Consume Directory

Docspell Documentation


The is a bash script that works in two modes:

  • Go through all files in given directories (recursively, if -r is specified) and sent each to docspell.
  • Watch one or more directories for new files and upload them to docspell.

It can watch or go through one or more directories. Files can be uploaded to multiple urls.

Run the script with the -h or --help option, to see a short help text. The help text will also show the values for any given option.

The script requires curl for uploading. It requires the inotifywait command if directories should be watched for new files.

Example for watching two directories:

./tools/ --path ~/Downloads --path ~/pdfs -m -dv http://localhost:7880/api/v1/open/upload/item/5DxhjkvWf9S-CkWqF3Kr892-WgoCspFWDo7-XBykwCyAUxQ

The script by default watches the given directories. If the -o or --once option is used, it will instead go through these directories and upload all files in there.

Example for uploading all immediatly (the same as above only with -o added):

$ -o --path ~/Downloads --path ~/pdfs/ -m -dv http://localhost:7880/api/v1/open/upload/item/5DxhjkvWf9S-CkWqF3Kr892-WgoCspFWDo7-XBykwCyAUxQ

The URL can be any docspell url that accepts uploads without authentication. This is usually a source url. It is also possible to use the script with the integration endpoint.

The script can be run multiple times and on on multiple machines, the files are transferred via HTTP to the docspell server. For example, it is convenient to set it up on your workstation, so that you can drop files into some local folder to be immediatly transferred to docspell (e.g. when downloading something from the browser).

Integration EndpointπŸ”—

When given the -i or --integration option, the script changes its behaviour slightly to work with the integration endpoint.

First, if -i is given, it implies -r – so the directories are watched or traversed recursively. The script then assumes that there is a subfolder with the collective name. Files must not be placed directly into a folder given by -p, but below a sub-directory that matches a collective name. In order to know for which collective the file is, the script uses the first subfolder.

If the endpoint is protected, the credentials can be specified as arguments --iuser and --iheader, respectively. The format is for both <name>:<value>, so the username cannot contain a colon character (but the password can).


$ -i -iheader 'Docspell-Integration:test123' -m -p ~/Downloads/ http://localhost:7880/api/v1/open/integration/item

The url is the integration endpoint url without the collective, as this is amended by the script.

This watches the folder ~/Downloads. If a file is placed in this folder directly, say ~/Downloads/test.pdf the upload will fail, because the collective cannot be determined. Create a subfolder below ~/Downloads with the name of a collective, for example ~/Downloads/family and place files somewhere below this family subfolder, like ~/Downloads/family/test.pdf.


With the -m option, the script will not upload files that already exist at docspell. For this the sha256sum command is required.

So you can move and rename files in those folders without worring about duplicates. This allows to keep your files organized using the file-system and have them mirrored into docspell as well.

Network Filesystems (samba cifs, nfs)πŸ”—

Watching a directory for changes relies on inotify subsystem on linux. This doesn't work on network filesystems like nfs or cifs. Here are some ideas to get around this limitation:

  1. The is just a shell script and doesn't need to run on the same machine as docspell. (Note that the default docker setup is mainly for demoing and quickstart, it's not required to run all of them on one machine). So the best option is to put the consumedir on the machine that contains the local filesystem. All files are send via HTTP to the docspell server anyways, so there is no need to first transfer them via a network filesystem or rsync.
  2. If option 1 is not possible for some reason, and you need to check a network filesystem, the only option left (that I know) is to periodically poll this directory. This is also possible with consumedir, using the -o or --once option (see above). You'd need to setup the systemd unit file a bit differently and add a timer to it (or you can use cron or something more fancy…).
  3. Copy the files to the machine that runs consumedir, via rsync for example. Note that this has no advantage over otpion 1, as you now need to setup rsync on the other machine to run either periodically or when some file arrives. Then you can as well run the consumedir script. But it might be more convenient, if rsync is already running.


The script can be used with systemd to run as a service. This is an example unit file:

Description=Docspell Consumedir


ExecStart=/bin/su -s /bin/bash someuser -c " --path '/a/path/' -m 'http://localhost:7880/api/v1/open/upload/item/5DxhjkvWf9S-CkWqF3Kr892-WgoCspFWDo7-XBykwCyAUxQ'"

This unit file is just an example, it needs some fiddling. It assumes an existing user someuser that is used to run this service. The url http://localhost:7880/api/v1/open/upload/... is an anonymous upload url as described here.


The provided docker-compose setup runs this script to watch a single directory, ./docs in current directory, for new files. If a new file is detected, it is pushed to docspell.

This utilizes the integration endpoint, which is enabled in the config file, to allow uploading documents for all collectives. A subfolder must be created for each registered collective. The docker containers are configured to use http-header protection for the integration endpoint. This requires you to provide a secret, that is shared between the rest-server and the script. This can be done by defining an environment variable which gets picked up by the containers defined in docker-compose.yml:

export DOCSPELL_HEADER_VALUE="my-secret"
docker-compose up

Now you can create a folder ./docs/<collective-name> and place all files in there that you want to import. Once dropped in this folder the consumedir container will push it to docspell.