular user. Without these, the access tokens might still work for a short period of time
# after the user has already invalidated them.
#
# This will also cache errors from the homeserver.
#
# Note that when this config block is used outside of a per-domain config, all hosts will be
# subject to the same cache. This also means that application services on limited homeservers
# could be authorized on the wrong domain.
#
# ***************************************************************************
# *  IT IS HIGHLY RECOMMENDED TO USE PER-DOMAIN CONFIGS WITH THIS FEATURE.  *
# ***************************************************************************
accessTokens:
  # The maximum time a cached access token will be considered valid. Set to zero (the default)
  # to disable the cache and constantly hit the homeserver. This is recommended to be set to
  # 43200 (12 hours) on servers with the logout endpoints proxied through the media repo, and
  # zero for servers who do not proxy the endpoints through.
  maxCacheTimeSeconds: 0

  # Whether or not to use the `appservices` config option below. If disabled (the default),
  # the regular access token cache will be used for each user, potentially leading to high
  # memory usage.
  useLocalAppserviceConfig: false

  # The application services (and their namespaces) registered on the homeserver. Only used
  # if `useLocalAppserviceConfig` is enabled (recommended).
  #
  # Usually the appservice will provide you with these config details - they'll just need
  # translating from the appservice registration to here. Note that this does not require
  # all options from the registration, and only requires the bare minimum required to run
  # the media repo.
  appservices:
    - id: Name_of_appservice_for_your_reference
      asToken: Secret_token_for_appservices_to_use
      senderUserId: "@_example_bridge:yourdomain.com"
      userNamespaces:
        - regex: "@_example_bridge_.+:yourdomain.com"
          # A note about regexes: it is best to suffix *all* namespaces with the homeserver
          # domain users are valid for, as otherwise the appservice can use any user with
          # any domain name it feels like, even if that domain is not configured with the
          # media repo. This will lead to inaccurate reporting in the case of the media
          # repo, and potentially leading to media being considered "remote".

# These users have full access to the administrative functions of the media repository.
# See docs/admin.md for information on what these people can do. They must belong to one of the
# configured homeservers above.
admins:
  - "@your_username:example.org"

# Shared secret auth is useful for applications building on top of the media repository, such
# as a management interface. The `token` provided here is treated as a repository administrator
# when shared secret auth is enabled: if the `token` is used in place of an access token, the'
# request will be authorized. This is not limited to any particular domain, giving applications
# the ability to use it on any configured hostname.
sharedSecretAuth:
  # Set this to true to enable shared secret auth.
  enabled: false

  # Use a secure value here to prevent unauthorized access to the media repository.
  token: "PutSomeRandomSecureValueHere"

# Datastores are places where media should be persisted. This isn't dedicated for just uploads:
# thumbnails and other misc data is also stored in these places. The media repo, when looking
# for a datastore to use, will always use the smallest datastore first.
datastores:
  - type: file
    enabled: false # Enable this to set up data storage.
    # Datastores can be split into many areas when handling uploads. Media is still de-duplicated
    # across all datastores (local content which duplicates remote content will re-use the remote
    # content's location). This option is useful if your datastore is becoming very large, or if
    # you want faster storage for a particular kind of media.
    #
    # The kinds available are:
    #   thumbnails    - Used to store thumbnails of media (local and remote).
    #   remote_media  - Original copies of remote media (servers not configured by this repo).
    #   local_media   - Original uploads for local media.
    #   archives      - Archives of content (GDPR and similar requests).
    forKinds: ["thumbnails"]
    opts:
      path: /var/matrix/media

  - type: s3
    enabled: false # Enable this to set up s3 uploads
    forKinds: ["thumbnails", "remote_media", "local_media", "archives"]
    opts:
      # The s3 uploader needs a temporary location to buffer files to reduce memory usage on
      # small file uploads. If the file size is unknown, the file is written to this location
      # before being uploaded to s3 (then the file is deleted). If you aren't concerned about
      # memory usage, set this to an empty string.
      tempPath: "/tmp/mediarepo_s3_upload"
      endpoint: sfo2.digitaloceanspaces.com
      accessKeyId: ""
      accessSecret: ""
      ssl: true
      bucketName: "your-media-bucket"
      # An optional region for where this S3 endpoint is located. Typically not needed, though
      # some providers will need this (like Scaleway). Uncomment to use.
      #region: "sfo2"
      # An optional storage class for tuning how the media is stored at s3.
      # See https://aws.amazon.com/s3/storage-classes/ for details; uncomment to use.
      #storageClass: STANDARD

