From 4e843761a91709287727bfa8b3b21e589077496f Mon Sep 17 00:00:00 2001 From: ValueOn AG Date: Tue, 17 Mar 2026 22:51:05 +0100 Subject: [PATCH] Cleaned backend and enhanced ai call failover modes --- modules/aicore/aicoreBase.py | 25 + modules/aicore/aicoreModelSelector.py | 23 +- modules/aicore/aicorePluginAnthropic.py | 20 +- modules/aicore/aicorePluginMistral.py | 35 +- modules/aicore/aicorePluginOpenai.py | 33 +- modules/connectors/connectorProviderBase.py | 15 +- .../providerGoogle/connectorGoogle.py | 39 +- .../connectors/providerMsft/connectorMsft.py | 26 +- modules/datamodels/datamodelBilling.py | 4 +- modules/datamodels/datamodelChat.py | 8 - .../automation/routeFeatureAutomation.py | 44 +- modules/features/chatplayground/__init__.py | 6 - .../interfaceFeatureChatplayground.py | 137 ---- .../chatplayground/mainChatplayground.py | 381 --------- .../routeFeatureChatplayground.py | 722 ------------------ modules/features/codeeditor/__init__.py | 1 - .../codeeditor/codeEditorProcessor.py | 280 ------- .../codeeditor/datamodelCodeeditor.py | 122 --- .../features/codeeditor/fileContextManager.py | 84 -- modules/features/codeeditor/mainCodeeditor.py | 248 ------ modules/features/codeeditor/promptAssembly.py | 183 ----- modules/features/codeeditor/responseParser.py | 184 ----- .../codeeditor/routeFeatureCodeeditor.py | 395 ---------- modules/features/codeeditor/toolRegistry.py | 157 ---- modules/features/workspace/__init__.py | 2 +- modules/features/workspace/mainWorkspace.py | 15 +- .../workspace/routeFeatureWorkspace.py | 209 ++++- modules/interfaces/interfaceAiObjects.py | 187 ++++- modules/interfaces/interfaceBootstrap.py | 36 +- modules/interfaces/interfaceDbManagement.py | 4 +- modules/routes/routeSystem.py | 8 +- .../serviceAgent/actionToolAdapter.py | 8 +- .../services/serviceAgent/datamodelAgent.py | 19 +- .../services/serviceAgent/mainServiceAgent.py | 94 ++- .../services/serviceAi/mainServiceAi.py | 15 +- .../serviceBilling/mainServiceBilling.py | 2 +- .../serviceKnowledge/mainServiceKnowledge.py | 98 ++- modules/system/mainSystem.py | 10 - modules/workflows/automation/__init__.py | 6 +- modules/workflows/automation/mainWorkflow.py | 12 +- scripts/migrate_async_to_sync.py | 5 - tests/functional/test03_ai_operations.py | 2 +- 42 files changed, 811 insertions(+), 3093 deletions(-) delete mode 100644 modules/features/chatplayground/__init__.py delete mode 100644 modules/features/chatplayground/interfaceFeatureChatplayground.py delete mode 100644 modules/features/chatplayground/mainChatplayground.py delete mode 100644 modules/features/chatplayground/routeFeatureChatplayground.py delete mode 100644 modules/features/codeeditor/__init__.py delete mode 100644 modules/features/codeeditor/codeEditorProcessor.py delete mode 100644 modules/features/codeeditor/datamodelCodeeditor.py delete mode 100644 modules/features/codeeditor/fileContextManager.py delete mode 100644 modules/features/codeeditor/mainCodeeditor.py delete mode 100644 modules/features/codeeditor/promptAssembly.py delete mode 100644 modules/features/codeeditor/responseParser.py delete mode 100644 modules/features/codeeditor/routeFeatureCodeeditor.py delete mode 100644 modules/features/codeeditor/toolRegistry.py diff --git a/modules/aicore/aicoreBase.py b/modules/aicore/aicoreBase.py index 2f998f1d..70dd67c4 100644 --- a/modules/aicore/aicoreBase.py +++ b/modules/aicore/aicoreBase.py @@ -11,11 +11,36 @@ IMPORTANT: Model Registration Requirements - If duplicate displayNames are detected during registration, an error will be raised """ +import re as _re + from abc import ABC, abstractmethod from typing import List, Dict, Any, Optional, AsyncGenerator, Union from modules.datamodels.datamodelAi import AiModel, AiModelCall, AiModelResponse +_RETRY_AFTER_PATTERN = _re.compile(r"try again in (\d+(?:\.\d+)?)\s*s", _re.IGNORECASE) + + +def _parseRetryAfterSeconds(message: str) -> float: + """Extract retry-after seconds from provider error messages like 'Please try again in 6.558s'.""" + match = _RETRY_AFTER_PATTERN.search(message) + return float(match.group(1)) if match else 0.0 + + +class RateLimitExceededException(Exception): + """Raised when a provider's rate limit (TPM / RPM) is exceeded.""" + def __init__(self, message: str = "Rate limit exceeded", retryAfterSeconds: float = 0.0): + super().__init__(message) + if retryAfterSeconds <= 0: + retryAfterSeconds = _parseRetryAfterSeconds(message) + self.retryAfterSeconds = retryAfterSeconds + + +class ContextLengthExceededException(Exception): + """Raised when the input exceeds a model's context window.""" + pass + + class BaseConnectorAi(ABC): """ Base class for all AI connectors. diff --git a/modules/aicore/aicoreModelSelector.py b/modules/aicore/aicoreModelSelector.py index ea0a4198..d3df1e45 100644 --- a/modules/aicore/aicoreModelSelector.py +++ b/modules/aicore/aicoreModelSelector.py @@ -19,26 +19,29 @@ class ModelSelector: """Model selector with priority scoring and recent-failure cooldown.""" def __init__(self): - self._failureLog: Dict[str, float] = {} + self._failureLog: Dict[str, Tuple[float, float]] = {} logger.info("ModelSelector initialized with failure cooldown support") - def reportFailure(self, modelName: str): + def reportFailure(self, modelName: str, cooldownSeconds: float = 0.0): """Record that a model just failed (rate limit, error, etc.). - The model will be deprioritized for COOLDOWN_DURATION seconds.""" - self._failureLog[modelName] = time.time() - logger.info(f"ModelSelector: Recorded failure for {modelName}, cooldown {_COOLDOWN_DURATION}s") + The model will be deprioritized for *cooldownSeconds* (default: _COOLDOWN_DURATION).""" + if cooldownSeconds <= 0: + cooldownSeconds = _COOLDOWN_DURATION + self._failureLog[modelName] = (time.time(), cooldownSeconds) + logger.info(f"ModelSelector: Recorded failure for {modelName}, cooldown {cooldownSeconds:.1f}s") def _getCooldownPenalty(self, modelName: str) -> float: """Return a score penalty (0.0 = no penalty, large negative = recently failed).""" - failedAt = self._failureLog.get(modelName) - if failedAt is None: + entry = self._failureLog.get(modelName) + if entry is None: return 0.0 + failedAt, cooldown = entry elapsed = time.time() - failedAt - if elapsed > _COOLDOWN_DURATION: + if elapsed > cooldown: del self._failureLog[modelName] return 0.0 - remaining = _COOLDOWN_DURATION - elapsed - return -(remaining / _COOLDOWN_DURATION) * 5000.0 + remaining = cooldown - elapsed + return -(remaining / cooldown) * 5000.0 def selectModel(self, prompt: str, diff --git a/modules/aicore/aicorePluginAnthropic.py b/modules/aicore/aicorePluginAnthropic.py index 85bbfa75..8b6ec197 100644 --- a/modules/aicore/aicorePluginAnthropic.py +++ b/modules/aicore/aicorePluginAnthropic.py @@ -7,7 +7,7 @@ import os from typing import Dict, Any, List, AsyncGenerator, Union from fastapi import HTTPException from modules.shared.configuration import APP_CONFIG -from .aicoreBase import BaseConnectorAi +from .aicoreBase import BaseConnectorAi, RateLimitExceededException from modules.datamodels.datamodelAi import AiModel, PriorityEnum, ProcessingModeEnum, OperationTypeEnum, AiModelCall, AiModelResponse, createOperationTypeRatings # Configure logger @@ -203,11 +203,12 @@ class AiAnthropic(BaseConnectorAi): error_detail = f"Anthropic API error: {response.status_code} - {response.text}" logger.error(error_detail) - # Provide more specific error messages based on status code + if response.status_code == 429: + raise RateLimitExceededException( + f"Rate limit exceeded for {model.name}: {response.text}" + ) if response.status_code == 529: error_message = "Anthropic API is currently overloaded. Please try again in a few minutes." - elif response.status_code == 429: - error_message = "Rate limit exceeded. Please wait before making another request." elif response.status_code == 401: error_message = "Invalid API key. Please check your Anthropic API configuration." elif response.status_code == 400: @@ -255,6 +256,8 @@ class AiAnthropic(BaseConnectorAi): metadata=metadata ) + except (RateLimitExceededException, HTTPException): + raise except Exception as e: error_msg = str(e) if str(e) else f"{type(e).__name__}" error_detail = f"Error calling Anthropic API: {error_msg}" @@ -296,7 +299,12 @@ class AiAnthropic(BaseConnectorAi): async with self.httpClient.stream("POST", model.apiUrl, json=payload) as response: if response.status_code != 200: body = await response.aread() - raise HTTPException(status_code=500, detail=f"Anthropic stream error: {response.status_code} - {body.decode()}") + bodyStr = body.decode() + if response.status_code == 429: + raise RateLimitExceededException( + f"Rate limit exceeded for {model.name}: {bodyStr}" + ) + raise HTTPException(status_code=500, detail=f"Anthropic stream error: {response.status_code} - {bodyStr}") async for line in response.aiter_lines(): if not line.startswith("data: "): @@ -354,7 +362,7 @@ class AiAnthropic(BaseConnectorAi): metadata=metadata, ) - except HTTPException: + except (RateLimitExceededException, HTTPException): raise except Exception as e: logger.error(f"Error streaming Anthropic API: {e}", exc_info=True) diff --git a/modules/aicore/aicorePluginMistral.py b/modules/aicore/aicorePluginMistral.py index a4f0e476..8c4fb6d9 100644 --- a/modules/aicore/aicorePluginMistral.py +++ b/modules/aicore/aicorePluginMistral.py @@ -6,20 +6,11 @@ import httpx from typing import List, Dict, Any, AsyncGenerator, Union from fastapi import HTTPException from modules.shared.configuration import APP_CONFIG -from .aicoreBase import BaseConnectorAi +from .aicoreBase import BaseConnectorAi, RateLimitExceededException, ContextLengthExceededException from modules.datamodels.datamodelAi import AiModel, PriorityEnum, ProcessingModeEnum, OperationTypeEnum, AiModelCall, AiModelResponse, createOperationTypeRatings -# Configure logger logger = logging.getLogger(__name__) -class ContextLengthExceededException(Exception): - """Exception raised when the context length exceeds the model's limit""" - pass - -class RateLimitExceededException(Exception): - """Exception raised when the provider's rate limit (TPM) is exceeded""" - pass - def loadConfigData(): """Load configuration data for Mistral connector""" return { @@ -264,7 +255,14 @@ class AiMistral(BaseConnectorAi): async with self.httpClient.stream("POST", model.apiUrl, json=payload) as response: if response.status_code != 200: body = await response.aread() - raise HTTPException(status_code=500, detail=f"Mistral stream error: {response.status_code} - {body.decode()}") + bodyStr = body.decode() + if response.status_code == 429: + try: + errorMsg = _json.loads(bodyStr).get("error", {}).get("message", "Rate limit exceeded") + except (ValueError, KeyError): + errorMsg = f"Rate limit exceeded for {model.name}" + raise RateLimitExceededException(f"Rate limit exceeded for {model.name}: {errorMsg}") + raise HTTPException(status_code=500, detail=f"Mistral stream error: {response.status_code} - {bodyStr}") async for line in response.aiter_lines(): if not line.startswith("data: "): @@ -289,7 +287,7 @@ class AiMistral(BaseConnectorAi): metadata={}, ) - except HTTPException: + except (RateLimitExceededException, ContextLengthExceededException, HTTPException): raise except Exception as e: logger.error(f"Error streaming Mistral API: {e}") @@ -317,6 +315,17 @@ class AiMistral(BaseConnectorAi): logger.error(errorMessage) if response.status_code == 429: raise RateLimitExceededException(f"Rate limit exceeded for {model.name}") + if response.status_code == 400: + try: + errorData = response.json() + errMsg = errorData.get("error", {}).get("message", "").lower() + errCode = errorData.get("error", {}).get("code", "") + if errCode == "context_length_exceeded" or "too many tokens" in errMsg or "maximum context length" in errMsg: + raise ContextLengthExceededException( + f"Embedding context length exceeded for {model.name}: {errorData.get('error', {}).get('message', '')}" + ) + except (ValueError, KeyError): + pass raise HTTPException(status_code=500, detail=errorMessage) responseJson = response.json() @@ -334,7 +343,7 @@ class AiMistral(BaseConnectorAi): }, metadata={"embeddings": embeddings}, ) - except RateLimitExceededException: + except (RateLimitExceededException, ContextLengthExceededException): raise except Exception as e: logger.error(f"Error calling Mistral Embedding API: {str(e)}") diff --git a/modules/aicore/aicorePluginOpenai.py b/modules/aicore/aicorePluginOpenai.py index 366f7dde..3b9f2c5f 100644 --- a/modules/aicore/aicorePluginOpenai.py +++ b/modules/aicore/aicorePluginOpenai.py @@ -6,20 +6,11 @@ import httpx from typing import List, Dict, Any, AsyncGenerator, Union from fastapi import HTTPException from modules.shared.configuration import APP_CONFIG -from .aicoreBase import BaseConnectorAi +from .aicoreBase import BaseConnectorAi, RateLimitExceededException, ContextLengthExceededException from modules.datamodels.datamodelAi import AiModel, PriorityEnum, ProcessingModeEnum, OperationTypeEnum, AiModelCall, AiModelResponse, createOperationTypeRatings, AiCallPromptImage -# Configure logger logger = logging.getLogger(__name__) -class ContextLengthExceededException(Exception): - """Exception raised when the context length exceeds the model's limit""" - pass - -class RateLimitExceededException(Exception): - """Exception raised when the provider's rate limit (TPM) is exceeded""" - pass - def loadConfigData(): """Load configuration data for OpenAI connector""" return { @@ -316,7 +307,14 @@ class AiOpenai(BaseConnectorAi): async with self.httpClient.stream("POST", model.apiUrl, json=payload) as response: if response.status_code != 200: body = await response.aread() - raise HTTPException(status_code=500, detail=f"OpenAI stream error: {response.status_code} - {body.decode()}") + bodyStr = body.decode() + if response.status_code == 429: + try: + errorMsg = _json.loads(bodyStr).get("error", {}).get("message", "Rate limit exceeded") + except (ValueError, KeyError): + errorMsg = f"Rate limit exceeded for {model.name}" + raise RateLimitExceededException(f"Rate limit exceeded for {model.name}: {errorMsg}") + raise HTTPException(status_code=500, detail=f"OpenAI stream error: {response.status_code} - {bodyStr}") async for line in response.aiter_lines(): if not line.startswith("data: "): @@ -362,7 +360,7 @@ class AiOpenai(BaseConnectorAi): metadata=metadata, ) - except HTTPException: + except (RateLimitExceededException, ContextLengthExceededException, HTTPException): raise except Exception as e: logger.error(f"Error streaming OpenAI API: {e}") @@ -390,6 +388,17 @@ class AiOpenai(BaseConnectorAi): logger.error(errorMessage) if response.status_code == 429: raise RateLimitExceededException(f"Rate limit exceeded for {model.name}") + if response.status_code == 400: + try: + errorData = response.json() + errMsg = errorData.get("error", {}).get("message", "").lower() + errCode = errorData.get("error", {}).get("code", "") + if errCode == "context_length_exceeded" or "too many tokens" in errMsg or "maximum context length" in errMsg: + raise ContextLengthExceededException( + f"Embedding context length exceeded for {model.name}: {errorData.get('error', {}).get('message', '')}" + ) + except (ValueError, KeyError): + pass raise HTTPException(status_code=500, detail=errorMessage) responseJson = response.json() diff --git a/modules/connectors/connectorProviderBase.py b/modules/connectors/connectorProviderBase.py index 71ad0ecf..107fe1c4 100644 --- a/modules/connectors/connectorProviderBase.py +++ b/modules/connectors/connectorProviderBase.py @@ -8,7 +8,16 @@ All ServiceAdapters share the same access token from the UserConnection. """ from abc import ABC, abstractmethod -from typing import List, Optional +from dataclasses import dataclass, field +from typing import List, Optional, Union + + +@dataclass +class DownloadResult: + """Rich return type for ServiceAdapter.download() when metadata is available.""" + data: bytes = field(default=b"", repr=False) + fileName: str = "" + mimeType: str = "" class ServiceAdapter(ABC): @@ -20,8 +29,8 @@ class ServiceAdapter(ABC): ... @abstractmethod - async def download(self, path: str) -> bytes: - """Download a file and return its content bytes.""" + async def download(self, path: str) -> Union[bytes, DownloadResult]: + """Download a file. Return bytes or DownloadResult with metadata.""" ... @abstractmethod diff --git a/modules/connectors/providerGoogle/connectorGoogle.py b/modules/connectors/providerGoogle/connectorGoogle.py index 216b9019..3928fa70 100644 --- a/modules/connectors/providerGoogle/connectorGoogle.py +++ b/modules/connectors/providerGoogle/connectorGoogle.py @@ -7,7 +7,7 @@ from typing import Any, Dict, List, Optional import aiohttp -from modules.connectors.connectorProviderBase import ProviderConnector, ServiceAdapter +from modules.connectors.connectorProviderBase import ProviderConnector, ServiceAdapter, DownloadResult from modules.datamodels.datamodelDataSource import ExternalEntry logger = logging.getLogger(__name__) @@ -192,8 +192,41 @@ class GmailAdapter(ServiceAdapter): )) return entries - async def download(self, path: str) -> bytes: - return b"" + async def download(self, path: str) -> DownloadResult: + """Download a Gmail message as RFC 822 EML via format=raw.""" + import base64 + import re + cleanPath = (path or "").strip("/") + msgId = cleanPath.split("/")[-1] if cleanPath else "" + if not msgId: + return DownloadResult() + + url = f"{_GMAIL_BASE}/users/me/messages/{msgId}?format=raw" + result = await _googleGet(self._token, url) + if "error" in result: + return DownloadResult() + + rawB64 = result.get("raw", "") + if not rawB64: + return DownloadResult() + + emlBytes = base64.urlsafe_b64decode(rawB64) + + metaUrl = f"{_GMAIL_BASE}/users/me/messages/{msgId}?format=metadata&metadataHeaders=Subject" + meta = await _googleGet(self._token, metaUrl) + subject = msgId + if "error" not in meta: + for h in meta.get("payload", {}).get("headers", []): + if h.get("name", "").lower() == "subject": + subject = h.get("value", msgId) + break + safeName = re.sub(r'[<>:"/\\|?