Compare commits

...

7 Commits

8 changed files with 478 additions and 7 deletions

View File

@@ -3,6 +3,10 @@ google_cloud_location: us-central1
firestore_db: bnt-orquestador-cognitivo-firestore-bdo-dev firestore_db: bnt-orquestador-cognitivo-firestore-bdo-dev
# Notifications configuration
notifications_collection_path: "artifacts/bnt-orquestador-cognitivo-dev/notifications"
notifications_max_to_notify: 5
mcp_remote_url: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app/mcp" mcp_remote_url: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app/mcp"
# audience sin la ruta, para emitir el ID Token: # audience sin la ruta, para emitir el ID Token:
mcp_audience: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app" mcp_audience: "https://ap01194-orq-cog-rag-connector-1007577023101.us-central1.run.app"

View File

@@ -13,6 +13,7 @@ dependencies = [
"google-cloud-firestore>=2.23.0", "google-cloud-firestore>=2.23.0",
"pydantic-settings[yaml]>=2.13.1", "pydantic-settings[yaml]>=2.13.1",
"google-auth>=2.34.0", "google-auth>=2.34.0",
"google-genai>=1.64.0",
] ]
[build-system] [build-system]

View File

@@ -1,34 +1,61 @@
"""ADK agent with vector search RAG tool.""" """ADK agent with vector search RAG tool."""
from functools import partial
from google import genai from google import genai
from google.adk.agents.llm_agent import Agent from google.adk.agents.llm_agent import Agent
from google.adk.runners import Runner from google.adk.runners import Runner
from google.adk.tools.mcp_tool import McpToolset from google.adk.tools.mcp_tool import McpToolset
from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams from google.adk.tools.mcp_tool.mcp_session_manager import StreamableHTTPConnectionParams
from google.cloud.firestore_v1.async_client import AsyncClient from google.cloud.firestore_v1.async_client import AsyncClient
from google.genai.types import Content, Part
from va_agent.auth import auth_headers_provider from va_agent.auth import auth_headers_provider
from va_agent.config import settings from va_agent.config import settings
from va_agent.dynamic_instruction import provide_dynamic_instruction
from va_agent.notifications import NotificationService
from va_agent.session import FirestoreSessionService from va_agent.session import FirestoreSessionService
from va_agent.governance import GovernancePlugin
# MCP Toolset for RAG knowledge search
toolset = McpToolset( toolset = McpToolset(
connection_params=StreamableHTTPConnectionParams(url=settings.mcp_remote_url), connection_params=StreamableHTTPConnectionParams(url=settings.mcp_remote_url),
header_provider=auth_headers_provider, header_provider=auth_headers_provider,
) )
agent = Agent(
model=settings.agent_model,
name=settings.agent_name,
instruction=settings.agent_instructions,
tools=[toolset],
)
# Shared Firestore client for session service and notifications
firestore_db = AsyncClient(database=settings.firestore_db)
# Session service with compaction
session_service = FirestoreSessionService( session_service = FirestoreSessionService(
db=AsyncClient(database=settings.firestore_db), db=firestore_db,
compaction_token_threshold=10_000, compaction_token_threshold=10_000,
genai_client=genai.Client(), genai_client=genai.Client(),
) )
# Notification service
notification_service = NotificationService(
db=firestore_db,
collection_path=settings.notifications_collection_path,
max_to_notify=settings.notifications_max_to_notify,
)
# Agent with static and dynamic instructions
governance = GovernancePlugin()
agent = Agent(
model=settings.agent_model,
name=settings.agent_name,
static_instruction=Content(
role="user",
parts=[Part(text=settings.agent_instructions)],
),
instruction=settings.agent_instructions,
tools=[toolset],
after_model_callback=governance.after_model_callback,
)
# Runner
runner = Runner( runner = Runner(
app_name="va_agent", app_name="va_agent",
agent=agent, agent=agent,

View File

@@ -26,6 +26,12 @@ class AgentSettings(BaseSettings):
# Firestore configuration # Firestore configuration
firestore_db: str firestore_db: str
# Notifications configuration
notifications_collection_path: str = (
"artifacts/bnt-orquestador-cognitivo-dev/notifications"
)
notifications_max_to_notify: int = 5
# MCP configuration # MCP configuration
mcp_audience: str mcp_audience: str
mcp_remote_url: str mcp_remote_url: str

