gateway/modules/features/chatbot/interfaceFeatureChatbot.py
ValueOn AG c8b7517209 refactor: modules/services/ abgeloest durch serviceCenter + serviceHub
serviceCenter = DI-Container (Resolver, Registry, Context) fuer Service-Instanziierung
serviceHub = Consumer-facing Aggregation (DB-Interfaces, Runtime-State, lazy Service-Resolution via serviceCenter)

- modules/serviceHub/ erstellt: ServiceHub, PublicService, getInterface()
- 22 Consumer-Dateien migriert (routes, features, tests): imports von modules.services auf serviceHub bzw. serviceCenter umgestellt
- resolver.py: legacy fallback auf altes services/ entfernt
- modules/services/ komplett geloescht (83 Dateien inkl. dead code mainAiChat.py)
- pre-extraction: progress callback durch chunk-pipeline propagiert, operationType DATA_EXTRACT->DATA_ANALYSE fuer guenstigeres Modell
2026-03-14 11:51:45 +01:00

1614 lines
76 KiB
Python

# Copyright (c) 2025 Patrick Motsch
# All rights reserved.
"""
Interface to Chatbot database and AI Connectors.
Uses the PostgreSQL connector for data access with user/feature-instance filtering.
Chatbot-specific models in poweron_chatbot (separate from workflow engine).
"""
import logging
import uuid
import math
from typing import Dict, Any, List, Optional, Union
from enum import Enum
from pydantic import BaseModel, Field
import asyncio
from modules.security.rbac import RbacClass
from modules.datamodels.datamodelRbac import AccessRuleContext
from modules.datamodels.datamodelUam import AccessLevel
from modules.datamodels.datamodelChat import UserInputRequest
from modules.shared.timeUtils import getUtcTimestamp, parseTimestamp
# =============================================================================
# Chatbot-specific Pydantic models for poweron_chatbot (per-instance isolation)
# =============================================================================
class ChatbotDocument(BaseModel):
"""Documents attached to chatbot messages."""
id: str = Field(default_factory=lambda: str(uuid.uuid4()), description="Primary key")
messageId: str = Field(description="Foreign key to message")
fileId: str = Field(description="Foreign key to file")
fileName: str = Field(description="Name of the file")
fileSize: int = Field(description="Size of the file")
mimeType: str = Field(description="MIME type of the file")
roundNumber: Optional[int] = Field(None, description="Round number in workflow")
taskNumber: Optional[int] = Field(None, description="Task number within round")
actionNumber: Optional[int] = Field(None, description="Action number within task")
actionId: Optional[str] = Field(None, description="ID of the action that created this document")
class ChatbotMessage(BaseModel):
"""Messages in chatbot conversations. Must match bridge format in memory.py."""
id: str = Field(default_factory=lambda: str(uuid.uuid4()), description="Primary key")
conversationId: str = Field(description="Foreign key to conversation")
parentMessageId: Optional[str] = Field(None, description="Parent message ID for threading")
documents: List[ChatbotDocument] = Field(default_factory=list, description="Associated documents")
documentsLabel: Optional[str] = Field(None, description="Label for the set of documents")
message: Optional[str] = Field(None, description="Message content")
role: str = Field(description="Role of the message sender")
status: str = Field(description="Status of the message (first, step, last)")
sequenceNr: Optional[int] = Field(default=0, description="Sequence number of the message")
publishedAt: Optional[float] = Field(default=None, description="When the message was published (UTC timestamp)")
success: Optional[bool] = Field(None, description="Whether the message processing was successful")
actionId: Optional[str] = Field(None, description="ID of the action that produced this message")
actionMethod: Optional[str] = Field(None, description="Method of the action that produced this message")
actionName: Optional[str] = Field(None, description="Name of the action that produced this message")
roundNumber: Optional[int] = Field(None, description="Round number in workflow")
taskNumber: Optional[int] = Field(None, description="Task number within round")
actionNumber: Optional[int] = Field(None, description="Action number within task")
taskProgress: Optional[str] = Field(None, description="Task progress status")
actionProgress: Optional[str] = Field(None, description="Action progress status")
class ChatbotLog(BaseModel):
"""Log entries for chatbot conversations."""
id: str = Field(default_factory=lambda: str(uuid.uuid4()), description="Primary key")
conversationId: str = Field(description="Foreign key to conversation")
message: str = Field(description="Log message")
type: str = Field(description="Log type (info, warning, error, etc.)")
timestamp: float = Field(default_factory=getUtcTimestamp, description="When the log entry was created (UTC timestamp)")
status: Optional[str] = Field(None, description="Status of the log entry")
progress: Optional[float] = Field(None, description="Progress indicator (0.0 to 1.0)")
performance: Optional[Dict[str, Any]] = Field(None, description="Performance metrics")
parentId: Optional[str] = Field(None, description="Parent operation ID")
operationId: Optional[str] = Field(None, description="Operation ID to group related log entries")
roundNumber: Optional[int] = Field(None, description="Round number in workflow")
taskNumber: Optional[int] = Field(None, description="Task number within round")
actionNumber: Optional[int] = Field(None, description="Action number within task")
class ChatbotWorkflowModeEnum(str, Enum):
WORKFLOW_CHATBOT = "Chatbot"
class ChatbotConversation(BaseModel):
"""Chatbot conversation container. Per feature-instance isolation via featureInstanceId."""
id: str = Field(default_factory=lambda: str(uuid.uuid4()), description="Primary key")
featureInstanceId: str = Field(description="Feature instance ID for per-instance isolation")
name: Optional[str] = Field(None, description="Name of the conversation")
status: str = Field(default="running", description="Current status")
currentRound: int = Field(default=0, description="Current round number")
lastActivity: float = Field(default_factory=getUtcTimestamp, description="Timestamp of last activity")
startedAt: float = Field(default_factory=getUtcTimestamp, description="When the conversation started")
workflowMode: ChatbotWorkflowModeEnum = Field(default=ChatbotWorkflowModeEnum.WORKFLOW_CHATBOT, description="Workflow mode")
maxSteps: int = Field(default=10, description="Maximum number of iterations")
# Hydrated from child tables (not stored in DB)
logs: List[ChatbotLog] = Field(default_factory=list, description="Conversation logs")
messages: List[ChatbotMessage] = Field(default_factory=list, description="Conversation messages")
import json
from modules.datamodels.datamodelUam import User
# DYNAMIC PART: Connectors to the Interface
from modules.connectors.connectorDbPostgre import DatabaseConnector
from modules.datamodels.datamodelPagination import PaginationParams, PaginatedResult
from modules.interfaces.interfaceRbac import getRecordsetWithRBAC
# Basic Configurations
from modules.shared.configuration import APP_CONFIG
logger = logging.getLogger(__name__)
# Singleton factory for Chat instances
_chatInterfaces = {}
def storeDebugMessageAndDocuments(message, currentUser, mandateId=None, featureInstanceId=None) -> None:
"""
Store message and documents (metadata and file bytes) for debugging purposes.