*\x00-\x1f]', "_", subject)[:80].strip(". ") or "email" + + return DownloadResult( + data=emlBytes, + fileName=f"{safeName}.eml", + mimeType="message/rfc822", + ) async def upload(self, path: str, data: bytes, fileName: str) -> dict: return {"error": "Gmail upload not applicable"} diff --git a/modules/connectors/providerMsft/connectorMsft.py b/modules/connectors/providerMsft/connectorMsft.py index 26aa3790..3654bad9 100644 --- a/modules/connectors/providerMsft/connectorMsft.py +++ b/modules/connectors/providerMsft/connectorMsft.py @@ -11,7 +11,7 @@ import aiohttp import asyncio from typing import Dict, Any, List, Optional -from modules.connectors.connectorProviderBase import ProviderConnector, ServiceAdapter +from modules.connectors.connectorProviderBase import ProviderConnector, ServiceAdapter, DownloadResult from modules.datamodels.datamodelDataSource import ExternalEntry logger = logging.getLogger(__name__) @@ -256,14 +256,24 @@ class OutlookAdapter(_GraphApiMixin, ServiceAdapter): for m in result.get("value", []) ] - async def download(self, path: str) -> bytes: - """Download a mail message as JSON bytes.""" - import json + async def download(self, path: str) -> DownloadResult: + """Download a mail message as RFC 822 EML via Graph API $value endpoint.""" + import re messageId = path.strip("/").split("/")[-1] - result = await self._graphGet(f"me/messages/{messageId}") - if "error" in result: - return b"" - return json.dumps(result, ensure_ascii=False).encode("utf-8") + + meta = await self._graphGet(f"me/messages/{messageId}?$select=subject") + subject = meta.get("subject", messageId) if "error" not in meta else messageId + safeName = re.sub(r'[<>:"/\\|?*\x00-\x1f]', "_", subject)[:80].strip(". ") or "email" + + emlBytes = await self._graphDownload(f"me/messages/{messageId}/$value") + if not emlBytes: + return DownloadResult() + + return DownloadResult( + data=emlBytes, + fileName=f"{safeName}.eml", + mimeType="message/rfc822", + ) async def upload(self, path: str, data: bytes, fileName: str) -> dict: """Not applicable for Outlook in the file sense.""" diff --git a/modules/datamodels/datamodelBilling.py b/modules/datamodels/datamodelBilling.py index f33c84e3..42d3da21 100644 --- a/modules/datamodels/datamodelBilling.py +++ b/modules/datamodels/datamodelBilling.py @@ -119,7 +119,7 @@ class BillingTransaction(BaseModel): # Context for workflow transactions workflowId: Optional[str] = Field(None, description="Workflow ID (for WORKFLOW transactions)") featureInstanceId: Optional[str] = Field(None, description="Feature instance ID") - featureCode: Optional[str] = Field(None, description="Feature code (e.g., chatplayground, automation)") + featureCode: Optional[str] = Field(None, description="Feature code (e.g., automation)") aicoreProvider: Optional[str] = Field(None, description="AICore provider (anthropic, openai, etc.)") aicoreModel: Optional[str] = Field(None, description="AICore model name (e.g., claude-4-sonnet, gpt-4o)") createdByUserId: Optional[str] = Field(None, description="User who created/caused this transaction") @@ -224,7 +224,7 @@ class UsageStatistics(BaseModel): # Breakdown by feature costByFeature: Dict[str, float] = Field( default_factory=dict, - description="Cost breakdown by feature (e.g., {'chatplayground': 15.00, 'automation': 5.80})" + description="Cost breakdown by feature (e.g., {'automation': 5.80, 'workspace': 3.20})" ) diff --git a/modules/datamodels/datamodelChat.py b/modules/datamodels/datamodelChat.py index 33a8ca7b..7002187a 100644 --- a/modules/datamodels/datamodelChat.py +++ b/modules/datamodels/datamodelChat.py @@ -247,8 +247,6 @@ class WorkflowModeEnum(str, Enum): WORKFLOW_DYNAMIC = "Dynamic" WORKFLOW_AUTOMATION = "Automation" WORKFLOW_CHATBOT = "Chatbot" - WORKFLOW_CODEEDITOR = "CodeEditor" - WORKFLOW_REACT = "React" # Legacy mode - kept for backward compatibility registerModelLabels( @@ -258,8 +256,6 @@ registerModelLabels( "WORKFLOW_DYNAMIC": {"en": "Dynamic", "fr": "Dynamique"}, "WORKFLOW_AUTOMATION": {"en": "Automation", "fr": "Automatisation"}, "WORKFLOW_CHATBOT": {"en": "Chatbot", "fr": "Chatbot"}, - "WORKFLOW_CODEEDITOR": {"en": "Code Editor", "fr": "Éditeur de code"}, - "WORKFLOW_REACT": {"en": "React (Legacy)", "fr": "React (Hérité)"}, }, ) @@ -298,10 +294,6 @@ class ChatWorkflow(BaseModel): "value": WorkflowModeEnum.WORKFLOW_CHATBOT.value, "label": {"en": "Chatbot", "fr": "Chatbot"}, }, - { - "value": WorkflowModeEnum.WORKFLOW_REACT.value, - "label": {"en": "React (Legacy)", "fr": "React (Hérité)"}, - }, ]}) maxSteps: int = Field(default=10, description="Maximum number of iterations in dynamic mode", json_schema_extra={"frontend_type": "integer", "frontend_readonly": False, "frontend_required": False}) expectedFormats: Optional[List[str]] = Field(None, description="List of expected file format extensions from user request (e.g., ['xlsx', 'pdf']). Extracted during intent analysis.", json_schema_extra={"frontend_type": "text", "frontend_readonly": True, "frontend_required": False}) diff --git a/modules/features/automation/routeFeatureAutomation.py b/modules/features/automation/routeFeatureAutomation.py index eacb90f4..6cdc4d44 100644 --- a/modules/features/automation/routeFeatureAutomation.py +++ b/modules/features/automation/routeFeatureAutomation.py @@ -17,7 +17,7 @@ from modules.features.automation.interfaceFeatureAutomation import getInterface from modules.features.automation.mainAutomation import getAutomationServices from modules.auth import limiter, getRequestContext, RequestContext from modules.features.automation.datamodelFeatureAutomation import AutomationDefinition, AutomationTemplate -from modules.datamodels.datamodelChat import ChatWorkflow, ChatMessage, ChatLog +from modules.datamodels.datamodelChat import ChatWorkflow, ChatMessage, ChatLog, UserInputRequest, WorkflowModeEnum from modules.datamodels.datamodelPagination import PaginationParams, PaginatedResponse, PaginationMetadata, normalize_pagination_dict from modules.shared.attributeUtils import getModelAttributeDefinitions from modules.interfaces import interfaceDbChat @@ -235,7 +235,7 @@ def get_available_actions( # ----------------------------------------------------------------------------- -# Workflow routes under /{instanceId}/workflows/ (instance-scoped, same as chatplayground) +# Workflow routes under /{instanceId}/workflows/ (instance-scoped) # ----------------------------------------------------------------------------- def _validateAutomationInstanceAccess(instanceId: str, context: RequestContext) -> Optional[str]: @@ -854,6 +854,46 @@ def delete_automation( detail=f"Error deleting automation: {str(e)}" ) +@router.post("/{instanceId}/start", response_model=ChatWorkflow) +@limiter.limit("120/minute") +async def start_automation_workflow( + request: Request, + instanceId: str = Path(..., description="Feature instance ID"), + workflowId: Optional[str] = Query(None, description="Optional ID of the workflow to continue"), + workflowMode: WorkflowModeEnum = Query(..., description="Workflow mode: 'Dynamic' or 'Automation' (mandatory)"), + userInput: UserInputRequest = Body(...), + context: RequestContext = Depends(getRequestContext) +) -> ChatWorkflow: + """Start a new workflow or continue an existing one.""" + try: + from modules.workflows.automation import chatStart + mandateId = _validateAutomationInstanceAccess(instanceId, context) + services = getAutomationServices( + context.user, + mandateId=mandateId, + featureInstanceId=instanceId, + ) + services.featureCode = "automation" + if hasattr(userInput, 'allowedProviders') and userInput.allowedProviders: + services.allowedProviders = userInput.allowedProviders + workflow = await chatStart( + context.user, + userInput, + workflowMode, + workflowId, + mandateId=mandateId, + featureInstanceId=instanceId, + featureCode="automation", + services=services, + ) + return workflow + except HTTPException: + raise + except Exception as e: + logger.error(f"Error in start_automation_workflow: {str(e)}") + raise HTTPException(status_code=500, detail=str(e)) + + @router.post("/{automationId}/execute", response_model=ChatWorkflow) @limiter.limit("5/minute") async def execute_automation_route( diff --git a/modules/features/chatplayground/__init__.py b/modules/features/chatplayground/__init__.py deleted file mode 100644 index 4b2f2bd4..00000000 --- a/modules/features/chatplayground/__init__.py +++ /dev/null @@ -1,6 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -Chat Playground Feature Container. -Provides workflow-based chat playground functionality. -""" diff --git a/modules/features/chatplayground/interfaceFeatureChatplayground.py b/modules/features/chatplayground/interfaceFeatureChatplayground.py deleted file mode 100644 index 3cddbc85..00000000 --- a/modules/features/chatplayground/interfaceFeatureChatplayground.py +++ /dev/null @@ -1,137 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -Chat Playground Feature Interface. -Wrapper around interfaceDbChat with feature instance context. -""" - -import logging -from typing import Dict, Any, List, Optional - -from modules.datamodels.datamodelUam import User -from modules.interfaces import interfaceDbChat - -logger = logging.getLogger(__name__) - -# Feature code constant -FEATURE_CODE = "chatplayground" - -# Singleton instances cache -_instances: Dict[str, "ChatPlaygroundObjects"] = {} - - -def getInterface(currentUser: User, mandateId: str = None, featureInstanceId: str = None) -> "ChatPlaygroundObjects": - """ - Factory function to get or create a ChatPlaygroundObjects instance. - Uses singleton pattern per user context. - - Args: - currentUser: Current user object - mandateId: Mandate ID - featureInstanceId: Feature instance ID - - Returns: - ChatPlaygroundObjects instance - """ - cacheKey = f"{currentUser.id}_{mandateId}_{featureInstanceId}" - - if cacheKey not in _instances: - _instances[cacheKey] = ChatPlaygroundObjects(currentUser, mandateId, featureInstanceId) - else: - # Update context if needed - _instances[cacheKey].setUserContext(currentUser, mandateId, featureInstanceId) - - return _instances[cacheKey] - - -class ChatPlaygroundObjects: - """ - Chat Playground feature interface. - Wraps the shared interfaceDbChat with feature instance context. - """ - - FEATURE_CODE = FEATURE_CODE - - def __init__(self, currentUser: User, mandateId: str = None, featureInstanceId: str = None): - """ - Initialize the Chat Playground interface. - - Args: - currentUser: Current user object - mandateId: Mandate ID - featureInstanceId: Feature instance ID - """ - self.currentUser = currentUser - self.mandateId = mandateId - self.featureInstanceId = featureInstanceId - - # Get the underlying chat interface - self._chatInterface = interfaceDbChat.getInterface( - currentUser, - mandateId=mandateId, - featureInstanceId=featureInstanceId - ) - - def setUserContext(self, currentUser: User, mandateId: str = None, featureInstanceId: str = None): - """ - Update the user context. - - Args: - currentUser: Current user object - mandateId: Mandate ID - featureInstanceId: Feature instance ID - """ - self.currentUser = currentUser - self.mandateId = mandateId - self.featureInstanceId = featureInstanceId - - # Update underlying interface - self._chatInterface = interfaceDbChat.getInterface( - currentUser, - mandateId=mandateId, - featureInstanceId=featureInstanceId - ) - - # ========================================================================= - # Delegated methods from interfaceDbChat - # ========================================================================= - - def getWorkflow(self, workflowId: str) -> Optional[Dict[str, Any]]: - """Get a workflow by ID.""" - return self._chatInterface.getWorkflow(workflowId) - - def getWorkflows(self, pagination=None) -> Dict[str, Any]: - """Get all workflows with pagination.""" - return self._chatInterface.getWorkflows(pagination=pagination) - - def getUnifiedChatData(self, workflowId: str, afterTimestamp: float = None) -> Dict[str, Any]: - """Get unified chat data for a workflow.""" - return self._chatInterface.getUnifiedChatData(workflowId, afterTimestamp) - - def createWorkflow(self, workflow) -> Dict[str, Any]: - """Create a new workflow.""" - return self._chatInterface.createWorkflow(workflow) - - def updateWorkflow(self, workflowId: str, updates: Dict[str, Any]) -> Optional[Dict[str, Any]]: - """Update a workflow.""" - return self._chatInterface.updateWorkflow(workflowId, updates) - - def deleteWorkflow(self, workflowId: str) -> bool: - """Delete a workflow.""" - return self._chatInterface.deleteWorkflow(workflowId) - - def getMessages(self, workflowId: str) -> List[Dict[str, Any]]: - """Get messages for a workflow.""" - return self._chatInterface.getMessages(workflowId) - - def createMessage(self, message) -> Dict[str, Any]: - """Create a new message.""" - return self._chatInterface.createMessage(message) - - def getLogs(self, workflowId: str) -> List[Dict[str, Any]]: - """Get logs for a workflow.""" - return self._chatInterface.getLogs(workflowId) - - def createLog(self, log) -> Dict[str, Any]: - """Create a new log entry.""" - return self._chatInterface.createLog(log) diff --git a/modules/features/chatplayground/mainChatplayground.py b/modules/features/chatplayground/mainChatplayground.py deleted file mode 100644 index ba1cd094..00000000 --- a/modules/features/chatplayground/mainChatplayground.py +++ /dev/null @@ -1,381 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -Chat Playground Feature Container - Main Module. -Handles feature initialization and RBAC catalog registration. -""" - -import logging -from typing import Dict, List, Any, Optional - -logger = logging.getLogger(__name__) - -# Feature metadata -FEATURE_CODE = "chatplayground" -FEATURE_LABEL = {"en": "Chat Playground", "de": "Chat Playground", "fr": "Chat Playground"} -FEATURE_ICON = "mdi-message-text" - -# UI Objects for RBAC catalog -UI_OBJECTS = [ - { - "objectKey": "ui.feature.chatplayground.playground", - "label": {"en": "Playground", "de": "Playground", "fr": "Playground"}, - "meta": {"area": "playground"} - }, - { - "objectKey": "ui.feature.chatplayground.workflows", - "label": {"en": "Workflows", "de": "Workflows", "fr": "Workflows"}, - "meta": {"area": "workflows"} - }, -] - -# Resource Objects for RBAC catalog -RESOURCE_OBJECTS = [ - { - "objectKey": "resource.feature.chatplayground.start", - "label": {"en": "Start Workflow", "de": "Workflow starten", "fr": "Démarrer workflow"}, - "meta": {"endpoint": "/api/chatplayground/{instanceId}/start", "method": "POST"} - }, - { - "objectKey": "resource.feature.chatplayground.stop", - "label": {"en": "Stop Workflow", "de": "Workflow stoppen", "fr": "Arrêter workflow"}, - "meta": {"endpoint": "/api/chatplayground/{instanceId}/workflows/{workflowId}/stop", "method": "POST"} - }, - { - "objectKey": "resource.feature.chatplayground.chatData", - "label": {"en": "Get Chat Data", "de": "Chat-Daten abrufen", "fr": "Récupérer données chat"}, - "meta": {"endpoint": "/api/chatplayground/{instanceId}/workflows/{workflowId}/chatData", "method": "GET"} - }, -] - -# Service requirements - services this feature needs from the service center -# Same as automation: chatplayground runs the same WorkflowManager and workflow methods -REQUIRED_SERVICES = [ - {"serviceKey": "chat", "meta": {"usage": "Workflow CRUD, messages, logs"}}, - {"serviceKey": "ai", "meta": {"usage": "AI planning for workflow execution"}}, - {"serviceKey": "utils", "meta": {"usage": "Timestamps, utilities"}}, - {"serviceKey": "billing", "meta": {"usage": "AI call billing"}}, - {"serviceKey": "extraction", "meta": {"usage": "Workflow method actions"}}, - {"serviceKey": "sharepoint", "meta": {"usage": "SharePoint actions (listDocuments, uploadDocument, etc.)"