  # The media repo does support an IPFS datastore, but only if the IPFS feature is enabled. If
  # the feature is not enabled, this will not work. Note that IPFS support is experimental at
  # the moment and not recommended for general use.
  #
  # NOTE: Everything you upload to IPFS will be publicly accessible, even when the media repo
  # puts authentication on the download endpoints. Only use this option for cases where you
  # expect your media to be publicly accessible.
  - type: ipfs
    enabled: false # Enable this to use IPFS support
    forKinds: ["local_media"]
    # The IPFS datastore currently has no options. It will use the daemon or HTTP API configured
    # in the IPFS section of your main config.
    opts: {}

# Options for controlling archives. Archives are exports of a particular user's content for
# the purpose of GDPR or moving media to a different server.
archiving:
  # Whether archiving is enabled or not. Default enabled.
  enabled: true
  # If true, users can request a copy of their own data. By default, only repository administrators
  # can request a copy.
  # This includes the ability for homeserver admins to request a copy of their own server's
  # data, as known to the repo.
  selfService: false
  # The number of bytes to target per archive before breaking up the files. This is independent
  # of any file upload limits and will require a similar amount of memory when performing an export.
  # The file size is also a target, not a guarantee - it is possible to have files that are smaller
  # or larger than the target. This is recommended to be approximately double the size of your
  # file upload limit, provided there is enough memory available for the demand of exporting.
  targetBytesPerPart: 209715200 # 200mb default

# The file upload settings for the media repository
uploads:
  # The maximum individual file size a user can upload.
  maxBytes: 104857600 # 100MB default, 0 to disable

  # The minimum number of bytes to let people upload. This is recommended to be non-zero to
  # ensure that the "cost" of running the media repo is worthwhile - small file uploads tend
  # to waste more CPU and database resources than small files, thus a default of 100 bytes
  # is applied here as an approximate break-even point.
  minBytes: 100 # 100 bytes by default

  # The number of bytes to claim as the maximum size for uploads for the limits API. If this
  # is not provided then the maxBytes setting will be used instead. This is useful to provide
  # if the media repo's settings and the reverse proxy do not match for maximum request size.
  # This is purely for informational reasons and does not actually limit any functionality.
  # Set this to -1 to indicate that there is no limit. Zero will force the use of maxBytes.
  #reportedMaxBytes: 104857600

  # Options for limiting how much content a user can upload. Quotas are applied to content
  # associated with a user regardless of de-duplication. Quotas which affect remote servers
  # or users will not take effect. When a user exceeds their quota they will be unable to
  # upload any more media.
  quotas:
    # Whether or not quotas are enabled/enforced. Note that even when disabled the media repo
    # will track how much media a user has uploaded. This is disabled by default.
    enabled: false

    # The quota rules that affect users. The first rule to match the uploader will take effect.
    # An implied rule which matches all users and has no quota is always last in this list,
    # meaning that if no rules are supplied then users will be able to upload anything. Similarly,
    # if no rules match a user then the implied rule will match, allowing the user to have no
    # quota. The quota will let the user upload to 1 media past their quota, meaning that from
    # a statistics perspective the user might exceed their quota however only by a small amount.
    users:
      - glob: "@*:*"  # Affect all users. Use asterisks (*) to match any character.
        maxBytes: 53687063712 # 50GB default, 0 to disable

# Settings related to downloading files from the media repository
downloads:
  # The maximum number of bytes to download from other servers
  maxBytes: 104857600 # 100MB default, 0 to disable

  # The number of workers to use when downloading remote media. Raise this number if remote
  # media is downloading slowly or timing out.
  #
  # Maximum memory usage = numWorkers multiplied by the maximum download size
  # Average memory usage is dependent on how many concurrent downloads your users are doing.
  numWorkers: 10

  # How long, in minutes, to cache errors related to downloading remote media. Once this time
  # has passed, the media is able to be re-requested.
  failureCacheMinutes: 5

  # The cache control settings for downloads. This can help speed up downloads for users by
  # keeping popular media in the cache. This cache is also used for thumbnails.
  cache:
    enabled: true

    # The maximum size of cache to have. Higher numbers are better.
    maxSizeBytes: 1048576000 # 1GB default

    # The maximum file size to cache. This should normally be the same size as your maximum
    # upload size.
    maxFileSizeBytes: 104857600 # 100MB default

