Self-improving AI Employees for everyone
Python ai-agentsai-employeeai-employeesassistantautonomous-agentsco-employeescompanygrok-botinstinctloop-agentsmuseown-your-datapersonalself-hosted
"""Wave 12: CloudEvents factories + dispatcher for twitter.
Plugin-specific event emission for Twitter/X inbound events (mentions,
search hits, timeline updates, DMs).
Per RFC plugin_authoring_rfc.md §6.4: plugin-specific factories live in
the plugin folder.
Legacy ``event_type`` (``"twitter_event_received"``) is preserved on
the dispatch path so the ``twitterReceive`` trigger node's
``event_type`` ClassVar still matches without a coordinated
registry-side rename.
Canary status (2026-05-15):
- Typed envelope + dispatcher infrastructure: ready (this file).
- Canary registry opt-in: **deferred**. ``twitterReceive`` is a plain
``TriggerNode`` with ``mode="polling"``; the Temporal-durable
polling canary path (``PollingTriggerWorkflow`` + ``as_poll_activity``)
requires the node class to subclass :class:`PollingTriggerNode` and
declare the 4 hooks (``setup_service`` / ``fetch_ids`` /
``fetch_detail`` / ``post_emit``) — same shape as ``GmailReceiveNode``.
Once that refactor lands, add
``register_canary_trigger_type("twitterReceive")`` to
``nodes/twitter/__init__.py`` and the deployment manager picks up
the polling canary path automatically.
"""
from __future__ import annotations
from typing import Any, Mapping
from services.events.envelope import WorkflowEvent
# Legacy event_type the event_waiter dispatches by; trigger nodes
# subscribe on this string (matches ``TwitterReceiveNode.event_type``).
_LEGACY_EVENT_TYPE = "twitter_event_received"
# ---- Typed factory ---------------------------------------------------------
def twitter_event_received(event_data: Mapping[str, Any]) -> WorkflowEvent:
"""Inbound Twitter/X event envelope (mention / search hit /
timeline update / DM).
``subject`` is the tweet/DM id when available so consumers can
dedup; the originating Twitter user id is in
``data.author_id`` or ``data.user_id``. The event ``type``
distinguishes the kind via the data payload, not the envelope —
every twitter inbound shares ``com.opencompany.twitter.event.received``
since the producer (polling loop) doesn't always know which kind a
given tweet is until the downstream filter runs.
"""
payload = dict(event_data)
subject = payload.get("tweet_id") or payload.get("id") or payload.get("dm_id")
return WorkflowEvent(
source="opencompany://nodes/twitter",
type="com.opencompany.twitter.event.received",
subject=str(subject) if subject else None,
data=payload,
)
# ---- Dispatcher wrapper ----------------------------------------------------
async def dispatch_twitter_event_received(event_data: Mapping[str, Any]) -> int:
"""Dispatch an incoming Twitter/X event to waiting ``twitterReceive``
trigger nodes.
Two delivery paths (mirrors :func:`email._events.dispatch_email_received`):
1. **Legacy event_waiter waiters** via :func:`event_waiter.dispatch`
(in-process collector/processor; the default while
``twitterReceive`` is not in the canary registry).
2. **Temporal-durable listeners** via
:func:`services.events.dispatch.emit`. ``emit`` is a pass-through
no-op when ``event_framework_enabled`` is off; otherwise it
fan-outs to running consumer workflows via Visibility query.
Returns the count of legacy waiters resolved.
"""
from services import event_waiter
from services.events.dispatch import emit
payload = dict(event_data)
resolved = event_waiter.dispatch(_LEGACY_EVENT_TYPE, payload)
await emit(
twitter_event_received(payload),
wire_routing_key=_LEGACY_EVENT_TYPE,
broadcast=False,
)
return resolved
__all__ = [
"dispatch_twitter_event_received",
"twitter_event_received",
]