}}, - {"serviceKey": "generation", "meta": {"usage": "Action completion messages, document creation from results"}}, -] -# Template roles for this feature -# Role names MUST follow convention: {featureCode}-{roleName} -TEMPLATE_ROLES = [ - { - "roleLabel": "chatplayground-viewer", - "description": { - "en": "Chat Playground Viewer - View chat playground (read-only)", - "de": "Chat Playground Betrachter - Chat Playground ansehen (nur lesen)", - "fr": "Visualiseur Chat Playground - Consulter le chat playground (lecture seule)" - }, - "accessRules": [ - # UI: only playground view, NO workflows - {"context": "UI", "item": "ui.feature.chatplayground.playground", "view": True}, - # RESOURCE: NO access (viewer cannot start/stop/access chat data) - # DATA access (own records, read-only) - {"context": "DATA", "item": None, "view": True, "read": "m", "create": "n", "update": "n", "delete": "n"}, - ] - }, - { - "roleLabel": "chatplayground-user", - "description": { - "en": "Chat Playground User - Use chat playground and workflows", - "de": "Chat Playground Benutzer - Chat Playground und Workflows nutzen", - "fr": "Utilisateur Chat Playground - Utiliser le chat playground et les workflows" - }, - "accessRules": [ - # UI: full access to all views - {"context": "UI", "item": "ui.feature.chatplayground.playground", "view": True}, - {"context": "UI", "item": "ui.feature.chatplayground.workflows", "view": True}, - # Resource access: can start/stop workflows and access chat data - {"context": "RESOURCE", "item": "resource.feature.chatplayground.start", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.chatplayground.stop", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.chatplayground.chatData", "view": True}, - # DATA access (own records) - {"context": "DATA", "item": None, "view": True, "read": "m", "create": "m", "update": "m", "delete": "m"}, - ] - }, - { - "roleLabel": "chatplayground-admin", - "description": { - "en": "Chat Playground Admin - Full access to chat playground", - "de": "Chat Playground Admin - Vollzugriff auf Chat Playground", - "fr": "Administrateur Chat Playground - Accès complet au chat playground" - }, - "accessRules": [ - # Full UI access - {"context": "UI", "item": None, "view": True}, - # Full resource access - {"context": "RESOURCE", "item": None, "view": True}, - # Full DATA access - {"context": "DATA", "item": None, "view": True, "read": "a", "create": "a", "update": "a", "delete": "a"}, - ] - }, -] - - -def getRequiredServiceKeys() -> List[str]: - """Return list of service keys this feature requires.""" - return [s["serviceKey"] for s in REQUIRED_SERVICES] - - -def getChatplaygroundServices( - user, - mandateId: Optional[str] = None, - featureInstanceId: Optional[str] = None, - workflow=None, -) -> "_ChatplaygroundServiceHub": - """ - Get a service hub for the chatplayground feature using the service center. - Resolves only the services declared in REQUIRED_SERVICES. - No legacy fallback - service center only. - - Returns a hub-like object with: chat, ai, utils, billing, extraction, - sharepoint, rbac, interfaceDbApp, interfaceDbComponent, interfaceDbChat. - """ - from modules.serviceCenter import getService - from modules.serviceCenter.context import ServiceCenterContext - - _workflow = workflow - if _workflow is None: - _workflow = type("_Placeholder", (), {"featureCode": FEATURE_CODE})() - ctx = ServiceCenterContext( - user=user, - mandate_id=mandateId, - feature_instance_id=featureInstanceId, - workflow=_workflow, - ) - - hub = _ChatplaygroundServiceHub() - hub.user = user - hub.mandateId = mandateId - hub.featureInstanceId = featureInstanceId - hub.workflow = workflow - hub.featureCode = FEATURE_CODE - hub.allowedProviders = None - - for spec in REQUIRED_SERVICES: - key = spec["serviceKey"] - try: - svc = getService(key, ctx) - setattr(hub, key, svc) - except Exception as e: - logger.warning(f"Could not resolve service '{key}' for chatplayground: {e}") - setattr(hub, key, None) - - # Copy interfaces from chat service for WorkflowManager compatibility - if hub.chat: - hub.interfaceDbApp = getattr(hub.chat, "interfaceDbApp", None) - hub.interfaceDbComponent = getattr(hub.chat, "interfaceDbComponent", None) - hub.interfaceDbChat = getattr(hub.chat, "interfaceDbChat", None) - - # RBAC for MethodBase action permission checks (workflow methods) - hub.rbac = getattr(hub.interfaceDbApp, "rbac", None) if hub.interfaceDbApp else None - - return hub - - -class _ChatplaygroundServiceHub: - """Lightweight hub exposing only services required by the chatplayground feature.""" - - user = None - mandateId = None - featureInstanceId = None - workflow = None - featureCode = "chatplayground" - allowedProviders = None - interfaceDbApp = None - interfaceDbComponent = None - interfaceDbChat = None - rbac = None - chat = None - ai = None - utils = None - billing = None - extraction = None - sharepoint = None - - -def getFeatureDefinition() -> Dict[str, Any]: - """Return the feature definition for registration.""" - return { - "code": FEATURE_CODE, - "label": FEATURE_LABEL, - "icon": FEATURE_ICON, - "autoCreateInstance": True, # Automatically create instance in root mandate during bootstrap - } - - -def getUiObjects() -> List[Dict[str, Any]]: - """Return UI objects for RBAC catalog registration.""" - return UI_OBJECTS - - -def getResourceObjects() -> List[Dict[str, Any]]: - """Return resource objects for RBAC catalog registration.""" - return RESOURCE_OBJECTS - - -def getTemplateRoles() -> List[Dict[str, Any]]: - """Return template roles for this feature.""" - return TEMPLATE_ROLES - - -def registerFeature(catalogService) -> bool: - """ - Register this feature's RBAC objects in the catalog. - - Args: - catalogService: The RBAC catalog service instance - - Returns: - True if registration was successful - """ - try: - # Register UI objects - for uiObj in UI_OBJECTS: - catalogService.registerUiObject( - featureCode=FEATURE_CODE, - objectKey=uiObj["objectKey"], - label=uiObj["label"], - meta=uiObj.get("meta") - ) - - # Register Resource objects - for resObj in RESOURCE_OBJECTS: - catalogService.registerResourceObject( - featureCode=FEATURE_CODE, - objectKey=resObj["objectKey"], - label=resObj["label"], - meta=resObj.get("meta") - ) - - # Sync template roles to database - _syncTemplateRolesToDb() - - logger.info(f"Feature '{FEATURE_CODE}' registered {len(UI_OBJECTS)} UI objects and {len(RESOURCE_OBJECTS)} resource objects") - return True - - except Exception as e: - logger.error(f"Failed to register feature '{FEATURE_CODE}': {e}") - return False - - -def _syncTemplateRolesToDb() -> int: - """ - Sync template roles and their AccessRules to the database. - Creates global template roles (mandateId=None) if they don't exist. - - Returns: - Number of roles created/updated - """ - try: - from modules.interfaces.interfaceDbApp import getRootInterface - from modules.datamodels.datamodelRbac import Role, AccessRule, AccessRuleContext - - rootInterface = getRootInterface() - - # Get existing template roles for this feature (Pydantic models) - existingRoles = rootInterface.getRolesByFeatureCode(FEATURE_CODE) - # Filter to template roles (mandateId is None) - templateRoles = [r for r in existingRoles if r.mandateId is None] - existingRoleLabels = {r.roleLabel: str(r.id) for r in templateRoles} - - createdCount = 0 - for roleTemplate in TEMPLATE_ROLES: - roleLabel = roleTemplate["roleLabel"] - - if roleLabel in existingRoleLabels: - roleId = existingRoleLabels[roleLabel] - # Ensure AccessRules exist for this role - _ensureAccessRulesForRole(rootInterface, roleId, roleTemplate.get("accessRules", [])) - else: - # Create new template role - newRole = Role( - roleLabel=roleLabel, - description=roleTemplate.get("description", {}), - featureCode=FEATURE_CODE, - mandateId=None, # Global template - featureInstanceId=None, - isSystemRole=False - ) - createdRole = rootInterface.db.recordCreate(Role, newRole.model_dump()) - roleId = createdRole.get("id") - - # Create AccessRules for this role - _ensureAccessRulesForRole(rootInterface, roleId, roleTemplate.get("accessRules", [])) - - logger.info(f"Created template role '{roleLabel}' with ID {roleId}") - createdCount += 1 - - if createdCount > 0: - logger.info(f"Feature '{FEATURE_CODE}': Created {createdCount} template roles") - - return createdCount - - except Exception as e: - logger.error(f"Error syncing template roles for feature '{FEATURE_CODE}': {e}") - return 0 - - -def _ensureAccessRulesForRole(rootInterface, roleId: str, ruleTemplates: List[Dict[str, Any]]) -> int: - """ - Ensure AccessRules exist for a role based on templates. - - Args: - rootInterface: Root interface instance - roleId: Role ID - ruleTemplates: List of rule templates - - Returns: - Number of rules created - """ - from modules.datamodels.datamodelRbac import AccessRule, AccessRuleContext - - # Get existing rules for this role (Pydantic models) - existingRules = rootInterface.getAccessRulesByRole(roleId) - - # Create a set of existing rule signatures to avoid duplicates - # IMPORTANT: Use .value for enum comparison, not str() which gives "AccessRuleContext.DATA" in Python 3.11+ - existingSignatures = set() - for rule in existingRules: - sig = (rule.context.value if rule.context else None, rule.item) - existingSignatures.add(sig) - - createdCount = 0 - for template in ruleTemplates: - context = template.get("context", "UI") - item = template.get("item") - sig = (context, item) - - if sig in existingSignatures: - continue - - # Map context string to enum - if context == "UI": - contextEnum = AccessRuleContext.UI - elif context == "DATA": - contextEnum = AccessRuleContext.DATA - elif context == "RESOURCE": - contextEnum = AccessRuleContext.RESOURCE - else: - contextEnum = context - - newRule = AccessRule( - roleId=roleId, - context=contextEnum, - item=item, - view=template.get("view", False), - read=template.get("read"), - create=template.get("create"), - update=template.get("update"), - delete=template.get("delete"), - ) - rootInterface.db.recordCreate(AccessRule, newRule.model_dump()) - createdCount += 1 - - if createdCount > 0: - logger.debug(f"Created {createdCount} AccessRules for role {roleId}") - - return createdCount diff --git a/modules/features/chatplayground/routeFeatureChatplayground.py b/modules/features/chatplayground/routeFeatureChatplayground.py deleted file mode 100644 index 1566c07b..00000000 --- a/modules/features/chatplayground/routeFeatureChatplayground.py +++ /dev/null @@ -1,722 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -Chat Playground Feature Routes. -Implements the endpoints for chat playground workflow management as a feature. -""" - -import json -import logging -from typing import Optional, Dict, Any -from fastapi import APIRouter, HTTPException, Depends, Body, Path, Query, Request, status - -# Import auth modules -from modules.auth import limiter, getRequestContext, RequestContext - -# Import interfaces -from modules.interfaces import interfaceDbChat -from modules.interfaces.interfaceDbBilling import getInterface as _getBillingInterface - -# Import models -from modules.datamodels.datamodelChat import ( - ChatWorkflow, - ChatMessage, - ChatLog, - UserInputRequest, - WorkflowModeEnum, -) -from modules.datamodels.datamodelPagination import ( - PaginationParams, - PaginatedResponse, - PaginationMetadata, - normalize_pagination_dict, -) - -# Import workflow control functions -from modules.workflows.automation import chatStart, chatStop -from modules.features.chatplayground.mainChatplayground import getChatplaygroundServices -from modules.shared.attributeUtils import getModelAttributeDefinitions - -# Configure logger -logger = logging.getLogger(__name__) - -# Model attributes for ChatWorkflow (workflow attributes endpoint) -workflowAttributes = getModelAttributeDefinitions(ChatWorkflow) - -# Create router for chat playground feature endpoints -router = APIRouter( - prefix="/api/chatplayground", - tags=["Chat Playground Feature"], - responses={404: {"description": "Not found"}} -) - - -def _getServiceChat(context: RequestContext, featureInstanceId: str = None): - """Get chat interface with feature instance context.""" - return interfaceDbChat.getInterface( - context.user, - mandateId=str(context.mandateId) if context.mandateId else None, - featureInstanceId=featureInstanceId - ) - - -def _validateInstanceAccess(instanceId: str, context: RequestContext) -> str: - """ - Validate that user has access to the feature instance. - - Args: - instanceId: Feature instance ID - context: Request context - - Returns: - mandateId for the instance - - Raises: - HTTPException if access is denied - """ - from modules.interfaces.interfaceDbApp import getRootInterface - - rootInterface = getRootInterface() - - # Get feature instance (Pydantic model) - instance = rootInterface.getFeatureInstance(instanceId) - if not instance: - raise HTTPException(status_code=404, detail=f"Feature instance {instanceId} not found") - - # Check user has access to this instance using interface method - featureAccess = rootInterface.getFeatureAccess(str(context.user.id), instanceId) - - if not featureAccess or not featureAccess.enabled: - raise HTTPException(status_code=403, detail="Access denied to this feature instance") - - return str(instance.mandateId) if instance.mandateId else None - - -# Workflow start endpoint -@router.post("/{instanceId}/start", response_model=ChatWorkflow) -@limiter.limit("120/minute") -async def start_workflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: Optional[str] = Query(None, description="Optional ID of the workflow to continue"), - workflowMode: WorkflowModeEnum = Query(..., description="Workflow mode: 'Dynamic' or 'Automation' (mandatory)"), - userInput: UserInputRequest = Body(...), - context: RequestContext = Depends(getRequestContext) -) -> ChatWorkflow: - """ - Starts a new workflow or continues an existing one. - - Args: - instanceId: Feature instance ID - workflowMode: "Dynamic" for iterative dynamic-style processing, "Automation" for automated workflow execution - """ - try: - # Validate access and get mandate ID - mandateId = _validateInstanceAccess(instanceId, context) - - # Get chatplayground services from service center (not automation) - services = getChatplaygroundServices( - context.user, - mandateId=mandateId, - featureInstanceId=instanceId, - ) - services.featureCode = 'chatplayground' - if hasattr(userInput, 'allowedProviders') and userInput.allowedProviders: - services.allowedProviders = userInput.allowedProviders - - # Start or continue workflow - workflow = await