When creating a Search index, you can't use the dot syntax to access to a field in an object inside an array. For example, take the following document into consideration:

{
   "movie": "Barbie",
   "topReviews": [
      { "score": 10, "review": "I'm just Ken" }
   ]
} 

If we want to create an index so we can search top reviews by text, it would be reasonable to try to define an index like this:

{
   "mappings": {
      "fields": {
         "topReviews.review": {
            "type": "string"
         }
      }
   }
} 

However, as review is an object inside an array of topReviews, this is not possible. We need to use the embeddedDocs type:

{
   "mappings": {
      "fields": {
         "topReviews": {
            "type": "embeddedDocument",            "fields": { "review": { "type": "string" } }
         }
      }
   }
} 

The autocomplete should be aware of the type of the document tree so it doesn't autocomplete with invalid index definitions.

More documentation:

• https://www.mongodb.com/docs/atlas/atlas-search/field-types/embedded-documents-type/#std-label-bson-data-types-embedded-documents
• https://www.mongodb.com/docs/atlas/atlas-search/embedded-document/