Structure: {log_dir}/debug/messages/m_round_task_action_timestamp/documentlist_label/
- message.json, message_text.txt
- document_###_metadata.json
- document_###_<original_filename> (actual file bytes)
Args:
message: ChatbotMessage or ChatMessage-like object to store
currentUser: Current user for component interface access
mandateId: Mandate ID for RBAC context (avoids overwriting singleton state)
featureInstanceId: Feature instance ID for RBAC context
"""
try:
import os
from datetime import datetime, UTC
from modules.shared.debugLogger import _getBaseDebugDir, _ensureDir
from modules.interfaces.interfaceDbManagement import getInterface
# Create base debug directory (use base debug dir, not prompts subdirectory)
baseDebugDir = _getBaseDebugDir()
debug_root = os.path.join(baseDebugDir, 'messages')
_ensureDir(debug_root)
# Generate timestamp
timestamp = datetime.now(UTC).strftime('%Y%m%d-%H%M%S-%f')[:-3]
# Create message folder name: m_round_task_action_timestamp
# Use actual values from message, not defaults
round_str = str(message.roundNumber) if message.roundNumber is not None else "0"
task_str = str(message.taskNumber) if message.taskNumber is not None else "0"
action_str = str(message.actionNumber) if message.actionNumber is not None else "0"
message_folder = f"{timestamp}_m_{round_str}_{task_str}_{action_str}"
message_path = os.path.join(debug_root, message_folder)
os.makedirs(message_path, exist_ok=True)
# Store message data - use dict() instead of model_dump() for compatibility
message_file = os.path.join(message_path, "message.json")
with open(message_file, "w", encoding="utf-8") as f:
# Convert message to dict manually to avoid model_dump() issues
message_dict = {
"id": message.id,
"workflowId": getattr(message, "conversationId", None) or getattr(message, "workflowId", ""),
"parentMessageId": message.parentMessageId,
"message": message.message,
"role": message.role,
"status": message.status,
"sequenceNr": message.sequenceNr,
"publishedAt": message.publishedAt,
"roundNumber": message.roundNumber,
"taskNumber": message.taskNumber,
"actionNumber": message.actionNumber,
"documentsLabel": message.documentsLabel,
"actionId": message.actionId,
"actionMethod": message.actionMethod,
"actionName": message.actionName,
"success": message.success,
"documents": []
}
json.dump(message_dict, f, indent=2, ensure_ascii=False, default=str)
# Store message content as text
if message.message:
message_text_file = os.path.join(message_path, "message_text.txt")
with open(message_text_file, "w", encoding="utf-8") as f:
f.write(str(message.message))
# Store documents if provided
if message.documents and len(message.documents) > 0:
# Group documents by documentsLabel
documents_by_label = {}
for doc in message.documents:
label = message.documentsLabel or 'default'
if label not in documents_by_label:
documents_by_label[label] = []
documents_by_label[label].append(doc)
# Create subfolder for each document label
for label, docs in documents_by_label.items():
# Sanitize label for filesystem
safe_label = "".join(c for c in str(label) if c.isalnum() or c in (' ', '-', '_')).rstrip()
safe_label = safe_label.replace(' ', '_')
if not safe_label:
safe_label = "default"
label_folder = os.path.join(message_path, safe_label)
_ensureDir(label_folder)
# Store each document
for i, doc in enumerate(docs):
# Create document metadata file
doc_meta = {
"id": doc.id,
"messageId": doc.messageId,
"fileId": doc.fileId,
"fileName": doc.fileName,
"fileSize": doc.fileSize,
"mimeType": doc.mimeType,
"roundNumber": doc.roundNumber,
"taskNumber": doc.taskNumber,
"actionNumber": doc.actionNumber,
"actionId": doc.actionId
}
doc_meta_file = os.path.join(label_folder, f"document_{i+1:03d}_metadata.json")
with open(doc_meta_file, "w", encoding="utf-8") as f:
json.dump(doc_meta, f, indent=2, ensure_ascii=False, default=str)
# Also store the actual file bytes next to metadata for debugging
try:
componentInterface = getInterface(currentUser, mandateId=mandateId, featureInstanceId=featureInstanceId)
file_bytes = componentInterface.getFileData(doc.fileId)
if file_bytes:
# Build a safe filename preserving original name
safe_name = doc.fileName or f"document_{i+1:03d}"
# Avoid path traversal
safe_name = os.path.basename(safe_name)
doc_file_path = os.path.join(label_folder, f"document_{i+1:03d}_" + safe_name)
with open(doc_file_path, "wb") as df:
df.write(file_bytes)
else:
pass
except Exception as e:
pass
except Exception as e:
# Silent fail - don't break main flow
pass
class ChatObjects:
"""
Interface to Chat database and AI Connectors.
Uses the JSON connector for data access with added language support.
"""
def __init__(self, currentUser: Optional[User] = None, mandateId: Optional[str] = None, featureInstanceId: Optional[str] = None):
"""Initializes the Chat Interface.
Args:
currentUser: The authenticated user
mandateId: The mandate ID from RequestContext (X-Mandate-Id header)
featureInstanceId: The feature instance ID from RequestContext (X-Feature-Instance-Id header)
"""
# Initialize variables
self.currentUser = currentUser # Store User object directly
self.userId = currentUser.id if currentUser else None
# Use mandateId from parameter (Request-Context), not from user object
self.mandateId = mandateId
self.featureInstanceId = featureInstanceId
self.featureCode = "chatbot" # For RBAC buildDataObjectKey
self.rbac = None # RBAC interface
# Initialize services
self._initializeServices()
# Initialize database
self._initializeDatabase()
# Set user context if provided
if currentUser:
self.setUserContext(currentUser, mandateId=mandateId, featureInstanceId=featureInstanceId)
# ===== Generic Utility Methods =====
def _isObjectField(self, fieldType) -> bool:
"""Check if a field type represents a complex object (not a simple type)."""
# Simple scalar types
if fieldType in (str, int, float, bool, type(None)):
return False
# Everything else is an object
return True
def _separateObjectFields(self, model_class, data: Dict[str, Any]) -> tuple[Dict[str, Any], Dict[str, Any]]:
"""Separate simple fields from object fields based on Pydantic model structure."""
simpleFields = {}
objectFields = {}
# Get field information from the Pydantic model
modelFields = model_class.model_fields
for fieldName, value in data.items():
# Check if this field should be stored as JSONB in the database
if fieldName in modelFields:
fieldInfo = modelFields[fieldName]
# Pydantic v2 only
fieldType = fieldInfo.annotation
# Always route relational/object fields to object_fields for separate handling
# These fields are stored in separate normalized tables, not as JSONB
if fieldName in ['documents', 'stats', 'logs', 'messages']:
objectFields[fieldName] = value
continue
# Check if this is a JSONB field (Dict, List, or complex types)
# Purely type-based detection - no hardcoded field names
if (fieldType == dict or
fieldType == list or
(hasattr(fieldType, '__origin__') and fieldType.__origin__ in (dict, list))):
# Store as JSONB - include in simple_fields for database storage
simpleFields[fieldName] = value
elif isinstance(value, (str, int, float, bool, type(None))):
# Simple scalar types
simpleFields[fieldName] = value
else:
# Complex objects that should be filtered out
objectFields[fieldName] = value
else:
# Field not in model - treat as scalar if simple, otherwise filter out
# BUT: always include metadata fields (_createdBy, _createdAt, etc.) as they're handled by connector
if fieldName.startswith("_"):
# Metadata fields should be passed through to connector
simpleFields[fieldName] = value
elif isinstance(value, (str, int, float, bool, type(None))):
simpleFields[fieldName] = value
else:
objectFields[fieldName] = value
return simpleFields, objectFields
def _initializeServices(self):
pass
def setUserContext(self, currentUser: User, mandateId: Optional[str] = None, featureInstanceId: Optional[str] = None):
"""Sets the user context for the interface.
