mail:account:train - not enough messages to train a classifier in account with thousands of messages

[sgofferj]

Steps to reproduce

1. Run ./occ mail:account:train

Expected behavior

That might be due to a lack of easily available documentation on the ML system of mail...
I would expect the ML model to be able to learn from an account with thousands of message..

Actual behavior

www-data@615d745f46a0:~/html$ ./occ mail:account:train 2   
[info] not enough messages to train a classifier
58MB of memory used
www-data@615d745f46a0:~/html$ ./occ mail:account:diagnose 2
IMAP capabilities:
- ACL
- ANNOTATEMORE
- BINARY
- CATENATE
- CHILDREN
- COMPRESS
- CONDSTORE
- ENABLE
- ESEARCH
- ID
- IDLE
- IMAP4REV1
- LIST-EXTENDED
- LITERAL+
- LOGINDISABLED
- MAILBOX-REFERRALS
- MULTIAPPEND
- NAMESPACE
- NO_ATOMIC_RENAME
- QRESYNC
- QUOTA
- RIGHTS
- SCAN
- SORT
- THREAD
- UIDPLUS
- UNSELECT
- URLAUTH
- WITHIN
- XLIST

Account has 41588 messages in 154 mailboxes

Mail app version

3.7.0

Mailserver or service

postfix/cyrus

Operating system

Linux/docker

PHP engine version

PHP 8.2

Web server

Nginx

Database

MariaDB

Additional info

No response

[ChristophWurst]

Account has 41588 messages in 154 mailboxes

This information is taken from IMAP. The ML training takes data found in the local database cache. If the messages have not been cached yet, you may get this warning. Give the background sync some time.

[sgofferj]

Can I manually trigger sync?
Edit: never mind, found it.

Where does the ML look for training data? Only the Inbox?

[sgofferj]

Give the background sync some time.

Jep, that's it.