Match by whitespace indentation in textmate language grammar - regex

Question

I have a DSL that I'm trying to write a textmate grammar for, so I can have syntax highlighting (in VS Code rather than textmate, but it uses their grammar)
The DSL represent tree nodes, is whitespace indented, and gets mapped to JSON
Most of it is pretty straightforward, but it has some weird constructs that I'm not sure how to match with the textmate grammar. An example of the DSL is below.
I can match the "tags" and simple scalar type "properties" and the single line "arrays" just fine, but I'm a bit lost on the multiline arrays and objects - I don't know how to stop matching their contents when the indentation changes. I thought I could just stop matching when I hit a known pattern, but its ambiguous between a single item on the last line of the array and a "tag" apart from their indentation level
Is there some way to match these based on indentation alone?
Example DSL file:
fooTag
foo An unquoted string
bar 42
baz "42"
qux true
barTag
qux{}
foo A string called foo on the qux object
bar null
foo[] in arrays each space separated token is a separate string "unless quoted"
bar[]
some arrays of space separated tokens look like this when it is a long
list with lots of items
baz[]
arrays can be mixed types like this one which contains the number 42 and
the constants true false and null and the rest of the tokens are all
strings and even nested arrays and objects as follows
[] 1 2 3
[]
indented array of tokens
{}
foo 42
bar "42"
some
more
tags
Example JSON output:
[
{
"name": "fooTag",
"model": {
"foo": "An unquoted string",
"bar": 42,
"baz": "42",
"qux": true
}
},
[
{
"name": "barTag",
"model": {
"qux": {
"foo": "A string called foo on the qux object",
"bar": null
},
"foo": [
"each", "space", "separated", "token", "is", "a", "separate",
"string", "unless quoted"
],
"bar": [
"some", "arrays", "of", "space", "separated", "tokens", "look",
"like", "this", "when", "it", "is", "a", "long", "list", "with",
"lots", "of", "items"
],
"baz": [
"arrays", "can", "be", "mixed", "types", "like", "this", "one",
"which", "contains", "the", "number", 42 "and", "the", "constants",
true, false, "and", null, "and", "the", "rest", "of", "the", "tokens",
"are", "all", "strings", "and", "even", "nested", "arrays", "and",
"objects", "as", "follows",
[ 1, 2, 3 ],
[ "indented", "array", "of", "tokens" ],
{
"foo": 42,
"bar": "42"
}
]
}
},
[
{
"name": "some",
"model": {}
}
],
[
{
"name": "more",
"model": {}
}
],
[
{
"name": "tags",
"model": {}
}
]
]
]