View File

@@ -0,0 +1,120 @@
"""Dynamic instruction provider for VAia agent."""
from __future__ import annotations
import logging
from typing import TYPE_CHECKING
if TYPE_CHECKING:
from google.adk.agents.readonly_context import ReadonlyContext
from va_agent.notifications import NotificationService
logger = logging.getLogger(__name__)
async def provide_dynamic_instruction(
notification_service: NotificationService,
ctx: ReadonlyContext | None = None,
) -> str:
"""Provide dynamic instructions based on pending notifications.
This function is called by the ADK agent on each message. It:
1. Checks if this is the first message in the session (< 2 events)
2. Queries Firestore for pending notifications
3. Marks them as notified
4. Returns a dynamic instruction for the agent to mention them
Args:
notification_service: Service for fetching/marking notifications
ctx: Agent context containing session information
Returns:
Dynamic instruction string (empty if no notifications or not first message)
"""
# Only check notifications on the first message
if not ctx or not ctx._invocation_context:
logger.debug("No context available for dynamic instruction")
return ""
session = ctx._invocation_context.session
if not session:
logger.debug("No session available for dynamic instruction")
return ""
# FOR TESTING: Always check for notifications (comment out to enable first-message-only)
# Only check on first message (when events list is empty or has only 1-2 events)
# Events include both user and agent messages, so < 2 means first interaction
# event_count = len(session.events) if session.events else 0
#
# if event_count >= 2:
# logger.debug(
# "Skipping notification check: not first message (event_count=%d)",
# event_count,
# )
# return ""
# Extract phone number from user_id (they are the same in this implementation)
phone_number = session.user_id
logger.info(
"First message detected for user %s, checking for pending notifications",
phone_number,
)
try:
# Fetch pending notifications
pending_notifications = await notification_service.get_pending_notifications(
phone_number
)
if not pending_notifications:
logger.info("No pending notifications for user %s", phone_number)
return ""
# Build dynamic instruction with notification details
notification_ids = [n.get("id_notificacion") for n in pending_notifications]
count = len(pending_notifications)
# Format notification details for the agent
notification_details = []
for notif in pending_notifications:
evento = notif.get("nombre_evento_dialogflow", "notificacion")
texto = notif.get("texto", "Sin texto")
notification_details.append(f" - Evento: {evento} | Texto: {texto}")
details_text = "\n".join(notification_details)
instruction = f"""
IMPORTANTE - NOTIFICACIONES PENDIENTES:
El usuario tiene {count} notificación(es) sin leer:
{details_text}
INSTRUCCIONES:
- Menciona estas notificaciones de forma natural en tu respuesta inicial
- No necesitas leerlas todas literalmente, solo hazle saber que las tiene
- Sé breve y directo según tu personalidad (directo y cálido)
- Si el usuario pregunta algo específico, prioriza responder eso primero y luego menciona las notificaciones
Ejemplo: "¡Hola! 👋 Antes de empezar, veo que tienes {count} notificación(es) pendiente(s) en tu cuenta. ¿Te gustaría revisarlas o prefieres que te ayude con algo más?"
"""
# Mark notifications as notified in Firestore
await notification_service.mark_as_notified(phone_number, notification_ids)
logger.info(
"Returning dynamic instruction with %d notification(s) for user %s",
count,
phone_number,
)
return instruction
except Exception:
logger.exception(
"Error building dynamic instruction for user %s", phone_number
)
return ""

