Python — Watch the Event Stream#

Invoking gets you one result; watching gets you the ongoing facts — what lets an agent recover from a partial failure instead of trusting a single return value (Submitted Is Not Completed).

Subscribe to typed events#

subscribe_typed takes your event type (a dataclass or a Pydantic BaseModel) and yields decoded instances:

from dataclasses import dataclass
from net_sdk import NetNode
 
@dataclass
class TemperatureReading:
    sensor_id: str
    celsius: float
 
with NetNode(shards=4) as node:
    for reading in node.subscribe_typed(TemperatureReading, limit=100, timeout=5.0):
        if reading.celsius > 80:
            print(f"HOT: {reading.sensor_id} at {reading.celsius}C")

Pydantic models work the same way — pass the model class as the first argument.

Subscriptions are hot: you see events emitted after you subscribe (plus whatever is still in the ring buffer), not the whole history. There's no replay-from-the-beginning on the bus — that's a durability decision (RedEX / an adapter), covered in Durable Logs.

subscribe(limit=…) yields the raw events; subscribe_typed(T, …) decodes each into T. Both take an optional timeout so the loop ends instead of blocking forever.

Location transparency#

The same subscribe code works whether the publisher is in-process or several hops away on the mesh. The concepts are in Channels and Events and Causality.