chatStart( - context.user, - userInput, - workflowMode, - workflowId, - mandateId=mandateId, - featureInstanceId=instanceId, - featureCode='chatplayground', - services=services, - ) - - return workflow - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error in start_workflow: {str(e)}") - raise HTTPException( - status_code=500, - detail=str(e) - ) - - -# Stop workflow endpoint (under /workflows/{workflowId}/ for consistency) -@router.post("/{instanceId}/workflows/{workflowId}/stop", response_model=ChatWorkflow) -@limiter.limit("120/minute") -async def stop_workflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow to stop"), - context: RequestContext = Depends(getRequestContext) -) -> ChatWorkflow: - """Stops a running workflow.""" - try: - # Validate access and get mandate ID - mandateId = _validateInstanceAccess(instanceId, context) - - # Get chatplayground services from service center (not automation) - services = getChatplaygroundServices( - context.user, - mandateId=mandateId, - featureInstanceId=instanceId, - ) - services.featureCode = 'chatplayground' - - # Stop workflow (pass featureInstanceId for proper RBAC filtering) - workflow = await chatStop( - context.user, - workflowId, - mandateId=mandateId, - featureInstanceId=instanceId, - featureCode='chatplayground', - services=services, - ) - - return workflow - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error in stop_workflow: {str(e)}") - raise HTTPException( - status_code=500, - detail=str(e) - ) - - -# Unified Chat Data Endpoint for Polling (under /workflows/{workflowId}/ for consistency) -@router.get("/{instanceId}/workflows/{workflowId}/chatData") -@limiter.limit("120/minute") -def get_workflow_chat_data( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - afterTimestamp: Optional[float] = Query(None, description="Unix timestamp to get data after"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """ - Get unified chat data (messages, logs, stats) for a workflow with timestamp-based selective data transfer. - Returns all data types in chronological order based on _createdAt timestamp. - """ - try: - # Validate access - _validateInstanceAccess(instanceId, context) - - # Get service with feature instance context - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - - # Verify workflow exists - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException( - status_code=404, - detail=f"Workflow with ID {workflowId} not found" - ) - - # Get workflow cost from billing transactions (single source of truth) - billingInterface = _getBillingInterface(context.user, context.mandateId) - workflowCost = billingInterface.getWorkflowCost(workflowId) - - chatData = chatInterface.getUnifiedChatData(workflowId, afterTimestamp, workflowCost=workflowCost) - return chatData - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting unified chat data: {str(e)}", exc_info=True) - raise HTTPException( - status_code=500, - detail=f"Error getting unified chat data: {str(e)}" - ) - - -# Get workflow attributes (ChatWorkflow model) -@router.get("/{instanceId}/workflows/attributes", response_model=Dict[str, Any]) -@limiter.limit("120/minute") -def get_workflow_attributes( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get attribute definitions for ChatWorkflow model.""" - _validateInstanceAccess(instanceId, context) - return {"attributes": workflowAttributes} - - -# Get workflows for this instance -@router.get("/{instanceId}/workflows", response_model=PaginatedResponse[ChatWorkflow]) -@limiter.limit("120/minute") -def get_workflows( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - pagination: Optional[str] = Query(None, description="JSON-encoded PaginationParams object"), - page: int = Query(1, ge=1, description="Page number (legacy)"), - pageSize: int = Query(20, ge=1, le=100, description="Items per page (legacy)"), - context: RequestContext = Depends(getRequestContext) -) -> PaginatedResponse[ChatWorkflow]: - """ - Get all workflows for this feature instance with optional pagination. - """ - try: - # Validate access - _validateInstanceAccess(instanceId, context) - - # Get service with feature instance context - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - - # Parse pagination parameter - paginationParams = None - if pagination: - try: - paginationDict = json.loads(pagination) - if paginationDict: - paginationDict = normalize_pagination_dict(paginationDict) - paginationParams = PaginationParams(**paginationDict) - except (json.JSONDecodeError, ValueError) as e: - raise HTTPException( - status_code=400, - detail=f"Invalid pagination parameter: {str(e)}" - ) - else: - paginationParams = PaginationParams(page=page, pageSize=pageSize) - - result = chatInterface.getWorkflows(pagination=paginationParams) - - if paginationParams: - return PaginatedResponse( - items=result.items, - pagination=PaginationMetadata( - currentPage=paginationParams.page, - pageSize=paginationParams.pageSize, - totalItems=result.totalItems, - totalPages=result.totalPages, - sort=paginationParams.sort, - filters=paginationParams.filters - ) - ) - else: - return PaginatedResponse(items=result, pagination=None) - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflows: {str(e)}", exc_info=True) - raise HTTPException( - status_code=500, - detail=f"Error getting workflows: {str(e)}" - ) - - -# Action Discovery Endpoints (must be before /{workflowId} to avoid path conflict) -@router.get("/{instanceId}/workflows/actions", response_model=Dict[str, Any]) -@limiter.limit("120/minute") -def get_all_workflow_actions( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get all available workflow actions for the current user (filtered by RBAC).""" - try: - mandateId = _validateInstanceAccess(instanceId, context) - services = getChatplaygroundServices( - context.user, - mandateId=mandateId, - featureInstanceId=instanceId, - ) - from modules.workflows.processing.shared.methodDiscovery import discoverMethods, methods - discoverMethods(services) - allActions = [] - for methodName, methodInfo in methods.items(): - if methodName.startswith('Method'): - continue - methodInstance = methodInfo['instance'] - methodActions = methodInstance.actions - for actionName, actionInfo in methodActions.items(): - actionResponse = { - "module": methodInstance.name, - "actionId": f"{methodInstance.name}.{actionName}", - "name": actionName, - "description": actionInfo.get('description', ''), - "parameters": actionInfo.get('parameters', {}) - } - allActions.append(actionResponse) - return {"actions": allActions} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting all actions: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Failed to get actions: {str(e)}") - - -@router.get("/{instanceId}/workflows/actions/{method}", response_model=Dict[str, Any]) -@limiter.limit("120/minute") -def get_method_workflow_actions( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - method: str = Path(..., description="Method name (e.g., 'outlook', 'sharepoint')"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get all available actions for a specific method.""" - try: - mandateId = _validateInstanceAccess(instanceId, context) - services = getChatplaygroundServices( - context.user, - mandateId=mandateId, - featureInstanceId=instanceId, - ) - from modules.workflows.processing.shared.methodDiscovery import discoverMethods, methods - discoverMethods(services) - methodInstance = None - for methodName, methodInfo in methods.items(): - if methodInfo['instance'].name == method: - methodInstance = methodInfo['instance'] - break - if not methodInstance: - raise HTTPException(status_code=404, detail=f"Method '{method}' not found") - actions = [] - for actionName, actionInfo in methodInstance.actions.items(): - actionResponse = { - "actionId": f"{methodInstance.name}.{actionName}", - "name": actionName, - "description": actionInfo.get('description', ''), - "parameters": actionInfo.get('parameters', {}) - } - actions.append(actionResponse) - return {"module": methodInstance.name, "description": methodInstance.description, "actions": actions} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting actions for method {method}: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Failed to get actions for method {method}: {str(e)}") - - -@router.get("/{instanceId}/workflows/actions/{method}/{action}", response_model=Dict[str, Any]) -@limiter.limit("120/minute") -def get_action_schema( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - method: str = Path(..., description="Method name (e.g., 'outlook', 'sharepoint')"), - action: str = Path(..., description="Action name (e.g., 'readEmails', 'uploadDocument')"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get action schema with parameter definitions for a specific action.""" - try: - mandateId = _validateInstanceAccess(instanceId, context) - services = getChatplaygroundServices( - context.user, - mandateId=mandateId, - featureInstanceId=instanceId, - ) - from modules.workflows.processing.shared.methodDiscovery import discoverMethods, methods - discoverMethods(services) - methodInstance = None - for methodName, methodInfo in methods.items(): - if methodInfo['instance'].name == method: - methodInstance = methodInfo['instance'] - break - if not methodInstance: - raise HTTPException(status_code=404, detail=f"Method '{method}' not found") - methodActions = methodInstance.actions - if action not in methodActions: - raise HTTPException(status_code=404, detail=f"Action '{action}' not found in method '{method}'") - actionInfo = methodActions[action] - return { - "method": methodInstance.name, - "action": action, - "actionId": f"{methodInstance.name}.{action}", - "description": actionInfo.get('description', ''), - "parameters": actionInfo.get('parameters', {}) - } - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting action schema for {method}.{action}: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Failed to get action schema: {str(e)}") - - -# Get single workflow by ID -@router.get("/{instanceId}/workflows/{workflowId}", response_model=ChatWorkflow) -@limiter.limit("120/minute") -def get_workflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - context: RequestContext = Depends(getRequestContext) -) -> ChatWorkflow: - """Get workflow by ID.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail="Workflow not found") - return workflow - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflow: {str(e)}") - raise HTTPException(status_code=500, detail=f"Failed to get workflow: {str(e)}") - - -# Update workflow -@router.put("/{instanceId}/workflows/{workflowId}", response_model=ChatWorkflow) -@limiter.limit("120/minute") -def update_workflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow to update"), - workflowData: Dict[str, Any] = Body(...), - context: RequestContext = Depends(getRequestContext) -) -> ChatWorkflow: - """Update workflow by ID.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail="Workflow not found") - if not chatInterface.checkRbacPermission(ChatWorkflow, "update", workflowId): - raise HTTPException(status_code=403, detail="You don't have permission to update this workflow") - updatedWorkflow = chatInterface.updateWorkflow(workflowId, workflowData) - if not updatedWorkflow: - raise HTTPException(status_code=500, detail="Failed to update workflow") - return updatedWorkflow - except HTTPException: - raise - except Exception as e: - logger.error(f"Error updating workflow: {str(e)}") - raise HTTPException(status_code=500, detail=f"Failed to update workflow: {str(e)}") - - -# Delete workflow -@router.delete("/{instanceId}/workflows/{workflowId}") -@limiter.limit("120/minute") -def delete_workflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow to delete"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Deletes a workflow and its associated data.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - if not chatInterface.checkRbacPermission(ChatWorkflow, "delete", workflowId): - raise HTTPException(status_code=403, detail="You don't have permission to delete this workflow") - success = chatInterface.deleteWorkflow(workflowId) - if not success: - raise HTTPException(status_code=500, detail="Failed to delete workflow") - return {"id": workflowId, "message": "Workflow and associated data deleted successfully"} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error deleting workflow: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error deleting workflow: {str(e)}") - - -# Get workflow status -@router.get("/{instanceId}/workflows/{workflowId}/status", response_model=ChatWorkflow) -@limiter.limit("120/minute") -def get_workflow_status( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - context: RequestContext = Depends(getRequestContext) -) -> ChatWorkflow: - """Get the current status of a workflow.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - return workflow - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflow status: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error getting workflow status: {str(e)}") - - -# Get workflow logs -@router.get("/{instanceId}/workflows/{workflowId}/logs", response_model=PaginatedResponse[ChatLog]) -@limiter.limit("120/minute") -def get_workflow_logs( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - logId: Optional[str] = Query(None, description="Optional log ID for selective data transfer"), - pagination: Optional[str] = Query(None, description="JSON-encoded PaginationParams object"), - context: RequestContext = Depends(getRequestContext) -) -> PaginatedResponse[ChatLog]: - """Get logs for a workflow with optional pagination.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - - paginationParams = None - if pagination: - try: - paginationDict = json.loads(pagination) - if paginationDict: - paginationDict = normalize_pagination_dict(paginationDict) - paginationParams = PaginationParams(**paginationDict) - except (json.JSONDecodeError, ValueError) as e: - raise HTTPException(status_code=400, detail=f"Invalid pagination parameter: {str(e)}") - - result = chatInterface.getLogs(workflowId, pagination=paginationParams) - - if logId: - allLogs = result.items if paginationParams else result - logIndex = next((i for i, log in enumerate(allLogs) if log.id == logId), -1) - if logIndex >= 0: - filteredLogs = allLogs[logIndex + 1:] - return PaginatedResponse(items=filteredLogs, pagination=None) - - if paginationParams: - return PaginatedResponse( - items=result.items, - pagination=PaginationMetadata( - currentPage=paginationParams.page, - pageSize=paginationParams.pageSize, - totalItems=result.totalItems, - totalPages=result.totalPages, - sort=paginationParams.sort, - filters=paginationParams.filters - ) - ) - return PaginatedResponse(items=result, pagination=None) - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflow logs: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error getting workflow logs: {str(e)}") - - -# Get workflow messages -@router.get("/{instanceId}/workflows/{workflowId}/messages", response_model=PaginatedResponse[ChatMessage]) -@limiter.limit("120/minute") -def get_workflow_messages( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - messageId: Optional[str] = Query(None, description="Optional message ID for selective data transfer"), - pagination: Optional[str] = Query(None, description="JSON-encoded PaginationParams object"), - context: RequestContext = Depends(getRequestContext) -) -> PaginatedResponse[ChatMessage]: - """Get messages for a workflow with optional pagination.