View File

@@ -0,0 +1,79 @@
"""GovernancePlugin: Guardrails for VAia, the virtual assistant for VA."""
import logging
import re
from google.adk.agents.callback_context import CallbackContext
from google.adk.models import LlmResponse
logger = logging.getLogger(__name__)
FORBIDDEN_EMOJIS = [
"🥵","🔪","🎰","🎲","🃏","😤","🤬","😡","😠","🩸","🧨","🪓","☠️","💀",
"💣","🔫","👗","💦","🍑","🍆","👄","👅","🫦","💩","⚖️","⚔️","✝️","🕍",
"🕌","","🍻","🍸","🥃","🍷","🍺","🚬","👹","👺","👿","😈","🤡","🧙",
"🧙‍♀️", "🧙‍♂️", "🧛", "🧛‍♀️", "🧛‍♂️", "🔞","🧿","💊", "💏"
]
class GovernancePlugin:
"""Guardrail executor for VAia requests as a Agent engine callbacks."""
def __init__(self) -> None:
"""Initialize guardrail model (structured output), prompt and emojis patterns."""
self._combined_pattern = self._get_combined_pattern()
def _get_combined_pattern(self):
person_pattern = r"(?:🧑|👩|👨)"
tone_pattern = r"[\U0001F3FB-\U0001F3FF]?"
# Unique pattern that combines all forbidden emojis, including complex ones with skin tones
combined_pattern = re.compile(
rf"{person_pattern}{tone_pattern}\u200d❤?\u200d💋\u200d{person_pattern}{tone_pattern}" # kiss
rf"|{person_pattern}{tone_pattern}\u200d❤?\u200d{person_pattern}{tone_pattern}" # lovers
rf"|🖕{tone_pattern}" # middle finger with all skin tone variations
rf"|{'|'.join(map(re.escape, sorted(FORBIDDEN_EMOJIS, key=len, reverse=True)))}" # simple emojis
rf"|\u200d|\uFE0F" # residual ZWJ and variation selectors
)
return combined_pattern
def _remove_emojis(self, text: str) -> tuple[str, list[str]]:
removed = self._combined_pattern.findall(text)
text = self._combined_pattern.sub("", text)
return text.strip(), removed
def after_model_callback(
self,
callback_context: CallbackContext | None = None,
llm_response: LlmResponse | None = None,
) -> None:
"""Guardrail post-processing.
Remove forbidden emojis from the model response.
"""
try:
text_out = ""
if llm_response and llm_response.content:
content = llm_response.content
parts = getattr(content, "parts", None)
if parts:
part = parts[0]
text_value = getattr(part, "text", "")
if isinstance(text_value, str):
text_out = text_value
if text_out:
new_text, deleted = self._remove_emojis(text_out)
if llm_response and llm_response.content and llm_response.content.parts:
llm_response.content.parts[0].text = new_text
if deleted:
if callback_context:
callback_context.state["removed_emojis"] = deleted
logger.warning(
"Removed forbidden emojis from response: %s",
deleted,
)
except Exception:
logger.exception("Error in after_model_callback")