Args:
currentUser: The authenticated user
mandateId: The mandate ID from RequestContext (X-Mandate-Id header)
featureInstanceId: The feature instance ID from RequestContext (X-Feature-Instance-Id header)
"""
self.currentUser = currentUser # Store User object directly
self.userId = currentUser.id
# Use mandateId from parameter (Request-Context), not from user object
self.mandateId = mandateId
self.featureInstanceId = featureInstanceId
if not self.userId:
raise ValueError("Invalid user context: id is required")
# Note: mandateId is ALWAYS optional here - it comes from Request-Context, not from User.
# Users are NOT assigned to mandates by design - they get mandate context from the request.
# sysAdmin users can additionally perform cross-mandate operations.
# Without mandateId, operations will be filtered to accessible mandates via RBAC.
# Add language settings
self.userLanguage = currentUser.language # Default user language
# Initialize RBAC interface
if not self.currentUser:
raise ValueError("User context is required for RBAC")
# Get DbApp connection for RBAC AccessRule queries
from modules.security.rootAccess import getRootDbAppConnector
dbApp = getRootDbAppConnector()
self.rbac = RbacClass(self.db, dbApp=dbApp)
# Update database context
self.db.updateContext(self.userId)
def __del__(self):
"""Cleanup method to close database connection."""
if hasattr(self, 'db') and self.db is not None:
try:
self.db.close()
except Exception as e:
logger.error(f"Error closing database connection: {e}")
def _initializeDatabase(self):
"""Initializes the database connection directly."""
try:
# Get configuration values with defaults
dbHost = APP_CONFIG.get("DB_HOST", "_no_config_default_data")
dbDatabase = "poweron_chatbot"
dbUser = APP_CONFIG.get("DB_USER")
dbPassword = APP_CONFIG.get("DB_PASSWORD_SECRET")
dbPort = int(APP_CONFIG.get("DB_PORT", 5432))
from modules.connectors.connectorDbPostgre import _get_cached_connector
self.db = _get_cached_connector(
dbHost=dbHost,
dbDatabase=dbDatabase,
dbUser=dbUser,
dbPassword=dbPassword,
dbPort=dbPort,
userId=self.userId
)
logger.info("Database initialized successfully")
except Exception as e:
logger.error(f"Failed to initialize database: {str(e)}")
raise
def _initRecords(self):
"""Initializes standard records in the database if they don't exist."""
pass
def checkRbacPermission(
self,
modelClass: type,
operation: str,
recordId: Optional[str] = None
) -> bool:
"""
Check RBAC permission for a specific operation on a table.
Args:
modelClass: Pydantic model class for the table
operation: Operation to check ('create', 'update', 'delete', 'read')
recordId: Optional record ID for specific record check
Returns:
Boolean indicating permission
"""
if not self.rbac or not self.currentUser:
return False
tableName = modelClass.__name__
from modules.interfaces.interfaceRbac import buildDataObjectKey
objectKey = buildDataObjectKey(tableName, featureCode=getattr(self, 'featureCode', 'chatbot'))
permissions = self.rbac.getUserPermissions(
self.currentUser,
AccessRuleContext.DATA,
objectKey,
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId
)
if operation == "create":
return permissions.create != AccessLevel.NONE
elif operation == "update":
return permissions.update != AccessLevel.NONE
elif operation == "delete":
return permissions.delete != AccessLevel.NONE
elif operation == "read":
return permissions.read != AccessLevel.NONE
else:
return False
def _applyFilters(self, records: List[Dict[str, Any]], filters: Dict[str, Any]) -> List[Dict[str, Any]]:
"""
Apply filter criteria to records.
Supports:
- General search: {"search": "text"} - searches across all text fields
- Field-specific filters:
- Simple: {"status": "running"} - equals match
- With operator: {"status": {"operator": "equals", "value": "running"}}
- Operators: equals, contains, gt, gte, lt, lte, in, notIn, startsWith, endsWith
Args:
records: List of record dictionaries to filter
filters: Filter criteria dictionary
Returns:
Filtered list of records
"""
if not filters or not records:
return records
filtered = []
for record in records:
matches = True
# Handle general search across text fields
if "search" in filters:
search_term = str(filters["search"]).lower()
if search_term:
# Search in all string fields
found = False
for key, value in record.items():
if isinstance(value, str) and search_term in value.lower():
found = True
break
elif isinstance(value, (int, float)) and search_term in str(value):
found = True
break
if not found:
matches = False
# Handle field-specific filters
for field_name, filter_value in filters.items():
if field_name == "search":
continue # Already handled above
if field_name not in record:
matches = False
break
record_value = record.get(field_name)
# Handle simple value (equals operator)
if not isinstance(filter_value, dict):
if record_value != filter_value:
matches = False
break
continue
# Handle filter with operator
operator = filter_value.get("operator", "equals")
filter_val = filter_value.get("value")
if operator in ["equals", "eq"]:
if record_value != filter_val:
matches = False
break
elif operator == "contains":
record_str = str(record_value).lower() if record_value is not None else ""
filter_str = str(filter_val).lower() if filter_val is not None else ""
if filter_str not in record_str:
matches = False
break
elif operator == "startsWith":
record_str = str(record_value).lower() if record_value is not None else ""
filter_str = str(filter_val).lower() if filter_val is not None else ""
if not record_str.startswith(filter_str):
matches = False
break
elif operator == "endsWith":
record_str = str(record_value).lower() if record_value is not None else ""
filter_str = str(filter_val).lower() if filter_val is not None else ""
if not record_str.endswith(filter_str):
matches = False
break
elif operator == "gt":
try:
record_num = float(record_value) if record_value is not None else float('-inf')
filter_num = float(filter_val) if filter_val is not None else float('-inf')
if record_num <= filter_num:
matches = False
break
except (ValueError, TypeError):
matches = False
break
elif operator == "gte":
try:
record_num = float(record_value) if record_value is not None else float('-inf')
filter_num = float(filter_val) if filter_val is not None else float('-inf')
if record_num < filter_num:
matches = False
break
except (ValueError, TypeError):
matches = False
break
elif operator == "lt":
try:
record_num = float(record_value) if record_value is not None else float('inf')
filter_num = float(filter_val) if filter_val is not None else float('inf')
if record_num >= filter_num:
matches = False
break
except (ValueError, TypeError):
matches = False
break
elif operator == "lte":
try:
record_num = float(record_value) if record_value is not None else float('inf')
filter_num = float(filter_val) if filter_val is not None else float('inf')
if record_num > filter_num:
matches = False
break
except (ValueError, TypeError):
matches = False
break
elif operator == "in":
if not isinstance(filter_val, list):
filter_val = [filter_val]
if record_value not in filter_val:
matches = False
break
elif operator == "notIn":
if not isinstance(filter_val, list):
filter_val = [filter_val]
if record_value in filter_val:
matches = False
break
else:
# Unknown operator - default to equals
if record_value != filter_val:
matches = False
break
if matches:
filtered.append(record)
return filtered
def _applySorting(self, records: List[Dict[str, Any]], sortFields: List[Any]) -> List[Dict[str, Any]]:
"""Apply multi-level sorting to records using stable sort (sorts from least to most significant field)."""