    # The number of minutes to track how many downloads a file gets
    trackedMinutes: 30

    # The number of downloads a file must receive in the window above (trackedMinutes) in
    # order to be cached.
    minDownloads: 5

    # The minimum amount of time an item should remain in the cache. This prevents the cache
    # from cycling out the file if it needs more room during this time. Note that the media
    # repo regularly cleans out media which is past this point from the cache, so this number
    # may need increasing depending on your use case. If the maxSizeBytes is reached for the
    # media repo, and some cached items are still under this timer, new items will not be able
    # to enter the cache. When this happens, consider raising maxSizeBytes or lowering this
    # timer.
    minCacheTimeSeconds: 300

    # The minimum amount of time an item should remain outside the cache once it is removed.
    minEvictedTimeSeconds: 60

  # How many days after a piece of remote content is downloaded before it expires. It can be
  # re-downloaded on demand, this just helps free up space in your datastore. Set to zero or
  # negative to disable. Defaults to disabled.
  expireAfterDays: 0

# URL Preview settings
urlPreviews:
  enabled: true # If enabled, the preview_url routes will be accessible
  maxPageSizeBytes: 10485760 # 10MB default, 0 to disable

  # If true, the media repository will try to provide previews for URLs with invalid or unsafe
  # certificates. If false (the default), the media repo will fail requests to said URLs.
  previewUnsafeCertificates: false

  # Note: URL previews are limited to a given number of words, which are then limited to a number
  # of characters, taking off the last word if it needs to. This also applies for the title.

  numWords: 50 # The number of words to include in a preview (maximum)
  maxLength: 200 # The maximum number of characters for a description

  numTitleWords: 30 # The maximum number of words to include in a preview's title
  maxTitleLength: 150 # The maximum number of characters for a title

  # The mime types to preview when OpenGraph previews cannot be rendered. OpenGraph previews are
  # calculated on anything matching "text/*". To have a thumbnail in the preview the URL must be
  # an image and the image's type must be allowed by the thumbnailer.
  filePreviewTypes:
    - "image/*"

  # The number of workers to use when generating url previews. Raise this number if url
  # previews are slow or timing out.
  #
  # Maximum memory usage = numWorkers multiplied by the maximum page size
  # Average memory usage is dependent on how many concurrent urls your users are previewing.
  numWorkers: 10

  # Either allowedNetworks or disallowedNetworks must be provided. If both are provided, they
  # will be merged. URL previews will be disabled if neither is supplied. Each entry must be
  # a CIDR range.
  disallowedNetworks:
    - "127.0.0.1/8"
    - "10.0.0.0/8"
    - "172.16.0.0/12"
    - "192.168.0.0/16"
    - "100.64.0.0/10"
    - "169.254.0.0/16"
    - '::1/128'
    - 'fe80::/64'
    - 'fc00::/7'
  allowedNetworks:
    - "0.0.0.0/0" # "Everything". The blacklist will help limit this.
                  # This is the default value for this field.

  # How many days after a preview is generated before it expires and is deleted. The preview
  # can be regenerated safely - this just helps free up some space in your database. Set to
  # zero or negative to disable. Defaults to disabled.
  expireAfterDays: 0

  # The default Accept-Language header to supply when generating URL previews when one isn't
  # supplied by the client.
  # Reference: https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Accept-Language
  defaultLanguage: "en-US,en"

  # Set the User-Agent header to supply when generating URL previews
  userAgent: "matrix-media-repo"

  # When true, oEmbed previews will be enabled. Typically these kinds of previews are used for
  # sites that do not support OpenGraph or page scraping, such as Twitter. For information on
  # specifying providers for oEmbed, including your own, see the following documentation:
  # https://docs.t2bot.io/matrix-media-repo/url-previews/oembed.html
  # Defaults to disabled.
  oEmbed: false

# The thumbnail configuration for the media repository.
thumbnails:
  # The maximum number of bytes an image can be before the thumbnailer refuses.
  maxSourceBytes: 10485760 # 10MB default, 0 to disable

  # The maximum number of pixels an image can have before the thumbnailer refuses. Note that
  # this only applies to image types: file types like audio and video are affected solely by
  # the maxSourceBytes.
  maxPixels: 32000000 # 32M default

  # The number of workers to use when generating thumbnails. Raise this number if thumbnails
  # are slow to generate or timing out.
  #
  # Maximum memory usage = numWorkers multiplied by the maximum image source size
  # Average memory usage is dependent on how many thumbnails are being generated by your users
  numWorkers: 100

