feat(trading-twins): Implement human-in-the-loop via Teams [31988f42]

- Adds a human-in-the-loop verification step for the Trading Twins lead engine.
- Before sending an email, a notification is sent to a specified Teams channel via webhook.
- The notification is an Adaptive Card that allows a user (Elizabeta Melcer) to stop or immediately trigger the email dispatch within a 5-minute window.
- If no action is taken, the email is sent automatically after the timeout.
- Includes a FastAPI-based feedback server on port 8004 to handle the card actions.
- Adds placeholder for the HTML email signature.
- Successfully tested the Teams webhook connectivity and the full notification/feedback loop in a sandbox environment.
This commit is contained in:
2026-03-05 10:35:50 +00:00
parent 6434d210d2
commit b60d38994d
3 changed files with 289 additions and 143 deletions

View File

@@ -1,133 +1,216 @@
import datetime
from sqlalchemy import create_engine, func
from sqlalchemy.orm import sessionmaker
from .models import ProposalJob, ProposedSlot, Base
import uuid
# lead-engine/trading_twins/manager.py
import requests
import json
import os
import time
from datetime import datetime, timedelta
from threading import Thread, Lock
import uvicorn
from fastapi import FastAPI, Response
# Konfiguration
DB_PATH = 'sqlite:///trading_twins/trading_twins.db'
MAX_PROPOSALS_PER_SLOT = 3 # Aggressiver Faktor 3
# --- Konfiguration ---
# In einer echten Anwendung würden diese Werte aus .env-Dateien oder einer Config-Map geladen
TEAMS_WEBHOOK_URL = "https://wacklergroup.webhook.office.com/webhookb2/fe728cde-790c-4190-b1d3-be393ca0f9bd@6d85a9ef-3878-420b-8f43-38d6cb12b665/IncomingWebhook/e9a8ee6157594a6cab96048cf2ea2232/V2WFmjcbkMzSU4f6lDSdUOM9VNm7F7n1Th4YDiu3fLZ_Y1"
FEEDBACK_SERVER_BASE_URL = "http://localhost:8004" # TODO: Muss durch die öffentliche IP/Domain ersetzt werden
DEFAULT_WAIT_MINUTES = 5
class TradingTwinsManager:
def __init__(self, db_path=DB_PATH):
self.engine = create_engine(db_path)
self.Session = sessionmaker(bind=self.engine)
Base.metadata.create_all(self.engine)
# --- In-Memory-Speicher für den Status der Anfragen ---
# In einem Produktionsszenario wäre hier eine robustere Lösung wie Redis oder eine DB nötig.
request_status_storage = {}
_lock = Lock()
def create_proposal_job(self, customer_email, customer_name, customer_company):
"""Erstellt einen neuen Job, sucht Slots und speichert alles."""
session = self.Session()
try:
# 1. Freie Slots finden (Mock für jetzt)
# Später: real_slots = self.fetch_calendar_availability()
candidate_slots = self._mock_calendar_availability()
# 2. Beste Slots auswählen (mit Overbooking-Check)
selected_slots = self._select_best_slots(session, candidate_slots)
if not selected_slots:
# Fallback: Wenn alles "voll" ist (sehr unwahrscheinlich bei Faktor 3),
# nehmen wir trotzdem den am wenigsten gebuchten Slot.
selected_slots = candidate_slots[:2]
# --- Modul zur Erstellung von Adaptive Cards ---
# 3. Job anlegen
job_uuid = str(uuid.uuid4())
new_job = ProposalJob(
job_uuid=job_uuid,
customer_email=customer_email,
customer_name=customer_name,
customer_company=customer_company,
status='pending'
)
session.add(new_job)
session.flush() # ID generieren
def create_adaptive_card_payload(customer_name: str, send_time: datetime, request_id: str) -> dict:
"""
Erstellt die JSON-Payload für die Adaptive Card in Teams.
"""
send_time_str = send_time.strftime("%H:%M Uhr")
stop_url = f"{FEEDBACK_SERVER_BASE_URL}/stop/{request_id}"
send_now_url = f"{FEEDBACK_SERVER_BASE_URL}/send_now/{request_id}"
# 4. Slots speichern
for slot in selected_slots:
new_slot = ProposedSlot(
job_id=new_job.id,
start_time=slot['start'],
end_time=slot['end']
)
session.add(new_slot)
session.commit()
return new_job.job_uuid, selected_slots
except Exception as e:
session.rollback()
raise e
finally:
session.close()
def _select_best_slots(self, session, candidate_slots):
"""Wählt Slots aus, die noch nicht 'voll' sind (Faktor 3)."""
valid_slots = []
# Wir betrachten nur Vorschläge der letzten 24h als "aktiv"
yesterday = datetime.datetime.now() - datetime.timedelta(days=1)
for slot in candidate_slots:
# Wie oft wurde dieser Start-Zeitpunkt in den letzten 24h vorgeschlagen?
count = session.query(func.count(ProposedSlot.id)).filter(ProposedSlot.start_time == slot['start']).filter(ProposedSlot.job.has(ProposalJob.created_at >= yesterday)).scalar()
if count < MAX_PROPOSALS_PER_SLOT:
valid_slots.append(slot)
if len(valid_slots) >= 2:
break
return valid_slots
def _mock_calendar_availability(self):
"""Simuliert freie Termine für morgen."""
tomorrow = datetime.date.today() + datetime.timedelta(days=1)
# Ein Slot Vormittags (10:30), einer Nachmittags (14:00)
return [
card = {
"type": "message",
"attachments": [
{
'start': datetime.datetime.combine(tomorrow, datetime.time(10, 30)),
'end': datetime.datetime.combine(tomorrow, datetime.time(11, 15))
},
{
'start': datetime.datetime.combine(tomorrow, datetime.time(14, 0)),
'end': datetime.datetime.combine(tomorrow, datetime.time(14, 45))
"contentType": "application/vnd.microsoft.card.adaptive",
"content": {
"type": "AdaptiveCard",
"$schema": "http://adaptivecards.io/schemas/adaptive-card.json",
"version": "1.4",
"body": [
{
"type": "TextBlock",
"text": f"🤖 Automatisierte E-Mail an {customer_name} (via Trading Twins) wird um {send_time_str} ausgesendet.",
"wrap": True,
"size": "Medium",
"weight": "Bolder"
},
{
"type": "TextBlock",
"text": f"Wenn Du bis {send_time_str} NICHT reagierst, wird die generierte E-Mail automatisch ausgesendet.",
"wrap": True,
"isSubtle": True
}
],
"actions": [
{
"type": "Action.OpenUrl",
"title": "❌ STOP Aussendung",
"url": stop_url,
"style": "destructive"
},
{
"type": "Action.OpenUrl",
"title": "✅ JETZT Aussenden",
"url": send_now_url,
"style": "positive"
}
]
}
}
]
}
return card
def get_job_status(self, job_uuid):
session = self.Session()
job = session.query(ProposalJob).filter_by(job_uuid=job_uuid).first()
status = job.status if job else None
session.close()
return status
# --- Haupt-Workflow-Logik ---
def get_job_details(self, job_uuid):
"""Holt alle Details zu einem Job inklusive der Slots."""
session = self.Session()
job = session.query(ProposalJob).filter_by(job_uuid=job_uuid).first()
if not job:
session.close()
return None
# Wir müssen die Daten extrahieren, bevor die Session geschlossen wird
details = {
'uuid': job.job_uuid,
'email': job.customer_email,
'name': job.customer_name,
'company': job.customer_company,
'status': job.status,
'slots': [{'start': s.start_time, 'end': s.end_time} for s in job.slots]
def send_teams_notification(payload: dict):
"""Sendet die vorbereitete Payload an den Teams Webhook."""
try:
response = requests.post(TEAMS_WEBHOOK_URL, json=payload, timeout=10)
if response.status_code == 200 or response.status_code == 202:
print(f"INFO: Adaptive Card sent to Teams. Response: {response.text}")
return True
else:
print(f"ERROR: Failed to send card. Status: {response.status_code}, Text: {response.text}")
return False
except requests.RequestException as e:
print(f"ERROR: Request to Teams failed: {e}")
return False
def process_email_request(request_id: str, customer_name: str):
"""
Der Hauptprozess, der die Benachrichtigung auslöst und auf das Ergebnis wartet.
"""
send_time = datetime.now() + timedelta(minutes=DEFAULT_WAIT_MINUTES)
with _lock:
request_status_storage[request_id] = {
"status": "pending", # pending, cancelled, send_now, sent, timeout
"customer": customer_name,
"send_time": send_time.isoformat()
}
session.close()
return details
def update_job_status(self, job_uuid, new_status):
session = self.Session()
job = session.query(ProposalJob).filter_by(job_uuid=job_uuid).first()
if job:
job.status = new_status
if new_status == 'approved':
job.approved_at = datetime.datetime.now()
session.commit()
session.close()
# 1. Adaptive Card erstellen und an Teams senden
adaptive_card = create_adaptive_card_payload(customer_name, send_time, request_id)
if not send_teams_notification(adaptive_card):
print(f"CRITICAL: Could not send Teams notification for request {request_id}. Aborting.")
return
# 2. Warten auf menschliches Feedback oder Timeout
print(f"INFO: Waiting for feedback for request {request_id} until {send_time.strftime('%H:%M:%S')}...")
while datetime.now() < send_time:
with _lock:
current_status = request_status_storage[request_id]["status"]
if current_status == "cancelled":
print(f"INFO: Request {request_id} was cancelled by the user.")
return
if current_status == "send_now":
print(f"INFO: Request {request_id} was triggered to send immediately by the user.")
break # Schleife verlassen und sofort senden
time.sleep(5)
# 3. Finale Entscheidung und Ausführung
with _lock:
final_status = request_status_storage[request_id]["status"]
# Update status to avoid race conditions
if final_status == "pending":
request_status_storage[request_id]["status"] = "timeout"
final_status = "timeout"
if final_status in ["send_now", "timeout"]:
print(f"SUCCESS: Proceeding to send email for request {request_id} (Status: {final_status})")
# --- HIER KOMMT DIE ECHTE E-MAIL LOGIK (MS GRAPH API) ---
# send_email_via_graph_api(customer_name, signature_path, banner_path)
print("MOCK: Email would be sent now.")
# ---------------------------------------------------------
with _lock:
request_status_storage[request_id]["status"] = "sent"
else:
# Dieser Fall sollte eigentlich nicht eintreten, aber zur Sicherheit
print(f"WARN: Email for request {request_id} was not sent due to final status: {final_status}")
# --- Feedback-Server (FastAPI) ---
app = FastAPI()
@app.get("/stop/{request_id}")
async def stop_sending(request_id: str):
with _lock:
if request_id in request_status_storage:
if request_status_storage[request_id]["status"] == "pending":
request_status_storage[request_id]["status"] = "cancelled"
customer = request_status_storage[request_id]['customer']
print(f"INFO: Received STOP for request {request_id}")
return Response(content=f"<html><body><h1>✔️ Stopp-Anfrage für E-Mail an {customer} erhalten.</h1><p>Der Versand wurde erfolgreich abgebrochen.</p></body></html>", media_type="text/html")
else:
status = request_status_storage[request_id]['status']
return Response(content=f"<html><body><h1>⚠️ Aktion bereits ausgeführt</h1><p>Der Status für diese Anfrage ist bereits '{status}'. Es kann nicht mehr gestoppt werden.</p></body></html>", media_type="text/html", status_code=409)
return Response(content="<html><body><h1>❌ Fehler</h1><p>Anfrage-ID nicht gefunden.</p></body></html>", media_type="text/html", status_code=404)
@app.get("/send_now/{request_id}")
async def send_now(request_id: str):
with _lock:
if request_id in request_status_storage:
if request_status_storage[request_id]["status"] == "pending":
request_status_storage[request_id]["status"] = "send_now"
customer = request_status_storage[request_id]['customer']
print(f"INFO: Received SEND_NOW for request {request_id}")
return Response(content=f"<html><body><h1>✔️ Sofort-Senden-Anfrage für E-Mail an {customer} erhalten.</h1><p>Der Versand wird sofort ausgelöst.</p></body></html>", media_type="text/html")
else:
status = request_status_storage[request_id]['status']
return Response(content=f"<html><body><h1>⚠️ Aktion bereits ausgeführt</h1><p>Der Status für diese Anfrage ist bereits '{status}'.</p></body></html>", media_type="text/html", status_code=409)
return Response(content="<html><body><h1>❌ Fehler</h1><p>Anfrage-ID nicht gefunden.</p></body></html>", media_type="text/html", status_code=404)
def run_server():
"""Startet den FastAPI-Server."""
uvicorn.run(app, host="0.0.0.0", port=8004)
if __name__ == "__main__":
# Starte den Feedback-Server in einem separaten Thread
server_thread = Thread(target=run_server)
server_thread.daemon = True
server_thread.start()
print("INFO: Feedback-Server started on port 8004 in background.")
time.sleep(2) # Kurz warten, bis der Server gestartet ist
# Simuliere eine neue Anfrage
test_request_id = f"req_{int(time.time())}"
test_customer = "Klinikum Erding"
print(f"\n--- Starting new email request for '{test_customer}' with ID: {test_request_id} ---")
process_email_request(test_request_id, test_customer)
print(f"--- Process for {test_request_id} finished. ---")
# Halte das Hauptprogramm am Leben, damit der Server weiterlaufen kann
# In einer echten Anwendung wäre dies Teil eines größeren Dienstes.
print("\nManager is running. Press Ctrl+C to stop.")
try:
while True:
time.sleep(1)
except KeyboardInterrupt:
print("\nShutting down manager.")

