Is there a way to write a headless Avro message to a file without deserializing its binary contents in Python?

Question

I'm using an Oracle DB to store some Avro messages of different schemas that represent some Kafka events in my application. Since there are multiple types of messages, I'm using a single table with a few columns and store their whole body in a CLOB column named BODY. I now need to export all of these messages into different files that are valid Avro messages.

When I'm fetching these data from the DB using something like:

# Return string or bytes instead of a locator (for CLOBs e.g. event.body)
oracledb.defaults.fetch_lobs = False

...

cursor.execute(
    'select body from EVENTS' +
    (f" where EVENT_TYPE='{event_type}'" if event_type else "") +
    " order by CREATED_ON"
)
for body in cursor:
    filename = fr'./exported_events/event_{SOME_UNIQUE_ID}.avro'
    os.makedirs(os.path.dirname(filename), exist_ok=True)
    with open(filename, 'wb+') as file:
        fastavro.writer(
                file,
                fastavro_schema.load_schema(fr"./event_{event_type}.avsc"),
                [body]
        )

I get an error saying basically that body is a bytes object which it cannot iterate over. This seems like I should serialize the body first before then deserializing it back in order to write it in the avro file.

Is there a way to avoid that?

PS. I have also tried converting the body to a dict using json.dumps(body) but the problem that emerges when doing this is that my message contains bytearrays and timestamp-millis types which are obviously wrongly converted by the json module and then raise errors when validating with the schema.