new Tokenizer().words() removes punctuation from keywords, chat names can include them so we need to keep them.

How does this work? Rather than storing kewords, we store whole prefixes. Let's take "test chat" name as an example. SearchIndex would store only keywords prefixes:

• t, te, tes, test
• c, ch, cha, chat

Using this approach would eventually hide our typeahead tooltip when user would press the space and that's not what we want while typing chat name in input bar while mentioning. New SearchIndex entries look like this:

• t, te, tes, test, "test ", test c, ..., test chat
• c, ch, cha, chat

Not sure if handler is the right name. How about e.g. addAllPrefixes?

This can be simplified