Can't get RegEx fields working

Sorry if this is a silly question, but I really can't seem to get RegEx fields working.

Background

I have documents like this. Basically a set of fields on the document that are associated with any (variable/dynamic) field (tenant ID in this case).

{
  ...
  "tenants": {
    "tenantA": {
      "balance": 0,
      ...
    },
    "tenantB": {
      "balance": 12.5,
      ...
    },
    "anyTenantId": {
      "balance": -3,
      ...
    },
  }
}

I want to be able to query by an exact defined (but dynamic) tenant ID. For examlpe, specifically tenant tenantAbc123 (or any other specific ID): ?filter_by=tenants.tenantAbc123.balance:>0, or ?sort_by=tenants.tenantDef456:desc.

I can't flatten the balance fields into an array, because I loose the association with the tenant ID.

What I've tried

With a schema like this:

{ "name": "tenants..*.balance", type: "float" }

I always get Cannot find a field named 'tenants.tenantAbc123.balance' in the schema. I've also tried flattening the fields like tenantAbc123_balance, but same issue.

The only thing that I got to work was { "name": "tenants", type: "object" } along with enable_nested_fields: true, but I run in to problems with type auto-detection, because it infers int64 from initial integer values and then fails when it comes accross a float value. I know exactly the types of the fields, don't need auto-detection, I just need a wildcard in the field name but can't make it work?

On a side note

The documentation about RegEx fields rather sparse, it mentions "you can use any RegEx expression to define a field name", but is that really the case? Does it support anchors and character classes, alternations etc, or is it just .* (because that's the only example provided)? How are single dots (.) treated in a field name of a normal nested field (e.g. nested_field.field1)? Do they match any character and need to be escaped?

Typesense Version: 0.24.1

Start by investigating schema handling for tenants..*.balance, nested fields with enable_nested_fields, and the type auto-detection behavior described in the issue. Done would require a clear decision on wildcard and regex semantics, support for querying dynamic tenant IDs with declared types, and documentation covering regex syntax and dots in nested field names.