Hi there! We're trying to make use of semantic search embeddings with the VueInstantSearch copmonents and have an Autocomplete component that makes the following search. The autocomplete component correctly returns semantic results, but the same configuration for the results component does not. I suspect this is due to the addition of the filter_by configuration required for faceting.

{
  "searches": [
    {
      "query_by": "name, sku, embedding",
      "collection": "products",
      "q": "*",
      "filter_by": "name:lemon",
      "page": 1,
      "per_page": 24
    }
  ]
}

and a results faceting component that makes the following search

{"searches":[{"query_by":"name, sku, embedding","vector_query":"embedding:([], k: 4, distance_threshold: 0.25)","collection":"products","q":"lemon","page":1},

but they're somehow both using the same

collectionSpecificSearchParameters

collectionSpecificSearchParameters: {
    products: {
      query_by: "name, sku, embedding",
      vector_query: "embedding:([], k: 4, distance_threshold: 0.25)"
    },
    categories: {
      query_by: "name",
    },
    recommended_content: {
      query_by: "title",
    },
    popular_queries: {
      query_by: "q",
      highlight_full_fields: '',
    },
  },

Does this imply that I can't do a text search using embeddings and can only do it in a non-filtering q:"*" scenario?

I'm not sure I fully understand the full context behind the question... But

collectionSpecificSearchParameters

will be added to any requests that are to that collection. Here's demo that uses semantic search with instantsearch.js: https://github.com/typesense/showcase-hn-comments-semantic-search/blob/c77cde41f8bc49f7dcf7e1423faba712d01616d6/src/js/index.js#L95-L110 If you can fork that repo and adapt it to match your use-case, and explain your question using that example, that will help provide more context