if not sortFields:
return records
# Start with a copy to avoid modifying original
sortedRecords = list(records)
# Sort from least significant to most significant field (reverse order)
# Python's sort is stable, so this creates proper multi-level sorting
for sortField in reversed(sortFields):
# Handle both dict and object formats
if isinstance(sortField, dict):
fieldName = sortField.get("field")
direction = sortField.get("direction", "asc")
else:
fieldName = getattr(sortField, "field", None)
direction = getattr(sortField, "direction", "asc")
if not fieldName:
continue
isDesc = (direction == "desc")
def sortKey(record):
value = record.get(fieldName)
# Handle None values - place them at the end for both directions
if value is None:
# Use a special value that sorts last
return (1, "") # (is_none_flag, empty_value) - sorts after (0, ...)
else:
# Return tuple with type indicator for proper comparison
if isinstance(value, (int, float)):
return (0, value)
elif isinstance(value, str):
return (0, value)
elif isinstance(value, bool):
return (0, value)
else:
return (0, str(value))
# Sort with reverse parameter for descending
sortedRecords.sort(key=sortKey, reverse=isDesc)
return sortedRecords
# Utilities
def getInitialId(self, model_class: type) -> Optional[str]:
"""Returns the initial ID for a table."""
return self.db.getInitialId(model_class)
# Conversation methods (chatbot-specific, poweron_chatbot)
def getConversations(self, pagination: Optional[PaginationParams] = None) -> Union[List[Dict[str, Any]], PaginatedResult]:
"""
Returns conversations for current feature instance based on user access.
Supports optional pagination, sorting, and filtering.
Args:
pagination: Optional pagination parameters. If None, returns all items.
Returns:
If pagination is None: List[Dict[str, Any]]
If pagination is provided: PaginatedResult with items and metadata
"""
# Use RBAC filtering with featureInstanceId for instance-level isolation
filteredConversations = getRecordsetWithRBAC(self.db,
ChatbotConversation,
self.currentUser,
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot"
)
# If no pagination requested, return all items (no sorting - frontend handles it)
if pagination is None:
return filteredConversations
# Apply filtering (if filters provided)
if pagination.filters:
filteredConversations = self._applyFilters(filteredConversations, pagination.filters)
# Apply sorting (in order of sortFields) - only if provided by frontend
if pagination.sort:
filteredConversations = self._applySorting(filteredConversations, pagination.sort)
# Count total items after filters
totalItems = len(filteredConversations)
totalPages = math.ceil(totalItems / pagination.pageSize) if totalItems > 0 else 0
# Apply pagination (skip/limit)
startIdx = (pagination.page - 1) * pagination.pageSize
endIdx = startIdx + pagination.pageSize
pagedConversations = filteredConversations[startIdx:endIdx]
return PaginatedResult(
items=pagedConversations,
totalItems=totalItems,
totalPages=totalPages
)
def getWorkflows(self, pagination: Optional[PaginationParams] = None) -> Union[List[Dict[str, Any]], PaginatedResult]:
"""Backward-compat alias for getConversations."""
return self.getConversations(pagination)
def getConversation(self, conversationId: str) -> Optional[ChatbotConversation]:
"""Returns a conversation by ID if user has access."""
# Use RBAC filtering with featureInstanceId for instance-level isolation
conversations = getRecordsetWithRBAC(self.db,
ChatbotConversation,
self.currentUser,
recordFilter={"id": conversationId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot"
)
if not conversations:
return None
conv = conversations[0]
try:
# Load related data from normalized tables
logs = self.getLogs(conversationId)
messages = self.getMessages(conversationId)
# Build ChatbotConversation with hydrated logs/messages
return ChatbotConversation(
id=conv["id"],
featureInstanceId=conv.get("featureInstanceId") or self.featureInstanceId or "",
name=conv.get("name"),
status=conv.get("status", "running"),
currentRound=conv.get("currentRound", 0) or 0,
lastActivity=conv.get("lastActivity", getUtcTimestamp()),
startedAt=conv.get("startedAt", getUtcTimestamp()),
workflowMode=ChatbotWorkflowModeEnum(conv.get("workflowMode", "Chatbot")),
maxSteps=conv.get("maxSteps") if conv.get("maxSteps") is not None else 10,
logs=logs,
messages=messages
)
except Exception as e:
logger.error(f"Error validating conversation data: {str(e)}")
return None
def getWorkflow(self, workflowId: str) -> Optional[ChatbotConversation]:
"""Backward-compat alias: workflowId maps to conversationId."""
return self.getConversation(workflowId)
def getWorkflowMinimal(self, workflowId: str) -> Optional[ChatbotConversation]:
"""Lightweight fetch: conversation record only, no logs/messages. For resume path."""
conversations = getRecordsetWithRBAC(
self.db,
ChatbotConversation,
self.currentUser,
recordFilter={"id": workflowId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot",
)
if not conversations:
return None
conv = conversations[0]
max_steps = conv.get("maxSteps")
return ChatbotConversation(
id=conv["id"],
featureInstanceId=conv.get("featureInstanceId") or self.featureInstanceId or "",
name=conv.get("name"),
status=conv.get("status", "running"),
currentRound=conv.get("currentRound", 0) or 0,
lastActivity=conv.get("lastActivity", getUtcTimestamp()),
startedAt=conv.get("startedAt", getUtcTimestamp()),
workflowMode=ChatbotWorkflowModeEnum(conv.get("workflowMode", "Chatbot")),
maxSteps=max_steps if max_steps is not None else 10,
logs=[],
messages=[],
)
def getMessageCount(self, conversationId: str) -> int:
"""Returns message count for a conversation (single query, no document fetch)."""
messages = getRecordsetWithRBAC(
self.db,
ChatbotMessage,
self.currentUser,
recordFilter={"conversationId": conversationId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot",
)
return len(messages) if messages else 0
def updateWorkflowMinimal(
self, workflowId: str, workflowData: Dict[str, Any]
) -> ChatbotConversation:
"""Lightweight update: no logs/messages fetch. For resume when caller has minimal workflow."""
if not self.checkRbacPermission(ChatbotConversation, "update", workflowId):
raise PermissionError(f"No permission to update conversation {workflowId}")
simpleFields, _ = self._separateObjectFields(ChatbotConversation, workflowData)
simpleFields["lastActivity"] = getUtcTimestamp()
updated = self.db.recordModify(ChatbotConversation, workflowId, simpleFields)
max_steps = updated.get("maxSteps")
return ChatbotConversation(
id=updated["id"],
featureInstanceId=updated.get("featureInstanceId") or self.featureInstanceId or "",
name=updated.get("name"),
status=updated.get("status", "running"),
currentRound=updated.get("currentRound", 0) or 0,
lastActivity=updated.get("lastActivity", getUtcTimestamp()),
startedAt=updated.get("startedAt", getUtcTimestamp()),
workflowMode=ChatbotWorkflowModeEnum(updated.get("workflowMode", "Chatbot")),
maxSteps=max_steps if max_steps is not None else 10,
logs=[],
messages=[],
)
def createConversation(self, conversationData: Dict[str, Any]) -> ChatbotConversation:
"""Creates a new conversation if user has permission."""