  # All thumbnails are generated into one of the sizes listed here. The first size is used as
  # the default for when no width or height is requested. The media repository will return
  # either an exact match or the next largest size of thumbnail.
  sizes:
    - width: 32
      height: 32
    - width: 96
      height: 96
    - width: 320
      height: 240
    - width: 640
      height: 480
    - width: 768   # This size is primarily used for audio thumbnailing.
      height: 240
    - width: 800
      height: 600

  # To allow for thumbnails to be any size, not just in the sizes specified above, set this to
  # true (default false). When enabled, whatever size requested by the client will be generated
  # up to a maximum of the largest possible dimensions in the `sizes` list. For best results,
  # specify only one size in the `sizes` list when this option is enabled.
  dynamicSizing: false

  # The content types to thumbnail when requested. Types that are not supported by the media repo
  # will not be thumbnailed (adding application/json here won't work). Clients may still not request
  # thumbnails for these types - this won't make clients automatically thumbnail these file types.
  types:
    - "image/jpeg"
    - "image/jpg"
    - "image/png"
    - "image/apng"
    - "image/gif"
    - "image/heif"
    - "image/webp"
    #- "image/svg+xml" # Be sure to have ImageMagick installed to thumbnail SVG files
    - "audio/mpeg"
    - "audio/ogg"
    - "audio/wav"
    - "audio/flac"
    #- "video/mp4" # Be sure to have ffmpeg installed to thumbnail video files

  # Animated thumbnails can be CPU intensive to generate. To disable the generation of animated
  # thumbnails, set this to false. If disabled, regular thumbnails will be returned.
  allowAnimated: true

  # Default to animated thumbnails, if available
  defaultAnimated: false

  # The maximum file size to thumbnail when a capable animated thumbnail is requested. If the image
  # is larger than this, the thumbnail will be generated as a static image.
  maxAnimateSizeBytes: 10485760 # 10MB default, 0 to disable

  # On a scale of 0 (start of animation) to 1 (end of animation), where should the thumbnailer try
  # and thumbnail animated content? Defaults to 0.5 (middle of animation).
  stillFrame: 0.5

  # How many days after a thumbnail is generated before it expires and is deleted. The thumbnail
  # can be regenerated safely - this just helps free up some space in your datastores. Set to
  # zero or negative to disable. Defaults to disabled.
  expireAfterDays: 0

# Controls for the rate limit functionality
rateLimit:
  # Set this to false if rate limiting is handled at a higher level or you don't want it enabled.
  enabled: true

  # The number of requests per second before an IP will be rate limited. Must be a whole number.
  requestsPerSecond: 1

  # The number of requests an IP can send at once before the rate limit is actually considered.
  burst: 10

# Identicons are generated avatars for a given username. Some clients use these to give users a
# default avatar after signing up. Identicons are not part of the official matrix spec, therefore
# this feature is completely optional.
identicons:
  enabled: true

# The quarantine media settings.
quarantine:
  # If true, when a thumbnail of quarantined media is requested an image will be returned. If no
  # image is given in the thumbnailPath below then a generated image will be provided. This does
  # not affect regular downloads of files.
  replaceThumbnails: true

  # If true, when media which has been quarantined is requested an image will be returned. If
  # no image is given in the thumbnailPath below then a generated image will be provided. This
  # will replace media which is not an image (ie: quarantining a PDF will replace the PDF with
  # an image).
  replaceDownloads: false

  # If provided, the given image will be returned as a thumbnail for media that is quarantined.
  #thumbnailPath: "/path/to/thumbnail.png"

  # If true, administrators of the configured homeservers may quarantine media for their server
  # only. Global administrators can quarantine any media (local or remote) regardless of this
  # flag.
  allowLocalAdmins: true

# The various timeouts that the media repo will use.
timeouts:
  # The maximum amount of time the media repo should spend trying to fetch a resource that is
  # being previewed.
  urlPreviewTimeoutSeconds: 10

  # The maximum amount of time the media repo will spend making remote requests to other repos
  # or homeservers. This is primarily used to download media.
  federationTimeoutSeconds: 120

  # The maximum amount of time the media repo will spend talking to your configured homeservers.
  # This is usually used to verify a user's identity.
  clientServerTimeoutSeconds: 30

# Prometheus metrics configuration
# For an example Grafana dashboard, import the following JSON:
# https://github.com/turt2live/matrix-media-repo/blob/master/docs/grafana.json
metrics:
  # If true, the bindAddress and port below will serve GET /metrics for Prometheus to scrape.
  enabled: false

