LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
"""API Webhook Notification Provider
Wraps single-event payloads (flat per-event dict) in the canonical envelope so
API webhook receivers always see the same ``{"summary": {...}, "events": [...]}``
shape. The decision is purely shape-based: a payload already in envelope shape
(it contains ``events`` — e.g. a backend buffer-rendered batch) passes through
unchanged; only a flat per-event dict is wrapped here.
"""
from typing import Any
from notification.providers.webhook_provider import WebhookProvider
from shared.infrastructure.logging import WorkerLogger
from unstract.core.notification_clubbed_renderer import build_envelope
logger = WorkerLogger.get_logger(__name__)
class APIWebhook(WebhookProvider):
"""Standard API webhook provider.
Normalises the payload to the canonical envelope before POSTing so
programmatic consumers parse one schema regardless of how the
notification was produced.
"""
def __init__(self) -> None:
"""Initialize API webhook provider."""
super().__init__()
self.provider_name = "APIWebhook"
def prepare_data(self, notification_data: dict[str, Any]) -> dict[str, Any]:
"""Prepare API webhook data.
Wraps a flat per-event payload in the canonical envelope; payloads
already in envelope shape (backend buffer-rendered) pass through.
"""
prepared_data = super().prepare_data(notification_data)
if "payload" in prepared_data:
payload = prepared_data["payload"]
if isinstance(payload, dict) and "events" not in payload:
prepared_data["payload"] = build_envelope(payloads=[payload])
logger.debug(f"Prepared API webhook data for {notification_data.get('url')}")
return prepared_data