Postgres jsonb advanced search

I've already posted a similar question but I have to precise a little bit...

This is the original post : [Postgres jsonb search in array with greater operator (with jsonb_array_elements)

If I sum up,

Here is the database declaration (simplified):

CREATE TABLE documents (
    document_id int4 NOT NULL GENERATED BY DEFAULT AS IDENTITY,
    data_block jsonb NULL,
    type varchar(10)
);

And this is an example of insert.

INSERT INTO documents (document_id, data_block)
VALUES(878979, 
    {"COMMONS": {"DATE": {"value": "2017-03-11"}},
     "CARS": [
         {"MODEL": {"value": "FERRARI F40"}}, 
         {"MODEL": {"value": "PORSCHE CAYENNE"}},
         {"MODEL": {"value": "FERRARI Testarossa"}}
     ]}, 'garage');
INSERT INTO documents (document_id, data_block)
VALUES(977656, 
    {"INVOICE": {"TOTAL_AMOUNT": {"value": "100.00"}},
     "PAYABLE_INVOICE_LINES": [
         {"AMOUNT": {"value": 75.00}},
         {"AMOUNT": {"value": 25.00}}
     ]}, 'invoices');
INSERT INTO documents (document_id, data_block)
VALUES(345, 
    {"INVOICE": {"TOTAL_AMOUNT": {"value": "200.00"}},
     "PAYABLE_INVOICE_LINES": [
         {"AMOUNT": {"value": 125.00}},
         {"AMOUNT": {"value": 75.00}}
     ]}, 'invoices');

In fact, I can store whatever in my JSONB column and now I want to search with specific operators.

Example of queries :

• All documents with at least one line in PAYABLE_INVOICE_LINES greater than 100.00 : data_block.PAYABLE_INVOICE_LINES.AMOUNT > 100.00

• All documents with a line in CARS starting with 'FERRARI' : data_block.CARS.MODEL like 'FERRARI%'

• All documents where the TOTAL_AMOUNT = 100.00

• All documents where the COMMONS.DATE > "2018-04-30"

• All documents where the CARS.MODEL in the list ('PORSCHE CAYENNE')

• All documents where the data_block.PAYABLE_INVOICE_LINES.AMOUNT is between 100.00 and 150.00

Bref, I want to query all I want. And I want to limit to 50 results with a sort.

My database contains millions of lines and I have some performance problem...

I've implemented the solution using a EXISTS :

select *
from documents d
where exists (
  select 1
  from jsonb_array_elements(d.data_block -> 'PAYABLE_INVOICE_LINES') as pil
  where (pil->'AMOUNT'->>'value')::decimal >= 1000)
limit 50;

It is good but not enough... Some queries are slow. > 20s.

I don't know the structure of the document. We can insert cars, invoices, identity card and the customer can create his own structure. So, I can't create index easily and I can't create function to prepare the results.

Queries are very slow when search are on the array in the JSON object. Like PAYABLE_INVOICE_LINES.

Do you have ideas to how increase performance... ?