View File

@@ -0,0 +1,232 @@
"""Notification management for VAia agent."""
from __future__ import annotations
import logging
import time
from typing import TYPE_CHECKING, Any
if TYPE_CHECKING:
from google.cloud.firestore_v1.async_client import AsyncClient
logger = logging.getLogger(__name__)
class NotificationService:
"""Service for fetching and managing user notifications from Firestore."""
def __init__(
self,
*,
db: AsyncClient,
collection_path: str,
max_to_notify: int = 5,
) -> None:
"""Initialize NotificationService.
Args:
db: Firestore async client
collection_path: Path to notifications collection
max_to_notify: Maximum number of notifications to return
"""
self._db = db
self._collection_path = collection_path
self._max_to_notify = max_to_notify
async def get_pending_notifications(
self, phone_number: str
) -> list[dict[str, Any]]:
"""Get pending notifications for a user.
Retrieves notifications that have not been notified by the agent yet,
ordered by timestamp (most recent first), limited to max_to_notify.
Args:
phone_number: User's phone number (used as document ID)
Returns:
List of notification dictionaries with structure:
{
"id_notificacion": str,
"texto": str,
"status": str,
"timestamp_creacion": timestamp,
"parametros": {...}
}
"""
try:
# Query Firestore document by phone number
doc_ref = self._db.collection(self._collection_path).document(
phone_number
)
doc = await doc_ref.get()
if not doc.exists:
logger.info(
"No notification document found for phone: %s", phone_number
)
return []
data = doc.to_dict() or {}
all_notifications = data.get("notificaciones", [])
if not all_notifications:
logger.info("No notifications in array for phone: %s", phone_number)
return []
# Filter notifications that have NOT been notified by the agent
pending = [
n
for n in all_notifications
if not n.get("notified_by_agent", False)
]
if not pending:
logger.info(
"All notifications already notified for phone: %s", phone_number
)
return []
# Sort by timestamp_creacion (most recent first)
pending.sort(
key=lambda n: n.get("timestamp_creacion", 0), reverse=True
)
# Return top N most recent
result = pending[: self._max_to_notify]
logger.info(
"Found %d pending notifications for phone: %s (returning top %d)",
len(pending),
phone_number,
len(result),
)
return result
except Exception:
logger.exception(
"Failed to fetch notifications for phone: %s", phone_number
)
return []
async def mark_as_notified(
self, phone_number: str, notification_ids: list[str]
) -> bool:
"""Mark notifications as notified by the agent.
Updates the notifications in Firestore by adding:
- notified_by_agent: true
- notified_at: current timestamp
Args:
phone_number: User's phone number (document ID)
notification_ids: List of id_notificacion values to mark
Returns:
True if update was successful, False otherwise
"""
if not notification_ids:
return True
try:
doc_ref = self._db.collection(self._collection_path).document(
phone_number
)
doc = await doc_ref.get()
if not doc.exists:
logger.warning(
"Cannot mark notifications as notified: document not found for %s",
phone_number,
)
return False
data = doc.to_dict() or {}
notificaciones = data.get("notificaciones", [])
if not notificaciones:
logger.warning(
"Cannot mark notifications: empty array for %s", phone_number
)
return False
# Update matching notifications
now = time.time()
updated_count = 0
for notif in notificaciones:
if notif.get("id_notificacion") in notification_ids:
notif["notified_by_agent"] = True
notif["notified_at"] = now
updated_count += 1
if updated_count == 0:
logger.warning(
"No notifications matched IDs for phone: %s", phone_number
)
return False
# Save back to Firestore
await doc_ref.update(
{
"notificaciones": notificaciones,
"ultima_actualizacion": now,
}
)
logger.info(
"Marked %d notification(s) as notified for phone: %s",
updated_count,
phone_number,
)
return True
except Exception:
logger.exception(
"Failed to mark notifications as notified for phone: %s",
phone_number,
)
return False
def format_notification_summary(
self, notifications: list[dict[str, Any]]
) -> str:
"""Format notifications into a human-readable summary.
Args:
notifications: List of notification dictionaries
Returns:
Formatted string summarizing the notifications
"""
if not notifications:
return ""
count = len(notifications)
summary_lines = [
f"El usuario tiene {count} notificación(es) pendiente(s):"
]
for i, notif in enumerate(notifications, 1):
texto = notif.get("texto", "Sin texto")
params = notif.get("parametros", {})
# Extract key parameters if available
amount = params.get("notification_po_amount")
tx_id = params.get("notification_po_transaction_id")
line = f"{i}. {texto}"
if amount:
line += f" (monto: ${amount})"
if tx_id:
line += f" [ID: {tx_id}]"
summary_lines.append(line)
return "\n".join(summary_lines)

2
uv.lock generated
View File

@@ -1924,6 +1924,7 @@ dependencies = [
{ name = "google-adk" }, { name = "google-adk" },
{ name = "google-auth" }, { name = "google-auth" },
{ name = "google-cloud-firestore" }, { name = "google-cloud-firestore" },
{ name = "google-genai" },
{ name = "pydantic-settings", extra = ["yaml"] }, { name = "pydantic-settings", extra = ["yaml"] },
] ]
@@ -1941,6 +1942,7 @@ requires-dist = [
{ name = "google-adk", specifier = ">=1.14.1" }, { name = "google-adk", specifier = ">=1.14.1" },
{ name = "google-auth", specifier = ">=2.34.0" }, { name = "google-auth", specifier = ">=2.34.0" },
{ name = "google-cloud-firestore", specifier = ">=2.23.0" }, { name = "google-cloud-firestore", specifier = ">=2.23.0" },
{ name = "google-genai", specifier = ">=1.64.0" },
{ name = "pydantic-settings", extras = ["yaml"], specifier = ">=2.13.1" }, { name = "pydantic-settings", extras = ["yaml"], specifier = ">=2.13.1" },
] ]