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - - paginationParams = None - if pagination: - try: - paginationDict = json.loads(pagination) - if paginationDict: - paginationDict = normalize_pagination_dict(paginationDict) - paginationParams = PaginationParams(**paginationDict) - except (json.JSONDecodeError, ValueError) as e: - raise HTTPException(status_code=400, detail=f"Invalid pagination parameter: {str(e)}") - - result = chatInterface.getMessages(workflowId, pagination=paginationParams) - - if messageId: - allMessages = result.items if paginationParams else result - messageIndex = next((i for i, msg in enumerate(allMessages) if msg.id == messageId), -1) - if messageIndex >= 0: - filteredMessages = allMessages[messageIndex + 1:] - return PaginatedResponse(items=filteredMessages, pagination=None) - - if paginationParams: - return PaginatedResponse( - items=result.items, - pagination=PaginationMetadata( - currentPage=paginationParams.page, - pageSize=paginationParams.pageSize, - totalItems=result.totalItems, - totalPages=result.totalPages, - sort=paginationParams.sort, - filters=paginationParams.filters - ) - ) - return PaginatedResponse(items=result, pagination=None) - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflow messages: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error getting workflow messages: {str(e)}") - - -# Delete message from workflow -@router.delete("/{instanceId}/workflows/{workflowId}/messages/{messageId}") -@limiter.limit("120/minute") -def delete_workflow_message( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - messageId: str = Path(..., description="ID of the message to delete"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Delete a message from a workflow.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - success = chatInterface.deleteMessage(workflowId, messageId) - if not success: - raise HTTPException(status_code=404, detail=f"Message with ID {messageId} not found in workflow {workflowId}") - return {"workflowId": workflowId, "messageId": messageId, "message": "Message deleted successfully"} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error deleting message: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error deleting message: {str(e)}") - - -# Delete file from message -@router.delete("/{instanceId}/workflows/{workflowId}/messages/{messageId}/files/{fileId}") -@limiter.limit("120/minute") -def delete_file_from_message( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="ID of the workflow"), - messageId: str = Path(..., description="ID of the message"), - fileId: str = Path(..., description="ID of the file to delete"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Delete a file reference from a message in a workflow.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow with ID {workflowId} not found") - success = chatInterface.deleteFileFromMessage(workflowId, messageId, fileId) - if not success: - raise HTTPException(status_code=404, detail=f"File with ID {fileId} not found in message {messageId}") - return {"workflowId": workflowId, "messageId": messageId, "fileId": fileId, "message": "File reference deleted successfully"} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error deleting file reference: {str(e)}", exc_info=True) - raise HTTPException(status_code=500, detail=f"Error deleting file reference: {str(e)}") diff --git a/modules/features/codeeditor/__init__.py b/modules/features/codeeditor/__init__.py deleted file mode 100644 index d6cca46d..00000000 --- a/modules/features/codeeditor/__init__.py +++ /dev/null @@ -1 +0,0 @@ -"""CodeEditor Feature - Cursor-style AI file editing via chat interface.""" diff --git a/modules/features/codeeditor/codeEditorProcessor.py b/modules/features/codeeditor/codeEditorProcessor.py deleted file mode 100644 index 7e472016..00000000 --- a/modules/features/codeeditor/codeEditorProcessor.py +++ /dev/null @@ -1,280 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""CodeEditor processor -- single-shot (Phase 1) and agent loop (Phase 2). -Orchestrates file loading, prompt building, AI calls, response parsing, and SSE emission.""" - -import logging -from typing import List, Dict, Any - -from modules.features.codeeditor import fileContextManager, promptAssembly, responseParser -from modules.features.codeeditor.datamodelCodeeditor import ( - FileEditProposal, SegmentTypeEnum, AgentState -) -from modules.features.codeeditor import toolRegistry -from modules.shared.timeUtils import getUtcTimestamp - -logger = logging.getLogger(__name__) - - -async def processMessage( - workflowId: str, - userPrompt: str, - selectedFileIds: List[str], - dbManagement, - interfaceAi, - chatInterface, - eventManager, - agentMode: bool = False -): - """Process a user message. Dispatches to single-shot or agent loop based on mode.""" - if agentMode: - await _processAgentMessage( - workflowId, userPrompt, dbManagement, interfaceAi, chatInterface, eventManager - ) - else: - await _processSingleShot( - workflowId, userPrompt, selectedFileIds, dbManagement, interfaceAi, chatInterface, eventManager - ) - - -async def _processSingleShot( - workflowId, userPrompt, selectedFileIds, dbManagement, interfaceAi, chatInterface, eventManager -): - """Phase 1: Single AI call with pre-loaded file context.""" - try: - await _emitStatus(eventManager, workflowId, "Loading files...") - fileContexts = await fileContextManager.loadFileContexts(dbManagement, selectedFileIds) - - await _emitStatus(eventManager, workflowId, "Building prompt...") - chatHistory = _loadChatHistory(chatInterface, workflowId) - aiRequest = promptAssembly.buildRequest(userPrompt, fileContexts, chatHistory) - - await _emitStatus(eventManager, workflowId, "AI is processing...") - aiResponse = await interfaceAi.callWithTextContext(aiRequest) - - if aiResponse.errorCount > 0: - await _emitError(eventManager, workflowId, aiResponse.content) - return - - segments = responseParser.parseResponse(aiResponse.content) - await _emitSegments(eventManager, workflowId, segments, fileContexts) - _logAiStats(aiResponse, workflowId) - - await eventManager.emit_event(workflowId, "complete", { - "workflowId": workflowId, - "modelName": aiResponse.modelName, - "priceCHF": aiResponse.priceCHF, - "processingTime": aiResponse.processingTime - }) - - except Exception as e: - logger.error(f"CodeEditor single-shot failed for {workflowId}: {e}", exc_info=True) - await eventManager.emit_event(workflowId, "error", { - "workflowId": workflowId, "error": str(e) - }) - - -async def _processAgentMessage( - workflowId, userPrompt, dbManagement, interfaceAi, chatInterface, eventManager -): - """Phase 2: Agent loop -- multiple AI calls with tool execution until done.""" - state = AgentState(workflowId=workflowId) - - try: - await _emitStatus(eventManager, workflowId, "Agent: Scanning available files...") - fileListContext = fileContextManager.buildFileListContext(dbManagement) - - state.conversationHistory.append({"role": "user", "content": userPrompt}) - - aiRequest = promptAssembly.buildAgentRequest( - userPrompt=userPrompt, - fileListContext=fileListContext, - conversationHistory=[] - ) - - while state.status == "running" and state.currentRound < state.maxRounds: - state.currentRound += 1 - state.totalAiCalls += 1 - - await _emitStatus(eventManager, workflowId, - f"Agent round {state.currentRound}: AI is thinking...") - - await eventManager.emit_event(workflowId, "chatdata", { - "type": "agent_progress", - "item": { - "round": state.currentRound, - "totalAiCalls": state.totalAiCalls, - "totalToolCalls": state.totalToolCalls, - "costCHF": round(state.totalCostCHF, 4), - } - }) - - aiResponse = await interfaceAi.callWithTextContext(aiRequest) - state.totalCostCHF += aiResponse.priceCHF - state.totalProcessingTime += aiResponse.processingTime - - if aiResponse.errorCount > 0: - logger.error(f"Agent AI call failed in round {state.currentRound}: {aiResponse.content}") - await _emitError(eventManager, workflowId, aiResponse.content) - state.status = "error" - break - - _logAiStats(aiResponse, workflowId) - - state.conversationHistory.append({"role": "assistant", "content": aiResponse.content}) - - segments = responseParser.parseResponse(aiResponse.content) - - textAndEditSegments = [s for s in segments if s.type != SegmentTypeEnum.TOOL_CALL] - if textAndEditSegments: - await _emitSegments(eventManager, workflowId, textAndEditSegments, []) - - toolCallSegments = [s for s in segments if s.type == SegmentTypeEnum.TOOL_CALL] - - if not toolCallSegments: - state.status = "completed" - break - - toolResultTexts = [] - for tc in toolCallSegments: - state.totalToolCalls += 1 - await _emitStatus(eventManager, workflowId, - f"Agent: Running {tc.toolName}...") - - result = await toolRegistry.dispatch(tc.toolName, tc.toolArgs or {}, dbManagement) - toolResultTexts.append(f"[{tc.toolName}] (success={result.success}):\n{result.result}") - - logger.info(f"Agent tool {tc.toolName}: success={result.success}, time={result.executionTime:.2f}s") - - combinedResults = "\n\n".join(toolResultTexts) - state.conversationHistory.append({ - "role": "tool_result", - "content": combinedResults, - "toolName": "batch" - }) - - aiRequest = promptAssembly.buildAgentRequest( - userPrompt=None, - fileListContext=fileListContext, - conversationHistory=state.conversationHistory - ) - - if state.currentRound >= state.maxRounds and state.status == "running": - state.status = "max_rounds" - await eventManager.emit_event(workflowId, "chatdata", { - "type": "message", - "item": { - "role": "system", - "content": f"Agent stopped: maximum rounds ({state.maxRounds}) reached.", - "createdAt": getUtcTimestamp() - } - }) - - await eventManager.emit_event(workflowId, "chatdata", { - "type": "agent_summary", - "item": { - "rounds": state.currentRound, - "totalAiCalls": state.totalAiCalls, - "totalToolCalls": state.totalToolCalls, - "costCHF": round(state.totalCostCHF, 4), - "processingTime": round(state.totalProcessingTime, 1), - "status": state.status, - } - }) - - await eventManager.emit_event(workflowId, "complete", { - "workflowId": workflowId, - "agentRounds": state.currentRound, - "totalCostCHF": round(state.totalCostCHF, 4), - "processingTime": round(state.totalProcessingTime, 1) - }) - - except Exception as e: - logger.error(f"CodeEditor agent loop failed for {workflowId}: {e}", exc_info=True) - await eventManager.emit_event(workflowId, "error", { - "workflowId": workflowId, "error": str(e) - }) - - -# --------------------------------------------------------------------------- -# Shared helpers -# --------------------------------------------------------------------------- - -async def _emitStatus(eventManager, workflowId: str, label: str): - await eventManager.emit_event(workflowId, "chatdata", { - "type": "status", "label": label - }) - - -async def _emitError(eventManager, workflowId: str, errorMsg: str): - await eventManager.emit_event(workflowId, "chatdata", { - "type": "message", - "item": {"role": "assistant", "content": f"Error: {errorMsg}"} - }) - await eventManager.emit_event(workflowId, "error", { - "workflowId": workflowId, "error": errorMsg - }) - - -async def _emitSegments(eventManager, workflowId: str, segments, fileContexts): - """Emit parsed segments as SSE events.""" - for segment in segments: - messageData = { - "role": "assistant", - "content": segment.content, - "type": segment.type.value, - "createdAt": getUtcTimestamp() - } - await eventManager.emit_event(workflowId, "chatdata", { - "type": "message", "item": messageData - }) - - if segment.type == SegmentTypeEnum.FILE_EDIT: - proposal = FileEditProposal( - workflowId=workflowId, - fileId=_resolveFileId(segment.fileName, fileContexts), - fileName=segment.fileName, - operation="edit", - oldContent=segment.oldContent, - newContent=segment.newContent - ) - await eventManager.emit_event(workflowId, "chatdata", { - "type": "file_edit_proposal", "item": proposal.model_dump() - }) - - -def _loadChatHistory(chatInterface, workflowId: str) -> List[Dict[str, Any]]: - """Load recent chat messages for multi-turn context.""" - try: - messages = chatInterface.getMessages(workflowId) - if not messages: - return [] - history = [] - for msg in messages: - role = msg.get("role", "unknown") if isinstance(msg, dict) else getattr(msg, "role", "unknown") - content = msg.get("content", "") if isinstance(msg, dict) else getattr(msg, "content", "") - history.append({"role": role, "content": content}) - return history - except Exception as e: - logger.warning(f"Could not load chat history for {workflowId}: {e}") - return [] - - -def _resolveFileId(fileName: str, fileContexts) -> str: - """Resolve a fileName to its fileId from the loaded contexts.""" - for fc in fileContexts: - if fc.fileName == fileName: - return fc.fileId - return f"unknown-{fileName}" - - -def _logAiStats(aiResponse, workflowId: str): - """Log AI call statistics.""" - logger.info( - f"CodeEditor AI call for {workflowId}: " - f"model={aiResponse.modelName}, " - f"provider={aiResponse.provider}, " - f"cost={aiResponse.priceCHF:.4f} CHF, " - f"time={aiResponse.processingTime:.1f}s, " - f"sent={aiResponse.bytesSent}B, received={aiResponse.bytesReceived}B" - ) diff --git a/modules/features/codeeditor/datamodelCodeeditor.py b/modules/features/codeeditor/datamodelCodeeditor.py deleted file mode 100644 index 448496bf..00000000 --- a/modules/features/codeeditor/datamodelCodeeditor.py +++ /dev/null @@ -1,122 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""Data models for the CodeEditor feature.""" - -from typing import List, Dict, Any, Optional -from enum import Enum -from pydantic import BaseModel, Field -from modules.shared.timeUtils import getUtcTimestamp -import uuid - - -class SegmentTypeEnum(str, Enum): - TEXT = "text" - CODE_BLOCK = "code_block" - FILE_EDIT = "file_edit" - TOOL_CALL = "tool_call" - - -class EditStatusEnum(str, Enum): - PENDING = "pending" - ACCEPTED = "accepted" - REJECTED = "rejected" - - -class FileContext(BaseModel): - """A text file loaded as context for the AI.""" - fileId: str - fileName: str - content: Optional[str] = None - mimeType: str - sizeBytes: int = 0 - modifiedAt: Optional[float] = None - tags: List[str] = Field(default_factory=list) - - -class ResponseSegment(BaseModel): - """A parsed segment from the AI response.""" - type: SegmentTypeEnum - content: str - language: Optional[str] = None - fileId: Optional[str] = None - fileName: Optional[str] = None - oldContent: Optional[str] = None - newContent: Optional[str] = None - toolName: Optional[str] = None - toolArgs: Optional[Dict[str, Any]] = None - - -class FileEditProposal(BaseModel): - """A proposed file edit from the AI, awaiting user accept/reject.""" - id: str = Field(default_factory=lambda: str(uuid.uuid4())) - workflowId: str - fileId: str - fileName: str - operation: str = "edit" - oldContent: Optional[str] = None - newContent: str - diffSummary: Optional[str] = None - status: EditStatusEnum = EditStatusEnum.PENDING - createdAt: float = Field(default_factory=getUtcTimestamp) - - -class FileVersion(BaseModel): - """A new version of a file created after accepting an edit proposal.""" - id: str = Field(default_factory=lambda: str(uuid.uuid4())) - sourceFileId: str - editProposalId: str - newFileId: str - createdAt: float = Field(default_factory=getUtcTimestamp) - - -class AgentState(BaseModel): - """Tracks state across an agent loop execution.""" - workflowId: str - currentRound: int = 0 - maxRounds: int = 50 - totalAiCalls: int = 0 - totalToolCalls: int = 0 - totalCostCHF: float = 0.0 - totalProcessingTime: float = 0.0 - conversationHistory: List[Dict[str, Any]] = Field(default_factory=list) - status: str = "running" - - -class ToolResult(BaseModel): - """Result from executing a tool.""" - toolName: str - result: str - success: bool = True - executionTime: float = 0.0 - - -TEXT_MIME_TYPES = { - "text/plain", "text/markdown", "text/html", "text/css", "text/csv", - "text/xml", "text/yaml", "text/x-python", "text/x-java", - "text/javascript", "text/x-typescript", "text/x-sql", - "application/json", "application/xml", "application/yaml", - "application/x-yaml", "application/javascript", -} - -TEXT_EXTENSIONS = { - ".md", ".txt", ".json", ".yaml", ".yml", ".xml", ".csv", - ".py", ".js", ".ts", ".tsx", ".jsx", ".html", ".htm", ".css", ".scss", - ".sql", ".sh", ".bash", ".zsh", ".ps1", ".bat", - ".toml", ".ini", ".cfg", ".conf", ".env", ".gitignore", - ".dockerfile", ".docker-compose", ".makefile", - ".java", ".kt", ".go", ".rs", ".rb", ".php", ".swift", ".c", ".cpp", ".h", - ".r", ".lua", ".dart", ".vue", ".svelte", -} - - -def isTextFile(mimeType: Optional[str], fileName: Optional[str] = None) -> bool: - """Check if a file is a text-based file suitable for the editor.""" - if mimeType and mimeType.lower() in TEXT_MIME_TYPES: - return True - if mimeType and mimeType.lower().startswith("text/"): - return True - if fileName: - ext = "." + fileName.rsplit(".", 1)[-1].lower() if "." in fileName else "" - if ext in TEXT_EXTENSIONS: - return True - return False diff --git a/modules/features/codeeditor/fileContextManager.py b/modules/features/codeeditor/fileContextManager.py deleted file mode 100644 index af13f09b..00000000 --- a/modules/features/codeeditor/fileContextManager.py +++ /dev/null @@ -1,84 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""File context manager for CodeEditor feature. -Loads text files from the database and provides them as context for AI calls.""" - -import logging -from typing import List, Optional - -from modules.features.codeeditor.datamodelCodeeditor import FileContext, isTextFile - -logger = logging.getLogger(__name__) - - -async def loadFileContexts(dbManagement, fileIds: List[str]) -> List[FileContext]: - """Load text files from DB and return as FileContext list. - - Args: - dbManagement: interfaceDbManagement instance with user context set - fileIds: list of file IDs to load - """ - contexts = [] - for fileId in fileIds: - fileItem = dbManagement.getFile(fileId) - if not fileItem: - logger.warning(f"File {fileId} not found or no access") - continue - - if not isTextFile(fileItem.mimeType, fileItem.fileName): - logger.warning(f"File {fileItem.fileName} ({fileItem.mimeType}) is not a text file, skipping") - continue - - fileData = dbManagement.getFileData(fileId) - if not fileData: - logger.warning(f"No data for file {fileId}") - continue - - try: - content = fileData.decode("utf-8") - except UnicodeDecodeError: - logger.warning(f"File {fileItem.fileName} is not valid UTF-8, skipping") - continue - - contexts.append(FileContext( - fileId=fileId, - fileName=fileItem.fileName, - content=content, - mimeType=fileItem.mimeType, - sizeBytes=fileItem.fileSize - )) - - logger.info(f"Loaded {len(contexts)} file contexts from {len(fileIds)} requested") - return contexts - - -def listTextFiles(dbManagement) -> List[FileContext]: - """List all text files accessible to the user (metadata only, no content).""" - allFiles = dbManagement.getAllFiles() - textFiles = [] - - if not allFiles: - return textFiles - - for fileItem in allFiles: - if isTextFile(fileItem.mimeType, fileItem.fileName): - modifiedAt = getattr(fileItem, "_modifiedAt", None) or getattr(fileItem, "creationDate", None) - textFiles.append(FileContext( - fileId=fileItem.id, - fileName=fileItem.fileName, - content=None, - mimeType=fileItem.mimeType, - sizeBytes=fileItem.fileSize, - modifiedAt=modifiedAt - )) - - return textFiles - - -def buildFileListContext(dbManagement) -> str: - """Build a compact file list string for the agent prompt (no content, just metadata).""" - textFiles = listTextFiles(dbManagement) - if not textFiles: - return "No text files available." - lines = [f"- {f.fileName} (id: {f.fileId}, size: {f.sizeBytes}B)" for f in textFiles] - return f"Total: {len(lines)} text files\n" + "\n".join(lines) diff --git a/modules/features/codeeditor/mainCodeeditor.py b/modules/features/codeeditor/mainCodeeditor.py deleted file mode 100644 index 4237f3ab..00000000 --- a/modules/features/codeeditor/mainCodeeditor.py +++ /dev/null @@ -1,248 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -CodeEditor Feature Container - Main Module. -Handles feature initialization and RBAC catalog registration. -Cursor-style AI file editing via chat interface. -""" - -import logging -from typing import Dict, List, Any - -logger = logging.getLogger(__name__) - -FEATURE_CODE = "codeeditor" -FEATURE_LABEL = {"en": "Code Editor", "de": "Code Editor", "fr": "Code Editor"} -FEATURE_ICON = "mdi-file-document-edit" - -UI_OBJECTS = [ - { - "objectKey": "ui.feature.codeeditor.editor", - "label": {"en": "Editor", "de": "Editor", "fr": "Editeur"}, - "meta": {"area": "editor"} - }, - { - "objectKey": "ui.feature.codeeditor.workflows", - "label": {"en": "Workflows", "de": "Workflows", "fr": "Workflows"}, - "meta": {"area": "workflows"} - }, -] - -RESOURCE_OBJECTS = [ - { - "objectKey": "resource.feature.codeeditor.start", - "label": {"en": "Start Workflow", "de": "Workflow starten", "fr": "Demarrer workflow"}, - "meta": {"endpoint": "/api/codeeditor/{instanceId}/start/stream", "method": "POST"} - }, - { - "objectKey": "resource.feature.codeeditor.stop", - "label": {"en": "Stop Workflow", "de": "Workflow stoppen", "fr": "Arreter workflow"}, - "meta": {"endpoint": "/api/codeeditor/{instanceId}/{workflowId}/stop", "method": "POST"} - }, - { - "objectKey": "resource.feature.codeeditor.chatData", - "label": {"en": "Get Chat Data", "de": "Chat-Daten abrufen", "fr": "Recuperer donnees chat"}, - "meta": {"endpoint": "/api/codeeditor/{instanceId}/{workflowId}/chatData", "method": "GET"} - }, - { - "objectKey": "resource.feature.codeeditor.files", - "label": {"en": "Manage Files", "de": "Dateien verwalten", "fr": "Gerer fichiers"}, - "meta": {"endpoint": "/api/codeeditor/{instanceId}/files", "method": "GET"} - }, - { - "objectKey": "resource.feature.codeeditor.apply", - "label": {"en": "Apply Edit", "de": "Aenderung anwenden", "fr": "Appliquer modification"}, - "meta": {"endpoint": "/api/codeeditor/{instanceId}/{workflowId}/apply", "method": "POST"} - }, -] - -TEMPLATE_ROLES = [ - { - "roleLabel": "codeeditor-viewer", - "description": { - "en": "Code Editor Viewer - View editor (read-only)", - "de": "Code Editor Betrachter - Editor ansehen (nur lesen)", - "fr": "Visualiseur Code Editor - Consulter l'editeur (lecture seule)" - }, - "accessRules": [ - {"context": "UI", "item": "ui.feature.codeeditor.editor", "view": True}, - {"context": "DATA", "item": None, "view": True, "read": "m", "create": "n", "update": "n", "delete": "n"}, - ] - }, - { - "roleLabel": "codeeditor-user", - "description": { - "en": "Code Editor User - Use editor and workflows", - "de": "Code Editor Benutzer - Editor und Workflows nutzen", - "fr": "Utilisateur Code Editor - Utiliser l'editeur et les workflows" - }, - "accessRules": [ - {"context": "UI", "item": "ui.feature.codeeditor.editor", "view": True}, - {"context": "UI", "item": "ui.feature.codeeditor.workflows", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.codeeditor.start", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.codeeditor.stop", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.codeeditor.chatData", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.codeeditor.files", "view": True}, - {"context": "RESOURCE", "item": "resource.feature.codeeditor.apply", "view": True}, - {"context": "DATA", "item": None, "view": True, "read": "m", "create": "m", "update": "m", "delete": "m"}, - ] - }, - { - "roleLabel": "codeeditor-admin", - "description": { - "en": "Code Editor Admin - Full access to code editor", - "de": "Code Editor Admin - Vollzugriff auf Code Editor", - "fr": "Administrateur Code Editor - Acces complet au code editor" - }, - "accessRules": [ - {"context": "UI", "item": None, "view": True}, - {"context": "RESOURCE", "item": None, "view": True}, - {"context": "DATA", "item": None, "view": True, "read": "a", "create": "a", "update": "a", "delete": "a"}, - ] - }, -] - - -def getFeatureDefinition() -> Dict[str, Any]: - """Return the feature definition for registration.""" - return { - "code": FEATURE_CODE, - "label": FEATURE_LABEL, - "icon": FEATURE_ICON, - "autoCreateInstance": True, - } - - -def getUiObjects() -> List[Dict[str, Any]]: - """Return UI objects for RBAC catalog registration.""" - return UI_OBJECTS - - -def getResourceObjects() -> List[Dict[str, Any]]: - """Return resource objects for RBAC catalog registration.""" - return RESOURCE_OBJECTS - - -def getTemplateRoles() -> List[Dict[str, Any]]: - """Return template roles for this feature.""" - return TEMPLATE_ROLES - - -def registerFeature(catalogService) -> bool: - """Register this feature's RBAC objects in the catalog.""" - try: - for uiObj in UI_OBJECTS: - catalogService.registerUiObject( - featureCode=FEATURE_CODE, - objectKey=uiObj["objectKey"], - label=uiObj["label"], - meta=uiObj.get("meta") - ) - - for resObj in RESOURCE_OBJECTS: - catalogService.registerResourceObject( - featureCode=FEATURE_CODE, - objectKey=resObj["objectKey"], - label=resObj["label"], - meta=resObj.get("meta") - ) - - _syncTemplateRolesToDb() - - logger.info(f"Feature '{FEATURE_CODE}' registered {len(UI_OBJECTS)} UI objects and {len(RESOURCE_OBJECTS)} resource objects") - return True - - except Exception as e: - logger.error(f"Failed to register feature '{FEATURE_CODE}': {e}") - return False - - -def _syncTemplateRolesToDb() -> int: - """Sync template roles and their AccessRules to the database.""" - try: - from modules.interfaces.interfaceDbApp import getRootInterface - from modules.datamodels.datamodelRbac import Role, AccessRule, AccessRuleContext - - rootInterface = getRootInterface() - - existingRoles = rootInterface.getRolesByFeatureCode(FEATURE_CODE) - templateRoles = [r for r in existingRoles if r.mandateId is None] - existingRoleLabels = {r.roleLabel: str(r.id) for r in templateRoles} - - createdCount = 0 - for roleTemplate in TEMPLATE_ROLES: - roleLabel = roleTemplate["roleLabel"] - - if roleLabel in existingRoleLabels: - roleId = existingRoleLabels[roleLabel] - _ensureAccessRulesForRole(rootInterface, roleId, roleTemplate.get("accessRules", [])) - else: - newRole = Role( - roleLabel=roleLabel, - description=roleTemplate.get("description", {}), - featureCode=FEATURE_CODE, - mandateId=None, - featureInstanceId=None, - isSystemRole=False - ) - createdRole = rootInterface.db.recordCreate(Role, newRole.model_dump()) - roleId = createdRole.get("id") - _ensureAccessRulesForRole(rootInterface, roleId, roleTemplate.get("accessRules", [])) - logger.info(f"Created template role '{roleLabel}' with ID {roleId}") - createdCount += 1 - - if createdCount > 0: - logger.info(f"Feature '{FEATURE_CODE}': Created {createdCount} template roles") - - return createdCount - - except Exception as e: - logger.error(f"Error syncing template roles for feature '{FEATURE_CODE}': {e}") - return 0 - - -def _ensureAccessRulesForRole(rootInterface, roleId: str, ruleTemplates: List[Dict[str, Any]]) -> int: - """Ensure AccessRules exist for a role based on templates.""" - from modules.datamodels.datamodelRbac import AccessRule, AccessRuleContext - - existingRules = rootInterface.getAccessRulesByRole(roleId) - existingSignatures = set() - for rule in existingRules: - sig = (rule.context.value if rule.context else None, rule.item) - existingSignatures.add(sig) - - createdCount = 0 - for template in ruleTemplates: - context = template.get("context", "UI") - item = template.get("item") - sig = (context, item) - - if sig in existingSignatures: - continue - - if context == "UI": - contextEnum = AccessRuleContext.UI - elif context == "DATA": - contextEnum = AccessRuleContext.DATA - elif context == "RESOURCE": - contextEnum = AccessRuleContext.RESOURCE - else: - contextEnum = context - - newRule = AccessRule( - roleId=roleId, - context=contextEnum, - item=item, - view=template.get("view", False), - read=template.get("read"), - create=template.get("create"), - update=template.get("update"), - delete=template.get("delete"), - ) - rootInterface.db.recordCreate(AccessRule, newRule.model_dump()) - createdCount += 1 - - if createdCount > 0: - logger.debug(f"Created {createdCount} AccessRules for role {roleId}") - - return createdCount diff --git a/modules/features/codeeditor/promptAssembly.py b/modules/features/codeeditor/promptAssembly.py deleted file mode 100644 index 95f4ce91..00000000 --- a/modules/features/codeeditor/promptAssembly.py +++ /dev/null @@ -1,183 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""Prompt assembly for the CodeEditor feature. -Builds Cursor-style system prompts with file context and format instructions.""" - -import logging -from typing import List, Optional, Dict, Any - -from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationTypeEnum -from modules.features.codeeditor.datamodelCodeeditor import FileContext - -logger = logging.getLogger(__name__) - -SYSTEM_PROMPT = """You are an AI assistant for text and code file editing. You receive files as context and can suggest changes. - -## Rules for file edits -- Use ```file_edit``` blocks for file changes -- Each file_edit block must contain: fileName, oldContent (exact text to replace), newContent (replacement text) -- Explain changes in normal text before or after the block -- oldContent must EXACTLY match existing content (including whitespace and indentation) -- You may propose edits to multiple files in one response - -## Response format -Normal text is displayed as explanation. -File changes must use this format: - -```file_edit -fileName: -oldContent: | - -newContent: | - -``` - -Code examples (without edits) use standard markdown code blocks: -```language -code here -``` - -## Important -- Only edit files that are provided in context -- Make minimal, targeted changes -- Preserve existing formatting and style -- If a task is unclear, ask for clarification instead of guessing""" - - -def buildRequest( - userPrompt: str, - fileContexts: List[FileContext], - chatHistory: Optional[List[Dict[str, Any]]] = None -) -> AiCallRequest: - """Build an AiCallRequest with system prompt, file context, and user prompt.""" - systemPart = SYSTEM_PROMPT - fileContextPart = _buildFileContext(fileContexts) - historyPart = _buildChatHistory(chatHistory) if chatHistory else "" - - fullPrompt = systemPart - if historyPart: - fullPrompt += f"\n\n## Previous conversation\n{historyPart}" - fullPrompt += f"\n\n## User request\n{userPrompt}" - - return AiCallRequest( - prompt=fullPrompt, - context=fileContextPart if fileContextPart else None, - options=AiCallOptions( - operationType=OperationTypeEnum.DATA_ANALYSE, - temperature=0.0, - compressPrompt=False, - compressContext=False, - resultFormat="txt" - ) - ) - - -def _buildFileContext(fileContexts: List[FileContext]) -> str: - """Build the file context string with line numbers.""" - if not fileContexts: - return "" - - parts = [] - for fc in fileContexts: - if not fc.content: - continue - lines = fc.content.split("\n") - numberedLines = [f"{i + 1}|{line}" for i, line in enumerate(lines)] - numbered = "\n".join(numberedLines) - parts.append(f"--- FILE: {fc.fileName} ---\n{numbered}\n--- END FILE ---") - - return "\n\n".join(parts) - - -def buildAgentRequest( - userPrompt: Optional[str], - fileListContext: str, - conversationHistory: List[Dict[str, Any]] -) -> AiCallRequest: - """Build an AiCallRequest for agent mode with tool definitions and conversation history.""" - from modules.features.codeeditor.toolRegistry import formatToolDefinitions - - systemPrompt = _AGENT_SYSTEM_PROMPT.replace("{{TOOL_DEFINITIONS}}", formatToolDefinitions()) - - if not conversationHistory: - fullPrompt = systemPrompt - context = f"## Available files\n{fileListContext}\n\n## Task\n{userPrompt}" - else: - fullPrompt = systemPrompt - historyText = _buildConversationHistory(conversationHistory) - context = f"## Available files\n{fileListContext}\n\n## Conversation\n{historyText}" - - return AiCallRequest( - prompt=fullPrompt, - context=context, - options=AiCallOptions( - operationType=OperationTypeEnum.DATA_ANALYSE, - temperature=0.0, - compressPrompt=False, - compressContext=False, - resultFormat="txt" - ) - ) - - -_AGENT_SYSTEM_PROMPT = """You are an AI agent for file analysis and editing. You work autonomously by using tools to read files, search content, and propose edits. - -## Available tools -{{TOOL_DEFINITIONS}} - -## How to call tools -Use this exact format for each tool call: - -```tool_call -tool: -args: {"param": "value"} -``` - -## Rules -- Read files ONE AT A TIME with read_file, never assume file contents -- First create a plan, then execute it step by step -- Use search_files to find relevant files before reading them -- Use list_files to discover what files are available -- For file changes, use ```file_edit``` blocks (same format as before) -- You may combine text explanations, tool calls, and file edits in one response -- When you are DONE and need no more tool calls, simply