if not self.checkRbacPermission(ChatbotConversation, "create"):
raise PermissionError("No permission to create conversations")
# Set timestamp if not present
currentTime = getUtcTimestamp()
if "startedAt" not in conversationData:
conversationData["startedAt"] = currentTime
if "lastActivity" not in conversationData:
conversationData["lastActivity"] = currentTime
# Set featureInstanceId from context (no mandateId in DB)
if "featureInstanceId" not in conversationData or not conversationData["featureInstanceId"]:
conversationData["featureInstanceId"] = self.featureInstanceId or ""
if not conversationData.get("featureInstanceId"):
conversationData["featureInstanceId"] = self.featureInstanceId or ""
# Use generic field separation - logs/messages go to objectFields, not stored
simpleFields, objectFields = self._separateObjectFields(ChatbotConversation, conversationData)
# Create conversation in database
created = self.db.recordCreate(ChatbotConversation, simpleFields)
return ChatbotConversation(
id=created["id"],
featureInstanceId=created.get("featureInstanceId") or self.featureInstanceId or "",
name=created.get("name"),
status=created.get("status", "running"),
currentRound=created.get("currentRound", 0) or 0,
lastActivity=created.get("lastActivity", currentTime),
startedAt=created.get("startedAt", currentTime),
workflowMode=ChatbotWorkflowModeEnum(created.get("workflowMode", "Chatbot")),
maxSteps=created.get("maxSteps", 10),
logs=[],
messages=[]
)
def createWorkflow(self, workflowData: Dict[str, Any]) -> ChatbotConversation:
"""Backward-compat alias: maps workflowData to conversationData."""
return self.createConversation(workflowData)
def updateConversation(self, conversationId: str, conversationData: Dict[str, Any]) -> ChatbotConversation:
"""Updates a conversation if user has access."""
conv = self.getConversation(conversationId)
if not conv:
return None
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
raise PermissionError(f"No permission to update conversation {conversationId}")
simpleFields, objectFields = self._separateObjectFields(ChatbotConversation, conversationData)
simpleFields["lastActivity"] = getUtcTimestamp()
updated = self.db.recordModify(ChatbotConversation, conversationId, simpleFields)
# Reuse logs/messages from conv — update only touches simple fields, not related data
return ChatbotConversation(
id=updated["id"],
featureInstanceId=updated.get("featureInstanceId") or conv.featureInstanceId or self.featureInstanceId or "",
name=updated.get("name", conv.name),
status=updated.get("status", conv.status),
currentRound=updated.get("currentRound", conv.currentRound),
lastActivity=updated.get("lastActivity", conv.lastActivity),
startedAt=updated.get("startedAt", conv.startedAt),
workflowMode=ChatbotWorkflowModeEnum(updated.get("workflowMode", conv.workflowMode.value)),
maxSteps=updated.get("maxSteps") if updated.get("maxSteps") is not None else conv.maxSteps,
logs=conv.logs,
messages=conv.messages
)
def updateWorkflow(self, workflowId: str, workflowData: Dict[str, Any]) -> ChatbotConversation:
"""Backward-compat alias."""
return self.updateConversation(workflowId, workflowData)
def deleteConversation(self, conversationId: str) -> bool:
"""Deletes a conversation and all related data if user has access."""
try:
conv = self.getConversation(conversationId)
if not conv:
return False
if not self.checkRbacPermission(ChatbotConversation, "delete", conversationId):
raise PermissionError(f"No permission to delete conversation {conversationId}")
# CASCADE DELETE: Delete all related data first
# 1. Delete all messages and their documents
messages = self.getMessages(conversationId)
for message in messages:
messageId = message.id
if messageId:
existing_docs = getRecordsetWithRBAC(self.db, ChatbotDocument, self.currentUser, recordFilter={"messageId": messageId}, featureCode="chatbot")
for doc in existing_docs:
self.db.recordDelete(ChatbotDocument, doc["id"])
self.db.recordDelete(ChatbotMessage, messageId)
# 2. Delete conversation logs
existing_logs = getRecordsetWithRBAC(self.db, ChatbotLog, self.currentUser, recordFilter={"conversationId": conversationId}, featureCode="chatbot")
for log in existing_logs:
self.db.recordDelete(ChatbotLog, log["id"])
# 3. Delete the conversation
success = self.db.recordDelete(ChatbotConversation, conversationId)
return success
except Exception as e:
logger.error(f"Error deleting conversation {conversationId}: {str(e)}")
return False
def deleteWorkflow(self, workflowId: str) -> bool:
"""Backward-compat alias."""
return self.deleteConversation(workflowId)
# Message methods
def getMessages(self, conversationId: str, pagination: Optional[PaginationParams] = None) -> Union[List[ChatbotMessage], PaginatedResult]:
"""
Returns messages for a conversation if user has access.
Supports optional pagination, sorting, and filtering.
Args:
conversationId: The conversation ID (workflowId for backward compat)
pagination: Optional pagination parameters. If None, returns all items.
Returns:
If pagination is None: List[ChatbotMessage]
If pagination is provided: PaginatedResult with items and metadata
"""
# Check conversation access first
conversations = getRecordsetWithRBAC(self.db,
ChatbotConversation,
self.currentUser,
recordFilter={"id": conversationId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot"
)
if not conversations:
if pagination is None:
return []
return PaginatedResult(items=[], totalItems=0, totalPages=0)
# Get messages for this conversation
messages = getRecordsetWithRBAC(self.db, ChatbotMessage, self.currentUser, recordFilter={"conversationId": conversationId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
# Convert raw messages to dict format for sorting/filtering
messageDicts = []
for msg in messages:
messageDicts.append({
"id": msg.get("id"),
"conversationId": msg.get("conversationId"),
"parentMessageId": msg.get("parentMessageId"),
"documentsLabel": msg.get("documentsLabel"),
"message": msg.get("message"),
"role": msg.get("role", "assistant"),
"status": msg.get("status", "step"),
"sequenceNr": msg.get("sequenceNr", 0),
"publishedAt": msg.get("publishedAt", msg.get("timestamp", getUtcTimestamp())),
"success": msg.get("success"),
"actionId": msg.get("actionId"),
"actionMethod": msg.get("actionMethod"),
"actionName": msg.get("actionName"),
"roundNumber": msg.get("roundNumber"),
"taskNumber": msg.get("taskNumber"),
"actionNumber": msg.get("actionNumber"),
"taskProgress": msg.get("taskProgress"),
"actionProgress": msg.get("actionProgress")
})
# Apply default sorting by publishedAt if no sort specified
if pagination is None or not pagination.sort:
messageDicts.sort(key=lambda x: x.get("publishedAt", getUtcTimestamp()))
# Apply filtering (if filters provided)
if pagination and pagination.filters:
messageDicts = self._applyFilters(messageDicts, pagination.filters)
# Apply sorting (in order of sortFields)
if pagination and pagination.sort:
messageDicts = self._applySorting(messageDicts, pagination.sort)
# If no pagination requested, return all items (batch-fetch documents to avoid N+1)
if pagination is None:
msg_ids = [m["id"] for m in messageDicts]
docs_by_message = self.getDocumentsForMessages(msg_ids) if msg_ids else {}
chat_messages = []
for msg in messageDicts:
documents = docs_by_message.get(msg["id"], [])
chat_message = ChatbotMessage(
id=msg["id"],
conversationId=msg["conversationId"],
