How to use a Compound Indexes

We have a collection online，The information is as follows：

curl -X GET 'http://localhost:6333/collections/questions_embeddings_v01_orionstar_text-embedding-bge_39' \
-H 'Content-Type: application/json'
{
    "result": {
        "status": "green", 
        "optimizer_status": "ok", 
        "vectors_count": 461279, 
        "indexed_vectors_count": 460177, 
        "points_count": 461279, 
        "segments_count": 3, 
        "config": {
            "params": {
                "vectors": {
                    "size": 1024, 
                    "distance": "Cosine"
                }, 
                "shard_number": 1, 
                "replication_factor": 1, 
                "write_consistency_factor": 1, 
                "on_disk_payload": true
            }, 
            "hnsw_config": {
                "m": 0, 
                "ef_construct": 100, 
                "full_scan_threshold": 10000, 
                "max_indexing_threads": 0, 
                "on_disk": false, 
                "payload_m": 16
            }, 
            "optimizer_config": {
                "deleted_threshold": 0.2, 
                "vacuum_min_vector_number": 1000, 
                "default_segment_number": 0, 
                "max_segment_size": null, 
                "memmap_threshold": null, 
                "indexing_threshold": 20000, 
                "flush_interval_sec": 5, 
                "max_optimization_threads": 1
            }, 
            "wal_config": {
                "wal_capacity_mb": 32, 
                "wal_segments_ahead": 0
            }, 
            "quantization_config": null
        }, 
        "payload_schema": {
            "metadata.data_id": {
                "data_type": "keyword", 
                "points": 461279
            }
        }
    }, 
    "status": "ok", 
    "time": 0.000023419
}

He has a payload index "metadata.data_id"，But when we use the following query, the query time is very long，About three seconds：

curl --location 'http://10.118.13.232:6333/collections/questions_embeddings_v01_orionstar_text-embedding-bge_39/points/scroll' \
--header 'Content-Type: application/json' \
--data '{
    "limit": 4,
    "with_payload": true,
    "with_vectors": true,
"query_vector": [0.016785502, ......]
    "filter": {
        "must": [
            {
                "key": "metadata.source_type",
                "match": {
                    "value": "intervention"
                }
            },
            {
                "key": "metadata.data_id",
                "match": {
                    "value": "7bd65f858d2e3144a193b9a3f2db1cdd"
                }
            },
            {
                "key": "metadata.is_using",
                "match": {
                    "value": 1
                }
            }
        ]
    }
}'

So I tried to find a way to add a Compound index, but I couldn't find it，and I tried to add an index for the "metadata.source_type" field，Below is the result of my addition

The same query is executed again, and the discovery takes only a few milliseconds

My problem is to add two separate indexes like the following：

curl -X PUT 'http://localhost:6333/collections/questions_embeddings_v01_orionstar_text-embedding-bge_39'/index \
-H 'Content-Type: application/json' \
--data '{
"field_name": "metadata.data_id",
"field_schema": "keyword"
}'

curl -X PUT 'http://localhost:6333/collections/questions_embeddings_v01_orionstar_text-embedding-bge_39'/index \
-H 'Content-Type: application/json' \
--data '{
"field_name": "metadata.source_type",
"field_schema": "keyword"
}'

Is that equivalent to
alter table add index idx_data_id_source_type(metadata.data_id,metadata.source_type); in mysql？

If not, what makes the same query faster？

Start with the collection information and points/scroll requests in the issue, comparing the payload indexes on metadata.data_id and metadata.source_type with the three-field filter. Clarify whether separate indexes provide compound-index behavior and what accounts for the observed latency difference; document the supported indexing behavior and its limits.