How to implement efficient secondary index filtering on non-key columns in TimeSeries containers?

I'm using GridDB CE v5.0 with the Java client for a time-series IoT data application. Each TimeSeries container logs data like this:

class Telemetry {
    Timestamp timestamp;
    String deviceId;
    float temperature;
    boolean isCalibrated;
}

My use case requires high-speed queries on non-key columns — specifically filtering by:

• deviceId (not the container name)

• isCalibrated = true

This must happen without full table scan, ideally using an index.

What I’ve done so far:

• I’ve created a TimeSeries container.

• Inserted ~10M rows from 1000 devices.

• I attempted to filter like this:

Query query = ts.query(
  "SELECT * WHERE deviceId = 'sensor-42' AND isCalibrated = true AND timestamp > TIMESTAMPADD(DAY, -1, NOW())"
);
RowSet<Row> rs = query.fetch();

But profiling shows full scan occurs, which leads to poor performance. I expected GridDB to optimize this using some internal index — but that doesn’t seem to happen.

Question

How can I create and utilize secondary indexes (on deviceId, isCalibrated, etc.) for efficient filtering in TimeSeries containers using Java API?