This is part of the schema I used:

{
      "name":   "published",
      "type":   "int64",
      "index":  true
    },
    {
      "name":   "published_str",
      "type":   "string",
      "index":  false,
      "store":  true
    },
    {
      "name":   "title",
      "type":   "string"
    },
    {
      "name":   "content_ind",
      "type":   "string"
    },
    {
      "name":   "content",
      "type":   "string",
      "index":  false
    },
    {
      "name": "embedding",
      "type": "float[]",
      "embed": {
        "from": [
          "title",
          "content_ind"
        ],
        "model_config": {
          "model_name": "ts/distiluse-base-multilingual-cased-v2"
        }
      }
    }
  ]
}

fields like

published

,

published_str

(string representation of date),

title

, are quite short. There are a few other, similar short fields not shown here. About

content

and `content_ind`: almost the same text there; the only difference: formatting/presence of HTML tags; "`_ind`" means "indexed, without HTML". Why two fields: this is a workaround to show in browser the

content

(with all the HTML tags/formatting) and to index the

content_ind

without HTML tags/formatting. Also I managed to show good-looking highlighting in hybrid search. The `content`/`content_ind` field in usually quite long, like half of screen/one screen. @Kishore Nallan - How wrong is this trick with two fields? any known side effects?