parentMessageId=msg.get("parentMessageId"),
documents=documents,
documentsLabel=msg.get("documentsLabel"),
message=msg.get("message"),
role=msg.get("role", "assistant"),
status=msg.get("status", "step"),
sequenceNr=msg.get("sequenceNr", 0),
publishedAt=msg.get("publishedAt", getUtcTimestamp()),
success=msg.get("success"),
actionId=msg.get("actionId"),
actionMethod=msg.get("actionMethod"),
actionName=msg.get("actionName"),
roundNumber=msg.get("roundNumber"),
taskNumber=msg.get("taskNumber"),
actionNumber=msg.get("actionNumber"),
taskProgress=msg.get("taskProgress"),
actionProgress=msg.get("actionProgress")
)
chat_messages.append(chat_message)
return chat_messages
# Count total items after filters
totalItems = len(messageDicts)
totalPages = math.ceil(totalItems / pagination.pageSize) if totalItems > 0 else 0
# Apply pagination (skip/limit)
startIdx = (pagination.page - 1) * pagination.pageSize
endIdx = startIdx + pagination.pageSize
pagedMessageDicts = messageDicts[startIdx:endIdx]
paged_msg_ids = [m["id"] for m in pagedMessageDicts]
docs_by_message = self.getDocumentsForMessages(paged_msg_ids) if paged_msg_ids else {}
chat_messages = []
for msg in pagedMessageDicts:
documents = docs_by_message.get(msg["id"], [])
chat_message = ChatbotMessage(
id=msg["id"],
conversationId=msg["conversationId"],
parentMessageId=msg.get("parentMessageId"),
documents=documents,
documentsLabel=msg.get("documentsLabel"),
message=msg.get("message"),
role=msg.get("role", "assistant"),
status=msg.get("status", "step"),
sequenceNr=msg.get("sequenceNr", 0),
publishedAt=msg.get("publishedAt", getUtcTimestamp()),
success=msg.get("success"),
actionId=msg.get("actionId"),
actionMethod=msg.get("actionMethod"),
actionName=msg.get("actionName"),
roundNumber=msg.get("roundNumber"),
taskNumber=msg.get("taskNumber"),
actionNumber=msg.get("actionNumber"),
taskProgress=msg.get("taskProgress"),
actionProgress=msg.get("actionProgress")
)
chat_messages.append(chat_message)
return PaginatedResult(
items=chat_messages,
totalItems=totalItems,
totalPages=totalPages
)
def createMessage(self, messageData: Dict[str, Any], event_manager=None) -> ChatbotMessage:
"""Creates a message for a conversation if user has access. Accepts workflowId (from bridge) or conversationId."""
try:
if "id" not in messageData or not messageData["id"]:
messageData["id"] = f"msg_{uuid.uuid4()}"
# Map workflowId to conversationId (bridge compatibility)
if "workflowId" in messageData and "conversationId" not in messageData:
messageData["conversationId"] = messageData["workflowId"]
requiredFields = ["id", "conversationId"]
for field in requiredFields:
if field not in messageData:
logger.error(f"Required field '{field}' missing in messageData")
raise ValueError(f"Required field '{field}' missing in message data")
conversationId = messageData["conversationId"]
conv = self.getConversation(conversationId)
if not conv:
raise PermissionError(f"No access to conversation {conversationId}")
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
raise PermissionError(f"No permission to modify conversation {conversationId}")
# Validate that ID is not None
if messageData["id"] is None:
messageData["id"] = f"msg_{uuid.uuid4()}"
logger.warning(f"Automatically generated ID for workflow message: {messageData['id']}")
# Set status if not present
if "status" not in messageData:
messageData["status"] = "step" # Default status for intermediate messages
# Ensure role and agentName are present
if "role" not in messageData:
messageData["role"] = "assistant" if messageData.get("agentName") else "user"
if "agentName" not in messageData:
messageData["agentName"] = ""
# Set roundNumber, taskNumber, actionNumber if not provided
if "roundNumber" not in messageData:
messageData["roundNumber"] = conv.currentRound
if "taskNumber" not in messageData:
messageData["taskNumber"] = 0
if "actionNumber" not in messageData:
messageData["actionNumber"] = 0
# Use generic field separation - no mandateId/featureInstanceId in message
simpleFields, objectFields = self._separateObjectFields(ChatbotMessage, messageData)
# Handle documents separately - they will be stored in normalized documents table
documents_to_create = objectFields.get("documents", [])
createdMessage = self.db.recordCreate(ChatbotMessage, simpleFields)
created_documents = []
logger.debug(f"Creating {len(documents_to_create)} document(s) for message {createdMessage['id']}")
for idx, doc_data in enumerate(documents_to_create):
try:
if isinstance(doc_data, ChatbotDocument):
doc_dict = doc_data.model_dump()
elif isinstance(doc_data, dict):
doc_dict = dict(doc_data)
else:
try:
doc_dict = ChatbotDocument(**doc_data).model_dump()
except Exception as e:
logger.error(f"Invalid document data type for message creation (document {idx + 1}/{len(documents_to_create)}): {e}")
continue
# Ensure messageId is set
doc_dict["messageId"] = createdMessage["id"]
logger.debug(f"Creating document {idx + 1}/{len(documents_to_create)}: fileName={doc_dict.get('fileName', 'unknown')}, fileId={doc_dict.get('fileId', 'unknown')}, messageId={doc_dict.get('messageId', 'unknown')}")
created_doc = self.createDocument(doc_dict)
if created_doc:
created_documents.append(created_doc)
logger.debug(f"Successfully created document {idx + 1}/{len(documents_to_create)}: {created_doc.fileName} (id: {created_doc.id})")
else:
logger.error(f"Failed to create document {idx + 1}/{len(documents_to_create)}: createDocument returned None for fileName={doc_dict.get('fileName', 'unknown')}")
except Exception as e:
logger.error(f"Error processing document {idx + 1}/{len(documents_to_create)}: {e}", exc_info=True)
logger.info(f"Created {len(created_documents)}/{len(documents_to_create)} document(s) for message {createdMessage['id']}")
chat_message = ChatbotMessage(
id=createdMessage["id"],
conversationId=createdMessage["conversationId"],
parentMessageId=createdMessage.get("parentMessageId"),
documents=created_documents,
documentsLabel=createdMessage.get("documentsLabel"),
message=createdMessage.get("message"),
role=createdMessage.get("role", "assistant"),
status=createdMessage.get("status", "step"),
sequenceNr=len(conv.messages) + 1,
publishedAt=createdMessage.get("publishedAt", getUtcTimestamp()),
roundNumber=createdMessage.get("roundNumber"),
taskNumber=createdMessage.get("taskNumber"),
actionNumber=createdMessage.get("actionNumber"),
success=createdMessage.get("success"),
actionId=createdMessage.get("actionId"),
actionMethod=createdMessage.get("actionMethod"),
actionName=createdMessage.get("actionName")
)
if event_manager:
try:
message_timestamp = parseTimestamp(chat_message.publishedAt, default=getUtcTimestamp())
asyncio.create_task(event_manager.emit_event(
context_id=conversationId,
event_type="chatdata",
data={
"type": "message",
"createdAt": message_timestamp,
"item": chat_message.model_dump()
},
event_category="chat"
))
except Exception as e:
logger.debug(f"Could not emit message event: {e}")
# Debug: Store message and documents for debugging - only if debug enabled
storeDebugMessageAndDocuments(chat_message, self.currentUser, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId)
return chat_message
except Exception as e:
logger.error(f"Error creating workflow message: {str(e)}")
return None
def updateMessage(self, messageId: str, messageData: Dict[str, Any]) -> Optional[ChatbotMessage]:
"""Updates a conversation message if user has access."""