respond with text only (no tool_call blocks) -- Keep responses focused and efficient - -## file_edit format (for changes) -```file_edit -fileName: -oldContent: | - -newContent: | - -```""" - - -def _buildConversationHistory(history: List[Dict[str, Any]]) -> str: - """Build the full conversation history for agent multi-turn context.""" - parts = [] - for msg in history: - role = msg.get("role", "unknown") - content = msg.get("content", "") - if role == "tool_result": - toolName = msg.get("toolName", "") - parts.append(f"[Tool Result - {toolName}]:\n{content}") - else: - parts.append(f"[{role}]:\n{content}") - return "\n\n".join(parts) - - -def _buildChatHistory(chatHistory: List[Dict[str, Any]]) -> str: - """Build a condensed chat history string for multi-turn context.""" - if not chatHistory: - return "" - - parts = [] - for msg in chatHistory[-10:]: - role = msg.get("role", "unknown") - content = msg.get("content", "") - if len(content) > 500: - content = content[:500] + "..." - parts.append(f"[{role}]: {content}") - - return "\n".join(parts) diff --git a/modules/features/codeeditor/responseParser.py b/modules/features/codeeditor/responseParser.py deleted file mode 100644 index 8003e7b2..00000000 --- a/modules/features/codeeditor/responseParser.py +++ /dev/null @@ -1,184 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""Response parser for the CodeEditor feature. -Parses AI responses into typed segments (text, code_block, file_edit, tool_call).""" - -import logging -import json -import re -from typing import List, Optional - -from modules.features.codeeditor.datamodelCodeeditor import ResponseSegment, SegmentTypeEnum - -logger = logging.getLogger(__name__) - -_FENCE_PATTERN = re.compile(r"^```(\w*)\s*$", re.MULTILINE) - - -def parseResponse(rawContent: str) -> List[ResponseSegment]: - """Parse an AI response into typed segments.""" - if not rawContent or not rawContent.strip(): - return [] - - segments = [] - lines = rawContent.split("\n") - i = 0 - - textBuffer = [] - - while i < len(lines): - line = lines[i] - - match = _FENCE_PATTERN.match(line) - if match: - if textBuffer: - _flushTextBuffer(textBuffer, segments) - textBuffer = [] - - lang = match.group(1).strip() - blockLines, endIdx = _collectBlock(lines, i + 1) - blockContent = "\n".join(blockLines) - - if lang == "file_edit": - segment = _parseFileEditBlock(blockContent) - if segment: - segments.append(segment) - else: - segments.append(ResponseSegment( - type=SegmentTypeEnum.CODE_BLOCK, - content=blockContent, - language="text" - )) - elif lang == "tool_call": - segment = _parseToolCallBlock(blockContent) - if segment: - segments.append(segment) - else: - segments.append(ResponseSegment( - type=SegmentTypeEnum.CODE_BLOCK, - content=blockContent, - language="text" - )) - else: - segments.append(ResponseSegment( - type=SegmentTypeEnum.CODE_BLOCK, - content=blockContent, - language=lang or "text" - )) - - i = endIdx + 1 - else: - textBuffer.append(line) - i += 1 - - if textBuffer: - _flushTextBuffer(textBuffer, segments) - - return segments - - -def hasToolCalls(segments: List[ResponseSegment]) -> bool: - """Check if any segments contain tool calls.""" - return any(s.type == SegmentTypeEnum.TOOL_CALL for s in segments) - - -def _collectBlock(lines: List[str], startIdx: int) -> tuple: - """Collect lines inside a fenced code block until closing ```.""" - blockLines = [] - idx = startIdx - while idx < len(lines): - if lines[idx].strip() == "```": - return blockLines, idx - blockLines.append(lines[idx]) - idx += 1 - return blockLines, idx - - -def _flushTextBuffer(buffer: List[str], segments: List[ResponseSegment]): - """Flush accumulated text lines into a text segment.""" - text = "\n".join(buffer).strip() - buffer.clear() - if text: - segments.append(ResponseSegment( - type=SegmentTypeEnum.TEXT, - content=text - )) - - -def _parseFileEditBlock(blockContent: str) -> Optional[ResponseSegment]: - """Parse a file_edit block into a ResponseSegment with fileName, oldContent, newContent.""" - fields = {"fileName": None, "oldContent": None, "newContent": None} - currentField = None - currentLines = [] - - for line in blockContent.split("\n"): - stripped = line.strip() - - newField = None - for key in ("fileName", "oldContent", "newContent"): - if stripped.startswith(f"{key}:"): - newField = key - break - - if newField: - if currentField and currentLines: - fields[currentField] = "\n".join(currentLines) - currentField = newField - value = stripped[len(f"{newField}:"):].strip() - if newField == "fileName": - fields["fileName"] = value if value else None - currentField = None - currentLines = [] - else: - currentLines = [value] if value and value != "|" else [] - else: - if currentField in ("oldContent", "newContent"): - dedented = line[2:] if line.startswith(" ") else line - currentLines.append(dedented) - - if currentField and currentLines: - fields[currentField] = "\n".join(currentLines) - - if not fields["fileName"]: - logger.warning("file_edit block missing fileName") - return None - if fields["newContent"] is None: - logger.warning(f"file_edit block for {fields['fileName']} missing newContent") - return None - - return ResponseSegment( - type=SegmentTypeEnum.FILE_EDIT, - content=f"Edit: {fields['fileName']}", - fileName=fields["fileName"], - oldContent=fields["oldContent"], - newContent=fields["newContent"] - ) - - -def _parseToolCallBlock(blockContent: str) -> Optional[ResponseSegment]: - """Parse a tool_call block into a ResponseSegment with toolName and toolArgs.""" - toolName = None - toolArgs = {} - - for line in blockContent.split("\n"): - stripped = line.strip() - if stripped.startswith("tool:"): - toolName = stripped[len("tool:"):].strip() - elif stripped.startswith("args:"): - argsStr = stripped[len("args:"):].strip() - try: - toolArgs = json.loads(argsStr) - except json.JSONDecodeError: - logger.warning(f"Could not parse tool args as JSON: {argsStr}") - toolArgs = {"raw": argsStr} - - if not toolName: - logger.warning("tool_call block missing tool name") - return None - - return ResponseSegment( - type=SegmentTypeEnum.TOOL_CALL, - content=f"Tool: {toolName}", - toolName=toolName, - toolArgs=toolArgs - ) diff --git a/modules/features/codeeditor/routeFeatureCodeeditor.py b/modules/features/codeeditor/routeFeatureCodeeditor.py deleted file mode 100644 index ef239b35..00000000 --- a/modules/features/codeeditor/routeFeatureCodeeditor.py +++ /dev/null @@ -1,395 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -""" -CodeEditor Feature Routes. -SSE-based endpoints for Cursor-style AI file editing. -""" - -import logging -import json -import asyncio -from typing import Optional, Dict, Any, List - -from fastapi import APIRouter, HTTPException, Depends, Body, Path, Query, Request -from fastapi.responses import StreamingResponse - -from modules.auth import limiter, getRequestContext, RequestContext -from modules.interfaces import interfaceDbChat, interfaceDbManagement -from modules.interfaces.interfaceAiObjects import AiObjects -from modules.datamodels.datamodelChat import UserInputRequest -from modules.serviceCenter.core.serviceStreaming import get_event_manager -from modules.features.codeeditor import codeEditorProcessor, fileContextManager -from modules.features.codeeditor.datamodelCodeeditor import FileEditProposal, EditStatusEnum - -logger = logging.getLogger(__name__) - -router = APIRouter( - prefix="/api/codeeditor", - tags=["Code Editor Feature"], - responses={404: {"description": "Not found"}} -) - -_aiObjects: Optional[AiObjects] = None - - -async def _getAiObjects() -> AiObjects: - """Lazy-init singleton for AiObjects.""" - global _aiObjects - if _aiObjects is None: - _aiObjects = await AiObjects.create() - return _aiObjects - - -def _getServiceChat(context: RequestContext, featureInstanceId: str = None): - """Get chat interface with feature instance context.""" - return interfaceDbChat.getInterface( - context.user, - mandateId=str(context.mandateId) if context.mandateId else None, - featureInstanceId=featureInstanceId - ) - - -def _getDbManagement(context: RequestContext, featureInstanceId: str = None): - """Get management interface with user context for file access.""" - return interfaceDbManagement.getInterface( - context.user, - mandateId=str(context.mandateId) if context.mandateId else None, - featureInstanceId=featureInstanceId - ) - - -def _validateInstanceAccess(instanceId: str, context: RequestContext) -> str: - """Validate user has access to the feature instance. Returns mandateId.""" - from modules.interfaces.interfaceDbApp import getRootInterface - - rootInterface = getRootInterface() - instance = rootInterface.getFeatureInstance(instanceId) - if not instance: - raise HTTPException(status_code=404, detail=f"Feature instance {instanceId} not found") - - featureAccess = rootInterface.getFeatureAccess(str(context.user.id), instanceId) - if not featureAccess or not featureAccess.enabled: - raise HTTPException(status_code=403, detail="Access denied to this feature instance") - - return str(instance.mandateId) if instance.mandateId else None - - -@router.post("/{instanceId}/start/stream") -@limiter.limit("60/minute") -async def streamCodeeditorStart( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: Optional[str] = Query(None, description="Optional workflow ID to continue"), - mode: str = Query("simple", description="Processing mode: 'simple' (single AI call) or 'agent' (multi-step with tools)"), - userInput: UserInputRequest = Body(...), - context: RequestContext = Depends(getRequestContext) -): - """Start or continue a CodeEditor workflow with SSE streaming. Supports simple and agent mode.""" - try: - mandateId = _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - dbManagement = _getDbManagement(context, featureInstanceId=instanceId) - aiObjects = await _getAiObjects() - eventManager = get_event_manager() - - if workflowId: - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow {workflowId} not found") - else: - workflow = chatInterface.createWorkflow({ - "workflowMode": "CodeEditor", - "status": "running", - "label": userInput.prompt[:80] if userInput.prompt else "CodeEditor Session", - }) - workflowId = workflow.get("id") if isinstance(workflow, dict) else workflow.id - - queue = eventManager.create_queue(workflowId) - - userMessage = { - "role": "user", - "content": userInput.prompt, - "selectedFiles": userInput.listFileId or [] - } - await eventManager.emit_event(workflowId, "chatdata", { - "type": "message", "item": userMessage - }) - - selectedFileIds = userInput.listFileId or [] - - agentMode = mode.lower() == "agent" - - asyncio.create_task( - codeEditorProcessor.processMessage( - workflowId=workflowId, - userPrompt=userInput.prompt, - selectedFileIds=selectedFileIds, - dbManagement=dbManagement, - interfaceAi=aiObjects, - chatInterface=chatInterface, - eventManager=eventManager, - agentMode=agentMode - ) - ) - - async def _eventStream(): - streamTimeout = 300 - lastActivity = asyncio.get_event_loop().time() - - while True: - now = asyncio.get_event_loop().time() - if now - lastActivity > streamTimeout: - yield f"data: {json.dumps({'type': 'error', 'error': 'Stream timeout'})}\n\n" - break - - if await request.is_disconnected(): - logger.info(f"Client disconnected for workflow {workflowId}") - break - - try: - event = await asyncio.wait_for(queue.get(), timeout=1.0) - lastActivity = asyncio.get_event_loop().time() - - eventType = event.get("type", "") - - if eventType == "chatdata": - yield f"data: {json.dumps(event.get('data', {}))}\n\n" - elif eventType in ("complete", "stopped", "error"): - yield f"data: {json.dumps({'type': eventType, **event.get('data', {})})}\n\n" - break - else: - yield f"data: {json.dumps(event)}\n\n" - - except asyncio.TimeoutError: - yield ": keepalive\n\n" - - await eventManager.cleanup(workflowId) - - return StreamingResponse( - _eventStream(), - media_type="text/event-stream", - headers={ - "Cache-Control": "no-cache", - "Connection": "keep-alive", - "X-Accel-Buffering": "no" - } - ) - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error in streamCodeeditorStart: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -@router.post("/{instanceId}/{workflowId}/stop") -@limiter.limit("120/minute") -async def stopWorkflow( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="Workflow ID"), - context: RequestContext = Depends(getRequestContext) -): - """Stop a running CodeEditor workflow.""" - try: - _validateInstanceAccess(instanceId, context) - eventManager = get_event_manager() - await eventManager.emit_event(workflowId, "stopped", { - "workflowId": workflowId - }, event_category="workflow", message="Workflow stopped by user") - return {"status": "stopped", "workflowId": workflowId} - except HTTPException: - raise - except Exception as e: - logger.error(f"Error stopping workflow: {e}") - raise HTTPException(status_code=500, detail=str(e)) - - -@router.get("/{instanceId}/{workflowId}/chatData") -@limiter.limit("120/minute") -def getWorkflowChatData( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="Workflow ID"), - afterTimestamp: Optional[float] = Query(None, description="Unix timestamp for incremental fetch"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get chat data for a workflow (polling fallback).""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - workflow = chatInterface.getWorkflow(workflowId) - if not workflow: - raise HTTPException(status_code=404, detail=f"Workflow {workflowId} not found") - return chatInterface.getUnifiedChatData(workflowId, afterTimestamp) - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting chat data: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -@router.get("/{instanceId}/workflows") -@limiter.limit("120/minute") -def getWorkflows( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - page: int = Query(1, ge=1), - pageSize: int = Query(20, ge=1, le=100), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """List workflows for this feature instance.""" - try: - _validateInstanceAccess(instanceId, context) - chatInterface = _getServiceChat(context, featureInstanceId=instanceId) - from modules.datamodels.datamodelPagination import PaginationParams - pagination = PaginationParams(page=page, pageSize=pageSize) - return chatInterface.getWorkflows(pagination=pagination) - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting workflows: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -@router.get("/{instanceId}/files") -@limiter.limit("120/minute") -def getFiles( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """List all text files accessible to the user.""" - try: - _validateInstanceAccess(instanceId, context) - dbManagement = _getDbManagement(context, featureInstanceId=instanceId) - textFiles = fileContextManager.listTextFiles(dbManagement) - return { - "files": [f.model_dump(exclude={"content"}) for f in textFiles], - "count": len(textFiles) - } - except HTTPException: - raise - except Exception as e: - logger.error(f"Error listing files: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -@router.get("/{instanceId}/files/{fileId}/content") -@limiter.limit("120/minute") -def getFileContent( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - fileId: str = Path(..., description="File ID"), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Get the text content of a file.""" - try: - _validateInstanceAccess(instanceId, context) - dbManagement = _getDbManagement(context, featureInstanceId=instanceId) - - fileItem = dbManagement.getFile(fileId) - if not fileItem: - raise HTTPException(status_code=404, detail=f"File {fileId} not found") - - fileData = dbManagement.getFileData(fileId) - if not fileData: - raise HTTPException(status_code=404, detail=f"No data for file {fileId}") - - try: - content = fileData.decode("utf-8") - except UnicodeDecodeError: - raise HTTPException(status_code=400, detail="File is not valid UTF-8 text") - - return { - "fileId": fileId, - "fileName": fileItem.fileName, - "mimeType": fileItem.mimeType, - "content": content - } - except HTTPException: - raise - except Exception as e: - logger.error(f"Error getting file