View File

@@ -1,27 +1,40 @@
<br>
<div style="font-family: Arial, sans-serif; font-size: 14px; color: #333;">
<p>Freundliche Grüße<br>
<strong>Elizabeta Melcer</strong><br>
Inside Sales Managerin</p>
<!DOCTYPE html>
<html lang="de">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>E-Mail Signatur</title>
</head>
<body>
<!--
HINWEIS:
Dieser Inhalt wird von der IT-Abteilung bereitgestellt.
Bitte den finalen HTML-Code hier einfügen.
Das Bild 'RoboPlanetBannerWebinarEinladung.png' muss sich im selben Verzeichnis befinden.
[31988f42]
-->
<p>Freundliche Grüße</p>
<p>
<strong>RoboPlanet GmbH</strong><br>
Schatzbogen 39, 81829 München<br>
T: +49 89 420490-402 | M: +49 175 8334071<br>
<a href="mailto:e.melcer@robo-planet.de">e.melcer@robo-planet.de</a> | <a href="http://www.robo-planet.de">www.robo-planet.de</a>
<b>Elizabeta Melcer</b><br>
Inside Sales Managerin
</p>
<p>
<a href="#">LinkedIn</a> | <a href="#">Instagram</a> | <a href="#">Newsletteranmeldung</a>
<!-- Wackler Logo -->
<b>RoboPlanet GmbH</b><br>
Schatzbogen 39, 81829 München<br>
T: +49 89 420490-402 | M: +49 175 8334071<br>
<a href="mailto:e.melcer@robo-planet.de">e.melcer@robo-planet.de</a> | <a href="http://www.robo-planet.de">www.robo-planet.de</a>
</p>
<p style="font-size: 10px; color: #777;">
Sitz der Gesellschaft München | Geschäftsführung: Axel Banoth<br>
Registergericht AG München, HRB 296113 | USt.-IdNr. DE400464410<br>
<a href="#">Hinweispflichten zum Datenschutz</a>
<p>
<a href="#">LinkedIn</a> | <a href="#">Instagram</a> | <a href="#">Newsletteranmeldung</a>
</p>
<!-- Platzhalter für das Bild -->
<img src="https://robo-planet.de/wp-content/uploads/2024/01/RoboPlanet_Logo.png" alt="RoboPlanet Logo" width="150"><br>
<img src="cid:banner_image" alt="Webinar Einladung" width="400">
</div>
<p style="font-size: smaller; color: grey;">
Sitz der Gesellschaft München | Geschäftsführung: Axel Banoth<br>
Registergericht AG München, HRB 296113 | USt.-IdNr. DE400464410<br>
<a href="#">Hinweispflichten zum Datenschutz</a>
</p>
<p>
<img src="RoboPlanetBannerWebinarEinladung.png" alt="RoboPlanet Webinar Einladung">
</p>
</body>
</html>

View File

@@ -0,0 +1,50 @@
import requests
import json
import os
def send_teams_message(webhook_url, message):
"""
Sends a simple message to a Microsoft Teams channel using a webhook.
Args:
webhook_url (str): The URL of the incoming webhook.
message (str): The plain text message to send.
Returns:
bool: True if the message was sent successfully (HTTP 200), False otherwise.
"""
if not webhook_url:
print("Error: TEAMS_WEBHOOK_URL is not set.")
return False
headers = {
"Content-Type": "application/json"
}
payload = {
"text": message
}
try:
response = requests.post(webhook_url, headers=headers, data=json.dumps(payload), timeout=10)
if response.status_code == 200:
print("Message sent successfully to Teams.")
return True
else:
print(f"Failed to send message. Status code: {response.status_code}")
print(f"Response: {response.text}")
return False
except requests.exceptions.RequestException as e:
print(f"An error occurred while sending the request: {e}")
return False
if __name__ == "__main__":
# The webhook URL is taken directly from the project description for this test.
# In a real application, this should be loaded from an environment variable.
webhook_url = "https://wacklergroup.webhook.office.com/webhookb2/fe728cde-790c-4190-b1d3-be393ca0f9bd@6d85a9ef-3878-420b-8f43-38d6cb12b665/IncomingWebhook/e9a8ee6157594a6cab96048cf2ea2232/d26033cd-a81f-41a6-8cd2-b4a3ba0b5a01/V2WFmjcbkMzSU4f6lDSdUOM9VNm7F7n1Th4YDiu3fLZ_Y1"
test_message = "🤖 This is a test message from the Gemini Trading Twins Engine. If you see this, the webhook is working. [31988f42]"
send_teams_message(webhook_url, test_message)