Streaming¶

Historical replay¶

Historical replay turns a finite historical range into an async stream:

async for event in client.replay_bars(
    "Apple",
    "2024-01-02 09:30",
    "2024-01-02 16:00",
    interval="5m",
    interval_seconds=0.1,
):
    print(event)

This is useful for testing downstream applications without waiting for the market to move in real time.

Delayed polling stream¶

stream_bars repeatedly fetches a recent lookback range and emits unseen bars:

async for event in client.stream_bars(
    ["Apple", "Microsoft"],
    interval="1m",
    lookback="1d",
    poll_seconds=30,
):
    print(event)

This is not exchange-grade real-time market data. It is a free-source polling stream designed for development, prototyping, and downstream pipeline tests.

Kafka¶

Kafka support is optional:

python -m pip install -r requirements-kafka.txt

Publish a stream:

from fincore.data.kafka import KafkaSink

sink = KafkaSink("localhost:9092", "market.bars")
await sink.publish_stream(
    client.replay_bars("Apple", "2024-01-01", "2024-01-05")
)

TimescaleDB direction¶

No TimescaleDB sink is implemented yet, but event envelopes are designed so they can be mapped cleanly into hypertables later:

event_time as the time column
symbol as a partition/query dimension
event_type as a data family
payload as normalized values
raw as source traceability