Have a look at what your text is transformed at search time:
POST docs/_analyze
{
"analyzer": "autocomplete_search",
"text": [ "123456789й" ]
}
{
"tokens": [
{
"token": "й",
"start_offset": 9,
"end_offset": 10,
"type": "word",
"position": 0
}
]
}
Have a look at the Doc.