Replacing antispam plugin with IMAPSieve

You will need at least pigeonhole v0.4.14 for this. If you have already configured sieve, please adjust the following to match your setup.

Caveats and possible pitfalls

  • INBOX name is case-sensitive

  • IMAP Sieve will only apply to IMAP. It will not apply to LDA or LMTP. Use Sieve normally for LDA/LMTP.

  • With this configuration, moving mails will slow down due to learn being done per email. If you want to avoid this, you need to think of something else. Probably piping things into a FIFO or perhaps using a socket based worker might work better.

  • Please read Sieve and sieve configuration to understand sieve configuration better.

  • Please read Sieve extensions and plugins for more information about sieve extensions.

  • If you run Spamassassin trough Amavis and you use a virtual users setup, you should instead configure Spamassassin to use MySQL/PostgreSQL as a backend, unless you want a headache with file permissions and lock files. You can find instructions here. In this case, the -u parameter passed to sa-learn (and the relevant sieve variables) is obsolete and can be safely removed.

  • Reloading dovecot doesn’t activate changes in this configuration, you’ll need to perform a full restart.


  • 2017/11/20 - Possibility of using spamc with SpamAssassin to mitigate multi-message delays

  • 2017/05/05 - Recommendation about Virtual Users and using an SQL Backend. Added brief info about RoundCube.

  • 2017/04/01 - Pass imap user to scripts.

  • 2017/03/19 - Added rspamd scripts and mention about sieve plugins.

  • 2017/02/13 - Improved documentation and added instructions for Spam->Trash. (Thanks for everyone who commented on mailing list)

  • 2017/02/10 - Removed imap_stats (it’s not needed).

  • 2018/04/11 - Added notes about sa-learn/spamc and warning about sieve script location.

Dovecot configuration

protocol imap {
  mail_plugins = $mail_plugins imap_sieve

plugin {
  sieve_plugins = sieve_imapsieve sieve_extprograms

  # From elsewhere to Spam folder
  imapsieve_mailbox1_name = Spam
  imapsieve_mailbox1_causes = COPY
  imapsieve_mailbox1_before = file:/usr/lib/dovecot/sieve/report-spam.sieve

  # From Spam folder to elsewhere
  imapsieve_mailbox2_name = *
  imapsieve_mailbox2_from = Spam
  imapsieve_mailbox2_causes = COPY
  imapsieve_mailbox2_before = file:/usr/lib/dovecot/sieve/report-ham.sieve

  sieve_pipe_bin_dir = /usr/lib/dovecot/sieve

  sieve_global_extensions = +vnd.dovecot.pipe +vnd.dovecot.environment

Sieve scripts


You cannot run scripts anywhere you want

Sieve allows you to only run scripts under sieve_pipe_bin_dir. You can’t use /usr/local/bin/, you have to put the script under sieve_pipe_bin_dir and use instead.

Create directory /usr/lib/dovecot/sieve and put following files to that:


require ["vnd.dovecot.pipe", "copy", "imapsieve", "environment", "variables"];

if environment :matches "imap.user" "*" {
  set "username" "${1}";

pipe :copy "" [ "${username}" ];


require ["vnd.dovecot.pipe", "copy", "imapsieve", "environment", "variables"];

if environment :matches "imap.mailbox" "*" {
  set "mailbox" "${1}";

if string "${mailbox}" "Trash" {

if environment :matches "imap.user" "*" {
  set "username" "${1}";

pipe :copy "" [ "${username}" ];

Shell scripts

For spamassassin



spamc interaction scripts are not tested yet.

# you can also use tcp/ip here, consult spamc(1)
exec /usr/bin/spamc -u ${1} -L spam -C report

# you can also use tcp/ip here, consult spamc(1)
exec /usr/bin/spamc -u ${1} -L ham -C report

You can also use sa-learn.

Note that using sa-learn often incurs significant start-up time for every message. This can cause “lockout” of the user until all the processes sequentially complete, potentially tens of seconds or minutes. If spamd is being used and the administrator is willing to accept the potential security issues of allowing unauthenticated learning of spam/ham, spamd can be envoked with the –allow-tell option and spamc with the –learntype= option. Please consult the man pages of spamd and spamc for further details.

exec /usr/bin/sa-learn -u ${1} --spam

exec /usr/bin/sa-learn -u ${1} --ham

For dspam

exec /usr/bin/dspam --client --user ${1} --class=spam --source=error

exec /usr/bin/dspam --client --user ${1} --class=innocent --source=error


CRLF handling

dspam may fail to read the mail if it contains CRLF line endings, add the Broken lineStripping option in dspam.conf if needed.

For rspamd

By default, rspamd does global learning. If you want per-user classification, or something more complex, see ` <>`__

Alternative scripts can be found from ` <>`__

exec /usr/bin/rspamc -h /run/rspamd/worker-controller.socket -P <secret> learn_spam

exec /usr/bin/rspamc -h /run/rspamd/worker-controller.socket -P <secret> learn_ham

Before running following commands, make sure dovecot.conf has all the sieve configuration you want. Then run following commands:

sievec /usr/lib/dovecot/sieve/report-spam.sieve
sievec /usr/lib/dovecot/sieve/report-ham.sieve
chmod +x /usr/lib/dovecot/sieve/ /usr/lib/dovecot/sieve/

Now your learn scripts should be invoked when you move mails between folders.


To debug, you need to import “vnd.dovecot.debug” extension. Then you can put, when required

debug_log "something"

variables are supported in this.


Recent versions of RoundCube include a markasjunk2 plugin for allowing users to mark Spam/Ham in a convenient way. Please make sure the Junk/Spam folder matches your configuration.