try:
if not messageId:
logger.error("No messageId provided for updateMessage")
raise ValueError("messageId cannot be empty")
if "workflowId" in messageData and "conversationId" not in messageData:
messageData["conversationId"] = messageData["workflowId"]
messages = getRecordsetWithRBAC(self.db, ChatbotMessage, self.currentUser, recordFilter={"id": messageId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
if not messages:
logger.warning(f"Message with ID {messageId} does not exist in database")
if "conversationId" in messageData:
conv = self.getConversation(messageData["conversationId"])
if not conv:
raise PermissionError(f"No access to conversation {messageData['conversationId']}")
if not self.checkRbacPermission(ChatbotConversation, "update", messageData["conversationId"]):
raise PermissionError(f"No permission to modify conversation")
logger.info(f"Creating new message with ID {messageId} for conversation {messageData['conversationId']}")
created = self.db.recordCreate(ChatbotMessage, messageData)
return ChatbotMessage(**created) if created else None
logger.error("Conversation ID missing for new message")
return None
existingMessage = messages[0]
conversationId = existingMessage.get("conversationId")
conv = self.getConversation(conversationId)
if not conv:
raise PermissionError(f"No access to conversation {conversationId}")
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
raise PermissionError(f"No permission to modify conversation {conversationId}")
simpleFields, objectFields = self._separateObjectFields(ChatbotMessage, messageData)
if "role" not in simpleFields and "role" not in existingMessage:
simpleFields["role"] = "assistant"
if 'id' not in simpleFields:
simpleFields['id'] = messageId
if "createdAt" in simpleFields and "startedAt" not in simpleFields:
simpleFields["startedAt"] = simpleFields["createdAt"]
del simpleFields["createdAt"]
updatedMessage = self.db.recordModify(ChatbotMessage, messageId, simpleFields)
if 'documents' in objectFields:
for doc_data in objectFields['documents']:
try:
if isinstance(doc_data, ChatbotDocument):
doc_dict = doc_data.model_dump()
elif isinstance(doc_data, dict):
doc_dict = dict(doc_data)
else:
doc_dict = ChatbotDocument(**doc_data).model_dump()
doc_dict["messageId"] = messageId
self.createDocument(doc_dict)
except Exception as e:
logger.error(f"Error updating message documents: {e}")
if not updatedMessage:
logger.warning(f"Failed to update message {messageId}")
return None
return ChatbotMessage(**updatedMessage)
except Exception as e:
logger.error(f"Error updating message {messageId}: {str(e)}", exc_info=True)
raise ValueError(f"Error updating message {messageId}: {str(e)}")
def deleteMessage(self, conversationId: str, messageId: str) -> bool:
"""Deletes a conversation message and related data if user has access."""
try:
conv = self.getConversation(conversationId)
if not conv:
logger.warning(f"No access to conversation {conversationId}")
return False
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
raise PermissionError(f"No permission to modify conversation {conversationId}")
messages = self.getMessages(conversationId)
message = next((m for m in messages if m.id == messageId), None)
if not message:
logger.warning(f"Message {messageId} for conversation {conversationId} not found")
return False
existing_docs = getRecordsetWithRBAC(self.db, ChatbotDocument, self.currentUser, recordFilter={"messageId": messageId}, featureCode="chatbot")
for doc in existing_docs:
self.db.recordDelete(ChatbotDocument, doc["id"])
success = self.db.recordDelete(ChatbotMessage, messageId)
return success
except Exception as e:
logger.error(f"Error deleting message {messageId}: {str(e)}")
return False
def deleteFileFromMessage(self, conversationId: str, messageId: str, fileId: str) -> bool:
"""Removes a file reference from a message if user has access."""
try:
conv = self.getConversation(conversationId)
if not conv:
logger.warning(f"No access to conversation {conversationId}")
return False
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
raise PermissionError(f"No permission to modify conversation {conversationId}")
documents = getRecordsetWithRBAC(self.db, ChatbotDocument, self.currentUser, recordFilter={"messageId": messageId}, featureCode="chatbot")
if not documents:
logger.warning(f"No documents found for message {messageId}")
return False
# Find and delete the specific document
removed = False
for doc in documents:
docId = doc.get("id")
fileIdValue = doc.get("fileId")
# Flexible matching approach
shouldRemove = (
(docId == fileId) or
(fileIdValue == fileId) or
(isinstance(docId, str) and str(fileId) in docId) or
(isinstance(fileIdValue, str) and str(fileId) in fileIdValue)
)
if shouldRemove:
# Delete the document from normalized table
success = self.db.recordDelete(ChatbotDocument, docId)
if success:
removed = True
else:
logger.warning(f"Failed to delete document {docId}")
if not removed:
logger.warning(f"No matching file {fileId} found in message {messageId}")
return False
except Exception as e:
logger.error(f"Error removing file {fileId} from message {messageId}: {str(e)}")
return False
# Document methods
def getDocuments(self, messageId: str) -> List[ChatbotDocument]:
"""Returns documents for a message from normalized table."""
try:
documents = getRecordsetWithRBAC(self.db, ChatbotDocument, self.currentUser, recordFilter={"messageId": messageId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
return [ChatbotDocument(**doc) for doc in documents]
except Exception as e:
logger.error(f"Error getting message documents: {str(e)}")
return []
def getDocumentsForMessages(self, messageIds: List[str]) -> Dict[str, List[ChatbotDocument]]:
"""Returns documents for multiple messages in one query. Returns {messageId: [doc, ...]}."""
if not messageIds:
return {}
try:
documents = getRecordsetWithRBAC(
self.db,
ChatbotDocument,
self.currentUser,
recordFilter={"messageId": messageIds},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot",
)
result: Dict[str, List[ChatbotDocument]] = {mid: [] for mid in messageIds}
for doc in documents:
mid = doc.get("messageId")
if mid in result:
result[mid].append(ChatbotDocument(**doc))
return result
except Exception as e:
logger.error(f"Error getting documents for messages: {e}")
return {mid: [] for mid in messageIds}
def createDocument(self, documentData: Dict[str, Any]) -> ChatbotDocument:
"""Creates a document for a message in normalized table."""
try:
document = ChatbotDocument(**documentData)
logger.debug(f"Creating document: fileName={document.fileName}, fileId={document.fileId}, messageId={document.messageId}")
created = self.db.recordCreate(ChatbotDocument, document.model_dump())
if created:
return ChatbotDocument(**created)
logger.error(f"Failed to create document for fileName={document.fileName}")
return None
except Exception as e:
logger.error(f"Error creating message document: {str(e)}", exc_info=True)
return None
# Log methods
def getLogs(self, conversationId: str, pagination: Optional[PaginationParams] = None) -> Union[List[ChatbotLog], PaginatedResult]:
"""
Returns logs for a conversation if user has access.
Supports optional pagination, sorting, and filtering.
