afew is an initial tagging script for notmuch mail:

• http://notmuchmail.org/
• http://notmuchmail.org/initial_tagging/

Its basic task is to provide automatic tagging each time new mail is registered with notmuch. In a classic setup, you might call it after 'notmuch new' in an offlineimap post sync hook.

In addition to more elementary features such as adding tags based on email headers or maildir folders, handling killed threads and spam, it can do some heavy magic in order to /learn/ how to initially tag your mails based on their content.

In move mode, afew will move mails between maildir folders according to configurable rules that can contain arbitrary notmuch queries to match against any searchable attributes.

fyi: afew plays nicely with alot, a GUI for notmuch mail ;)

• https://github.com/pazz/alot

afew is quite young, so expect a few user visible API or configuration format changes, though I'll try to minimize such disruptive events.

Please keep an eye on NEWS.md for important news.

• text classification, magic tags aka the mailing list without server
• spam handling (flush all tags, add spam)
• killed thread handling
• tags posts to lists with lists, $list-id
• autoarchives mails sent from you
• catchall -> remove new, add inbox
• can operate on new messages [default], --all messages or on custom query results
• can move mails based on arbitrary notmuch queries, so your sorting may show on your traditional mail client (well, almost ;))
• has a --dry-run mode for safe testing
• works with python 2.7, 3.1 and 3.2

You'll need dbacl for the text classification:

# aptitude install dbacl

And I'd like to suggest to install afew as your unprivileged user. If you do, make sure ~/.local/bin is in your path.

$ python setup.py install --prefix=~/.local
$ mkdir -p ~/.config/afew ~/.local/share/afew/categories

Make sure that ~/.notmuch-config reads:

[new]
tags=new

Put a list of filters into ~/.config/afew/config:

# This is the default filter chain
[SpamFilter]
[ClassifyingFilter]
[KillThreadsFilter]
[ListMailsFilter]
[ArchiveSentMailsFilter]
[InboxFilter]

And configure rules to sort mails on your disk, if you want:

[MailMover]
folders = INBOX Junk
max_age = 15

# rules
INBOX = 'tag:spam':Junk 'NOT tag:inbox':Archive
Junk = 'NOT tag:spam AND tag:inbox':INBOX 'NOT tag:spam':Archive

$ afew --help
Usage: afew [options] [--] [query]

Options:
  -h, --help            show this help message and exit

  Actions:
    Please specify exactly one action (both update actions can be
    specified simultaniously).

    -t, --tag           run the tag filters
    -l LEARN, --learn=LEARN
                        train the category with the messages matching the
                        given query
    -u, --update        update the categories [requires no query]
    -U, --update-reference
                        update the reference category (takes quite some time)
                        [requires no query]
    -c, --classify      classify each message matching the given query (to
                        test the trained categories)
    -m, --move-mails    move mail files between maildir folders

  Query modifiers:
    Please specify either --all or --new or a query string. The default
    query for the update actions is a random selection of
    REFERENCE_SET_SIZE mails from the last REFERENCE_SET_TIMEFRAME days.

    -a, --all           operate on all messages
    -n, --new           operate on all new messages

  General options:
    -C NOTMUCH_CONFIG, --notmuch-config=NOTMUCH_CONFIG
                        path to the notmuch configuration file [default:
                        $NOTMUCH_CONFIG or ~/.notmuch-config]
    -e ENABLE_FILTERS, --enable-filters=ENABLE_FILTERS
                        filter classes to use, separated by ',' [default:
                        filters specified in afew's config]
    -d, --dry-run       don't change the db [default: False]
    -R REFERENCE_SET_SIZE, --reference-set-size=REFERENCE_SET_SIZE
                        size of the reference set [default: 1000]
    -T DAYS, --reference-set-timeframe=DAYS
                        do not use mails older than DAYS days [default: 30]
    -v, --verbose       be more verbose, can be given multiple times

Adding --dry-run to any --tag or --sync-tags action prevents modification of the notmuch db. Add some -vv goodness to see some action.

Basic tagging stuff requires no configuration, just run

$ afew --tag --new

To do this automatically you can add the following hook into your ~/.offlineimaprc:

postsynchook = ionice -c 3 chrt --idle 0 /bin/sh -c "notmuch new && afew --tag --new"

Tag filters are plugin-like modules that encapsulate tagging functionality. There is a filter that handles the archiving of mails you sent, one that handles spam, one for killed threads, one for mailing list magic...

The tag filter concept allows you to easily extend afew's tagging abilities by writing your own filters. Take a look at the default configuration file (afew/defaults/afew.config) for a list of available filters and how to enable filters and create customized filter types.

To invoke afew in move mode, provide the --move-mails option on the command line. Move mode will respect --dry-run, so throw in --verbose and watch what effects a real run would have.

In move mode, afew will check all mails (or only recent ones) in the configured maildir folders, deciding whether they should be moved to another folder.

The decision is based on rules defined in your config file. A rule is bound to a source folder and specifies a target folder into which a mail will be moved that is matched by an associated query.

This way you will be able to transfer your sorting principles roughly to the classic folder based maildir structure understood by your traditional mail server. Tag your mails with notmuch, call afew --move-mails in an offlineimap presynchook and enjoy a clean inbox in your webinterface/GUI-client at work.

For information on how to configure rules for move mode, what you can do with it and what you can't, please refer to docs/move_mode.

Let's train on an existing tag spam:

$ afew --learn spam -- tag:spam

Let's build the reference category. This is important to reduce the false positive rate. This may take a while...

$ afew --update-reference

And now let's create a new tag from an arbitrary query result:

$ afew -vv --learn sourceforge -- sourceforge

Let's see how good the classification is:

$ afew --classify -- tag:inbox and not tag:killed
Sergio López <slpml@sinrega.org> (2011-10-08) (bug-hurd inbox lists unread) --> no match
Patrick Totzke <reply+i-1840934-9a702d09342dca2b120126b26b008d0deea1731e@reply.github.com> (2011-10-08) (alot inbox lists) --> alot
[...]

As soon as you trained some categories, afew will automatically tag your new mails using the classifier. If you want to disable this feature, either use the --enable-filters option to override the default set of filters or remove the files in your afew state dir:

$ ls ~/.local/share/afew/categories
alot juggling  reference_category  sourceforge  spam

You need to update the category files periodically. I'd suggest to run

$ afew --update

on a weekly and

$ afew --update-reference

on a monthly basis.

Have fun :)