  # The address to listen on. Typically "127.0.0.1" or "0.0.0.0" for all interfaces.
  bindAddress: "127.0.0.1"

  # The port to listen on. Cannot be the same as the general web server port.
  port: 9000

# Plugins are optional pieces of the media repo used to extend the functionality offered.
# Currently there are only antispam plugins, but in future there should be more options.
# Plugins are not supported on per-domain paths and are instead repo-wide. For more
# information on writing plugins, please visit #matrix-media-repo:t2bot.io on Matrix.
plugins:
  - exec: /path/to/plugin/executable
    # Note: the exact config varies by plugin.
    config:
      hello: world

  # An example OCR plugin to block images with certain text. Note that the Docker image
  # for the media repo automatically ships this at /plugins/plugin_antispam_ocr
#  - exec: /plugins/plugin_antispam_ocr
#    config:
#      # The URL to your OCR server (https://github.com/otiai10/ocrserver)
#      ocrServer: "http://localhost:8080"
#      # The keywords to scan for. The image must contain at least one of the keywords
#      # from each list to qualify for spam.
#      keywordGroups:
#        - - elon
#          - musk
#          - elonmusk
#        - - bitcoin
#      # The minimum (and maximum) sizes of images to process.
#      minSizeBytes: 20000
#      maxSizeBytes: 200000
#      # The types of files to process
#      types: ["image/png", "image/jpeg", "image/jpg"]
#      # The user ID regex to check against
#      userIds: "@telegram_.*"
#      # How much of the image's height, starting from the top, to consider before
#      # discarding the rest. Set to 1.0 to consider the whole image.
#      percentageOfHeight: 0.35

# Options for controlling various MSCs/unstable features of the media repo
# Sections of this config might disappear or be added over time. By default all
# features are disabled in here and must be explicitly enabled to be used.
featureSupport:
  # MSC2248 - Blurhash
  MSC2448:
    # Whether or not this MSC is enabled for use in the media repo
    enabled: false

    # Maximum dimensions for converting a blurhash to an image. When no width and
    # height options are supplied, the default will be half these values.
    maxWidth: 1024
    maxHeight: 1024

    # Thumbnail size in pixels to use to generate the blurhash string
    thumbWidth: 64
    thumbHeight: 64

    # The X and Y components to use. Higher numbers blur less, lower numbers blur more.
    xComponents: 4
    yComponents: 3

    # The amount of contrast to apply when converting a blurhash to an image. Lower values
    # make the effect more subtle, larger values make it stronger.
    punch: 1

  # IPFS Support
  # This is currently experimental and might not work at all.
  IPFS:
    # Whether or not IPFS support is enabled for use in the media repo.
    enabled: false

    # Options for the built in IPFS daemon
    builtInDaemon:
      # Enable this to spawn an in-process IPFS node to use instead of a localhost
      # HTTP agent. If this is disabled, the media repo will assume you have an HTTP
      # IPFS agent running and accessible. Defaults to using a daemon (true).
      enabled: true

      # If the Daemon is enabled, set this to the location where the IPFS files should
      # be stored. If you're using Docker, this should be something like "/data/ipfs"
      # so it can be mapped to a volume.
      repoPath: "./ipfs"

# Support for redis as a cache mechanism
#
# Note: Enabling Redis support will mean that the existing cache mechanism will do nothing.
# It can be safely disabled once Redis support is enabled.
#
# See docs/redis.md for more information on how this works and how to set it up.
redis:
  # Whether or not use Redis instead of in-process caching.
  enabled: false

  # The database number to use. Leave at zero if using a dedicated Redis instance.
  databaseNumber: 0

  # The Redis shards that should be used by the media repo in the ring. The names of the
  # shards are for your reference and have no bearing on the connection, but must be unique.
  shards:
    - name: "server1"
      addr: ":7000"
    - name: "server2"
      addr: ":7001"
    - name: "server3"
      addr: ":7002"

# Optional sentry (https://sentry.io/) configuration for the media repo
sentry:
  # Whether or not to set up error reporting. Defaults to off.
  enabled: false

  # Get this value from the setup instructions in Sentry
  dsn: "https://examplePublicKey@ingest.sentry.io/0"

  # Optional environment flag. Defaults to an empty string.
  environment: ""

  # Whether or not to turn on sentry's built in debugging. This will increase log output.
  debug: false