"""
conversations = getRecordsetWithRBAC(self.db,
ChatbotConversation,
self.currentUser,
recordFilter={"id": conversationId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot"
)
if not conversations:
if pagination is None:
return []
return PaginatedResult(items=[], totalItems=0, totalPages=0)
logs = getRecordsetWithRBAC(self.db, ChatbotLog, self.currentUser, recordFilter={"conversationId": conversationId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
logDicts = []
for log in logs:
logDicts.append({
"id": log.get("id"),
"conversationId": log.get("conversationId"),
"message": log.get("message"),
"type": log.get("type"),
"timestamp": log.get("timestamp", getUtcTimestamp()),
"status": log.get("status"),
"progress": log.get("progress")
})
# Apply default sorting by timestamp if no sort specified
if pagination is None or not pagination.sort:
logDicts.sort(key=lambda x: parseTimestamp(x.get("timestamp"), default=0))
# Apply filtering (if filters provided)
if pagination and pagination.filters:
logDicts = self._applyFilters(logDicts, pagination.filters)
# Apply sorting (in order of sortFields)
if pagination and pagination.sort:
logDicts = self._applySorting(logDicts, pagination.sort)
if pagination is None:
return [ChatbotLog(**log) for log in logDicts]
totalItems = len(logDicts)
totalPages = math.ceil(totalItems / pagination.pageSize) if totalItems > 0 else 0
startIdx = (pagination.page - 1) * pagination.pageSize
endIdx = startIdx + pagination.pageSize
pagedLogDicts = logDicts[startIdx:endIdx]
items = [ChatbotLog(**log) for log in pagedLogDicts]
return PaginatedResult(
items=items,
totalItems=totalItems,
totalPages=totalPages
)
def createLog(self, logData: Dict[str, Any], event_manager=None) -> Optional[ChatbotLog]:
"""Creates a log entry for a conversation if user has access. Accepts workflowId for backward compat."""
conversationId = logData.get("conversationId") or logData.get("workflowId")
if not conversationId:
logger.error("No conversationId/workflowId provided for createLog")
return None
conv = self.getConversation(conversationId)
if not conv:
logger.warning(f"No access to conversation {conversationId}")
return None
if not self.checkRbacPermission(ChatbotConversation, "update", conversationId):
logger.warning(f"No permission to modify conversation {conversationId}")
return None
if "timestamp" not in logData:
logData["timestamp"] = getUtcTimestamp()
logData["conversationId"] = conversationId
if "status" not in logData and "type" in logData:
logData["status"] = "error" if logData["type"] == "error" else "running"
if "progress" not in logData:
logData["progress"] = 1.0 if logData.get("type") == "error" else 0.5
try:
log_model = ChatbotLog(**logData)
except Exception as e:
logger.error(f"Invalid log data: {e}")
return None
createdLog = self.db.recordCreate(ChatbotLog, log_model.model_dump())
if not createdLog:
return None
if event_manager:
try:
log_timestamp = parseTimestamp(createdLog.get("timestamp"), default=getUtcTimestamp())
asyncio.create_task(event_manager.emit_event(
context_id=conversationId,
event_type="chatdata",
data={"type": "log", "createdAt": log_timestamp, "item": ChatbotLog(**createdLog).model_dump()},
event_category="log",
message="New log"
))
except Exception as e:
logger.debug(f"Could not emit log event: {e}")
return ChatbotLog(**createdLog)
def getUnifiedChatData(self, conversationId: str, afterTimestamp: Optional[float] = None) -> Dict[str, Any]:
"""
Returns unified chat data (messages, logs) for a conversation in chronological order.
Uses timestamp-based selective data transfer for efficient polling.
"""
conversations = getRecordsetWithRBAC(self.db,
ChatbotConversation,
self.currentUser,
recordFilter={"id": conversationId},
mandateId=self.mandateId,
featureInstanceId=self.featureInstanceId,
featureCode="chatbot"
)
if not conversations:
return {"items": []}
items = []
messages = getRecordsetWithRBAC(self.db, ChatbotMessage, self.currentUser, recordFilter={"conversationId": conversationId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
# Batch-fetch documents for all messages (avoids N+1)
message_ids = [m["id"] for m in messages if afterTimestamp is None or parseTimestamp(m.get("publishedAt"), default=getUtcTimestamp()) > afterTimestamp]
docs_by_message = self.getDocumentsForMessages(message_ids) if message_ids else {}
for msg in messages:
msgTimestamp = parseTimestamp(msg.get("publishedAt"), default=getUtcTimestamp())
if afterTimestamp is not None and msgTimestamp <= afterTimestamp:
continue
documents = docs_by_message.get(msg["id"], [])
chatMessage = ChatbotMessage(
id=msg["id"],
conversationId=msg["conversationId"],
parentMessageId=msg.get("parentMessageId"),
documents=documents,
documentsLabel=msg.get("documentsLabel"),
message=msg.get("message"),
role=msg.get("role", "assistant"),
status=msg.get("status", "step"),
sequenceNr=msg.get("sequenceNr", 0),
publishedAt=msg.get("publishedAt", getUtcTimestamp()),
success=msg.get("success"),
actionId=msg.get("actionId"),
actionMethod=msg.get("actionMethod"),
actionName=msg.get("actionName"),
roundNumber=msg.get("roundNumber"),
taskNumber=msg.get("taskNumber"),
actionNumber=msg.get("actionNumber"),
taskProgress=msg.get("taskProgress"),
actionProgress=msg.get("actionProgress")
)
items.append({"type": "message", "createdAt": msgTimestamp, "item": chatMessage})
logs = getRecordsetWithRBAC(self.db, ChatbotLog, self.currentUser, recordFilter={"conversationId": conversationId}, mandateId=self.mandateId, featureInstanceId=self.featureInstanceId, featureCode="chatbot")
for log in logs:
logTimestamp = parseTimestamp(log.get("timestamp"), default=getUtcTimestamp())
if afterTimestamp is not None and logTimestamp <= afterTimestamp:
continue
chatLog = ChatbotLog(**log)
items.append({"type": "log", "createdAt": logTimestamp, "item": chatLog})
items.sort(key=lambda x: parseTimestamp(x.get("createdAt"), default=0))
return {"items": items}
def getInterface(currentUser: Optional[User] = None, mandateId: Optional[str] = None, featureInstanceId: Optional[str] = None) -> 'ChatObjects':
"""
Returns a ChatObjects instance for the current user.
Handles initialization of database and records.
Args:
currentUser: The authenticated user
mandateId: The mandate ID from RequestContext (X-Mandate-Id header). Required.
featureInstanceId: The feature instance ID from RequestContext (X-Feature-Instance-Id header).
"""
if not currentUser:
raise ValueError("Invalid user context: user is required")
effectiveMandateId = str(mandateId) if mandateId else None
effectiveFeatureInstanceId = str(featureInstanceId) if featureInstanceId else None
# Create context key including featureInstanceId for proper isolation
contextKey = f"{effectiveMandateId}_{effectiveFeatureInstanceId}_{currentUser.id}"
# Create new instance if not exists
if contextKey not in _chatInterfaces:
_chatInterfaces[contextKey] = ChatObjects(currentUser, mandateId=effectiveMandateId, featureInstanceId=effectiveFeatureInstanceId)
else:
# Update user context if needed
_chatInterfaces[contextKey].setUserContext(currentUser, mandateId=effectiveMandateId, featureInstanceId=effectiveFeatureInstanceId)
return _chatInterfaces[contextKey]