content: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -@router.post("/{instanceId}/{workflowId}/apply") -@limiter.limit("60/minute") -async def applyEdit( - request: Request, - instanceId: str = Path(..., description="Feature instance ID"), - workflowId: str = Path(..., description="Workflow ID"), - proposalData: Dict[str, Any] = Body(...), - context: RequestContext = Depends(getRequestContext) -) -> Dict[str, Any]: - """Accept a file edit proposal. Updates existing file or creates new one.""" - try: - _validateInstanceAccess(instanceId, context) - dbManagement = _getDbManagement(context, featureInstanceId=instanceId) - - fileId = proposalData.get("fileId", "") - newContent = proposalData.get("newContent") - fileName = proposalData.get("fileName", "") - - if newContent is None: - raise HTTPException(status_code=400, detail="newContent is required") - - contentBytes = newContent.encode("utf-8") - isNewFile = not fileId or fileId.startswith("unknown-") - - if isNewFile: - mimeType = _guessMimeType(fileName) - fileItem = dbManagement.createFile(fileName, mimeType, contentBytes) - resultFileId = fileItem.id - resultFileName = fileItem.fileName - else: - fileItem = dbManagement.getFile(fileId) - if not fileItem: - raise HTTPException(status_code=404, detail=f"File {fileId} not found") - success = dbManagement.createFileData(fileId, contentBytes) - if not success: - raise HTTPException(status_code=500, detail="Failed to store updated file content") - resultFileId = fileId - resultFileName = fileName or fileItem.fileName - - eventManager = get_event_manager() - await eventManager.emit_event(workflowId, "chatdata", { - "type": "file_version", - "item": { - "fileId": resultFileId, - "fileName": resultFileName, - "status": "accepted", - "isNew": isNewFile - } - }) - - return { - "status": "accepted", - "fileId": resultFileId, - "fileName": resultFileName, - "isNew": isNewFile - } - - except HTTPException: - raise - except Exception as e: - logger.error(f"Error applying edit: {e}", exc_info=True) - raise HTTPException(status_code=500, detail=str(e)) - - -_MIME_MAP = { - ".md": "text/markdown", ".txt": "text/plain", ".json": "application/json", - ".yaml": "application/yaml", ".yml": "application/yaml", ".xml": "application/xml", - ".csv": "text/csv", ".py": "text/x-python", ".js": "text/javascript", - ".ts": "text/x-typescript", ".html": "text/html", ".css": "text/css", - ".sql": "text/x-sql", ".sh": "text/x-shellscript", -} - - -def _guessMimeType(fileName: str) -> str: - """Guess MIME type from file extension.""" - if not fileName or "." not in fileName: - return "text/plain" - ext = "." + fileName.rsplit(".", 1)[-1].lower() - return _MIME_MAP.get(ext, "text/plain") diff --git a/modules/features/codeeditor/toolRegistry.py b/modules/features/codeeditor/toolRegistry.py deleted file mode 100644 index 69256671..00000000 --- a/modules/features/codeeditor/toolRegistry.py +++ /dev/null @@ -1,157 +0,0 @@ -# Copyright (c) 2025 Patrick Motsch -# All rights reserved. -"""Tool registry and dispatcher for the CodeEditor agent loop. -Defines available tools and executes them against the file context manager.""" - -import logging -import time -import fnmatch -from typing import Dict, Any, List - -from modules.features.codeeditor.datamodelCodeeditor import ToolResult - -logger = logging.getLogger(__name__) - -TOOL_DEFINITIONS = [ - { - "name": "read_file", - "description": "Read the full content of a single file by its fileId.", - "parameters": {"fileId": "string (required)"} - }, - { - "name": "list_files", - "description": "List all available text files with metadata (name, size, mimeType). Optionally filter by glob pattern.", - "parameters": {"filter": "string (optional, glob pattern e.g. '*.py')"} - }, - { - "name": "search_files", - "description": "Search all file contents for a text query. Returns matching lines with file name and line number.", - "parameters": {"query": "string (required)", "fileType": "string (optional, extension e.g. 'py')"} - }, -] - - -async def dispatch(toolName: str, toolArgs: Dict[str, Any], dbManagement) -> ToolResult: - """Execute a tool and return the result.""" - startTime = time.time() - try: - if toolName == "read_file": - result = await _toolReadFile(toolArgs, dbManagement) - elif toolName == "list_files": - result = _toolListFiles(toolArgs, dbManagement) - elif toolName == "search_files": - result = await _toolSearchFiles(toolArgs, dbManagement) - else: - result = f"Unknown tool: {toolName}" - return ToolResult(toolName=toolName, result=result, success=False, - executionTime=time.time() - startTime) - - return ToolResult(toolName=toolName, result=result, success=True, - executionTime=time.time() - startTime) - except Exception as e: - logger.error(f"Tool {toolName} failed: {e}", exc_info=True) - return ToolResult(toolName=toolName, result=f"Error: {str(e)}", success=False, - executionTime=time.time() - startTime) - - -async def _toolReadFile(args: Dict[str, Any], dbManagement) -> str: - """Read a single file's content.""" - fileId = args.get("fileId", "") - if not fileId: - return "Error: fileId is required" - - fileItem = dbManagement.getFile(fileId) - if not fileItem: - return f"Error: File {fileId} not found" - - fileData = dbManagement.getFileData(fileId) - if not fileData: - return f"Error: No data for file {fileId}" - - try: - content = fileData.decode("utf-8") - except UnicodeDecodeError: - return f"Error: File {fileItem.fileName} is not valid UTF-8" - - lines = content.split("\n") - numbered = "\n".join([f"{i + 1}|{line}" for i, line in enumerate(lines)]) - return f"--- FILE: {fileItem.fileName} (id: {fileId}) ---\n{numbered}\n--- END FILE ---" - - -def _toolListFiles(args: Dict[str, Any], dbManagement) -> str: - """List all text files, optionally filtered by glob pattern.""" - from modules.features.codeeditor.datamodelCodeeditor import isTextFile - - filterPattern = args.get("filter", "") - allFiles = dbManagement.getAllFiles() - if not allFiles: - return "No files found." - - lines = [] - for f in allFiles: - if not isTextFile(f.mimeType, f.fileName): - continue - if filterPattern and not fnmatch.fnmatch(f.fileName, filterPattern): - continue - lines.append(f"- {f.fileName} (id: {f.id}, size: {f.fileSize}B, type: {f.mimeType})") - - if not lines: - return "No matching text files found." - return f"Available files ({len(lines)}):\n" + "\n".join(lines) - - -async def _toolSearchFiles(args: Dict[str, Any], dbManagement) -> str: - """Search file contents for a query string.""" - from modules.features.codeeditor.datamodelCodeeditor import isTextFile - - query = args.get("query", "") - if not query: - return "Error: query is required" - - fileType = args.get("fileType", "") - allFiles = dbManagement.getAllFiles() - if not allFiles: - return "No files to search." - - hits = [] - maxHits = 50 - queryLower = query.lower() - - for f in allFiles: - if not isTextFile(f.mimeType, f.fileName): - continue - if fileType and not f.fileName.endswith(f".{fileType}"): - continue - - fileData = dbManagement.getFileData(f.id) - if not fileData: - continue - - try: - content = fileData.decode("utf-8") - except UnicodeDecodeError: - continue - - for lineNum, line in enumerate(content.split("\n"), 1): - if queryLower in line.lower(): - hits.append(f"{f.fileName}:{lineNum}: {line.strip()}") - if len(hits) >= maxHits: - break - if len(hits) >= maxHits: - break - - if not hits: - return f"No matches found for '{query}'." - result = f"Search results for '{query}' ({len(hits)} matches):\n" + "\n".join(hits) - if len(hits) >= maxHits: - result += f"\n... (truncated at {maxHits} matches)" - return result - - -def formatToolDefinitions() -> str: - """Format tool definitions for inclusion in the system prompt.""" - parts = [] - for tool in TOOL_DEFINITIONS: - params = ", ".join([f"{k}: {v}" for k, v in tool["parameters"].items()]) - parts.append(f"- **{tool['name']}**: {tool['description']}\n Parameters: {{{params}}}") - return "\n".join(parts) diff --git a/modules/features/workspace/__init__.py b/modules/features/workspace/__init__.py index e4d7dac9..2e48ea1c 100644 --- a/modules/features/workspace/__init__.py +++ b/modules/features/workspace/__init__.py @@ -1,3 +1,3 @@ # Copyright (c) 2025 Patrick Motsch # All rights reserved. -"""Unified AI Workspace feature -- merges Codeeditor, Chatbot, and Playground.""" +"""Unified AI Workspace feature.""" diff --git a/modules/features/workspace/mainWorkspace.py b/modules/features/workspace/mainWorkspace.py index 0737bf13..5da67a45 100644 --- a/modules/features/workspace/mainWorkspace.py +++ b/modules/features/workspace/mainWorkspace.py @@ -3,7 +3,7 @@ """ Workspace Feature Container - Main Module. Handles feature initialization and RBAC catalog registration. -Unified AI Workspace combining Codeeditor, Chatbot, and Playground capabilities. +Unified AI Workspace feature. """ import logging @@ -21,6 +21,11 @@ UI_OBJECTS = [ "label": {"en": "Dashboard", "de": "Dashboard", "fr": "Tableau de bord"}, "meta": {"area": "dashboard"} }, + { + "objectKey": "ui.feature.workspace.editor", + "label": {"en": "Editor", "de": "Editor", "fr": "Editeur"}, + "meta": {"area": "editor"} + }, { "objectKey": "ui.feature.workspace.settings", "label": {"en": "Settings", "de": "Einstellungen", "fr": "Parametres"}, @@ -59,6 +64,11 @@ RESOURCE_OBJECTS = [ "label": {"en": "Voice Input/Output", "de": "Spracheingabe/-ausgabe", "fr": "Entree/sortie vocale"}, "meta": {"endpoint": "/api/workspace/{instanceId}/voice/*", "method": "POST"} }, + { + "objectKey": "resource.feature.workspace.edits", + "label": {"en": "Review File Edits", "de": "Datei-Aenderungen pruefen", "fr": "Verifier les modifications de fichiers"}, + "meta": {"endpoint": "/api/workspace/{instanceId}/edit/*", "method": "POST"} + }, ] TEMPLATE_ROLES = [ @@ -71,6 +81,7 @@ TEMPLATE_ROLES = [ }, "accessRules": [ {"context": "UI", "item": "ui.feature.workspace.dashboard", "view": True}, + {"context": "UI", "item": "ui.feature.workspace.editor", "view": True}, {"context": "UI", "item": "ui.feature.workspace.settings", "view": True}, {"context": "DATA", "item": None, "view": True, "read": "m", "create": "n", "update": "n", "delete": "n"}, ] @@ -84,6 +95,7 @@ TEMPLATE_ROLES = [ }, "accessRules": [ {"context": "UI", "item": "ui.feature.workspace.dashboard", "view": True}, + {"context": "UI", "item": "ui.feature.workspace.editor", "view": True}, {"context": "UI", "item": "ui.feature.workspace.settings", "view": True}, {"context": "RESOURCE", "item": "resource.feature.workspace.start", "view": True}, {"context": "RESOURCE", "item": "resource.feature.workspace.stop", "view": True}, @@ -91,6 +103,7 @@ TEMPLATE_ROLES = [ {"context": "RESOURCE", "item": "resource.feature.workspace.folders", "view": True}, {"context": "RESOURCE", "item": "resource.feature.workspace.datasources", "view": True}, {"context": "RESOURCE", "item": "resource.feature.workspace.voice", "view": True}, + {"context": "RESOURCE", "item": "resource.feature.workspace.edits", "view": True}, {"context": "DATA", "item": None, "view": True, "read": "m", "create": "m", "update": "m", "delete": "m"}, ] }, diff --git a/modules/features/workspace/routeFeatureWorkspace.py b/modules/features/workspace/routeFeatureWorkspace.py index 62ebfdbc..1ea8a93a 100644 --- a/modules/features/workspace/routeFeatureWorkspace.py +++ b/modules/features/workspace/routeFeatureWorkspace.py @@ -2,8 +2,7 @@ # All rights reserved. """Unified AI Workspace routes. -SSE-based endpoints that combine the capabilities of Codeeditor, Chatbot, -and Playground into a single agent-driven workspace. +SSE-based endpoints for the agent-driven AI Workspace. """ import logging @@ -19,7 +18,7 @@ from modules.auth import limiter, getRequestContext, RequestContext from modules.interfaces import interfaceDbChat, interfaceDbManagement from modules.interfaces.interfaceAiObjects import AiObjects from modules.serviceCenter.core.serviceStreaming import get_event_manager -from modules.serviceCenter.services.serviceAgent.datamodelAgent import AgentEventTypeEnum +from modules.serviceCenter.services.serviceAgent.datamodelAgent import AgentEventTypeEnum, PendingFileEdit logger = logging.getLogger(__name__) @@ -32,6 +31,41 @@ router = APIRouter( _aiObjects: Optional[AiObjects] = None +class _InstanceEdits: + """Pending file edits for a single workspace instance.""" + + def __init__(self): + self._edits: Dict[str, PendingFileEdit] = {} + + def add(self, edit: PendingFileEdit) -> None: + self._edits[edit.id] = edit + + def get(self, editId: str) -> Optional[PendingFileEdit]: + return self._edits.get(editId) + + def getPending(self) -> List[PendingFileEdit]: + return [e for e in self._edits.values() if e.status == "pending"] + + def items(self): + return self._edits.items() + + +class _PendingEditsStore: + """Global store for pending file edits across all workspace instances.""" + + def __init__(self): + self._instances: Dict[str, _InstanceEdits] = {} + + def forInstance(self, instanceId: str) -> _InstanceEdits: + """Get-or-create the edit collection for a workspace instance.""" + if instanceId not in self._instances: + self._instances[instanceId] = _InstanceEdits() + return self._instances[instanceId] + + +_pendingEditsStore = _PendingEditsStore() + + class WorkspaceInputRequest(BaseModel): """Prompt input for the unified workspace.""" prompt: str = Field(description="User prompt text") @@ -41,6 +75,7 @@ class WorkspaceInputRequest(BaseModel): voiceMode: bool = Field(default=False, description="Enable voice response") workflowId: Optional[str] = Field(default=None, description="Continue existing workflow") userLanguage: str = Field(default="en", description="User language code") + allowedProviders: List[str] = Field(default_factory=list, description="Restrict AI to these providers") async def _getAiObjects() -> AiObjects: @@ -264,6 +299,7 @@ async def streamWorkspaceStart( eventManager=eventManager, userLanguage=userInput.userLanguage, instanceConfig=instanceConfig, + allowedProviders=userInput.allowedProviders, ) ) eventManager.register_agent_task(queueId, agentTask) @@ -317,6 +353,7 @@ async def _runWorkspaceAgent( eventManager, userLanguage: str = "en", instanceConfig: Dict[str, Any] = None, + allowedProviders: List[str] = None, ): """Run the serviceAgent loop and forward events to the SSE queue.""" try: @@ -332,6 +369,9 @@ async def _runWorkspaceAgent( chatService = getService("chat", ctx) aiService = getService("ai", ctx) + if allowedProviders: + aiService.services.allowedProviders = allowedProviders + wfRecord = chatInterface.getWorkflow(workflowId) if workflowId else None wfName = "" if wfRecord: @@ -383,6 +423,35 @@ async def _runWorkspaceAgent( if event.type == AgentEventTypeEnum.CHUNK and event.content: accumulatedText += event.content + if event.type == AgentEventTypeEnum.FILE_EDIT_PROPOSAL and event.data: + editData = event.data + editId = editData.get("id", "") + if editId: + pendingEdit = PendingFileEdit( + id=editId, + fileId=editData.get("fileId", ""), + fileName=editData.get("fileName", ""), + mimeType=editData.get("mimeType", ""), + oldContent=editData.get("oldContent", ""), + newContent=editData.get("newContent", ""), + workflowId=workflowId, + ) + _pendingEditsStore.forInstance(instanceId).add(pendingEdit) + logger.info(f"Stored pending edit {editId} for file '{pendingEdit.fileName}' in instance {instanceId}") + await eventManager.emit_event(queueId, "fileEditProposal", { + "type": "fileEditProposal", + "workflowId": workflowId, + "item": { + "id": editId, + "fileId": editData.get("fileId", ""), + "fileName": editData.get("fileName", ""), + "mimeType": editData.get("mimeType", ""), + "oldSize": len(editData.get("oldContent", "")), + "newSize": len(editData.get("newContent", "")), + }, + }) + continue + sseEvent = { "type": event.type.value if hasattr(event.type, "value") else event.type, "workflowId": workflowId, @@ -1103,3 +1172,137 @@ async def testVoice( except Exception as e: logger.error(f"Voice test failed: {e}") raise HTTPException(status_code=500, detail=f"TTS test failed: {str(e)}") + + +# ============================================================================= +# FILE EDIT PROPOSAL ENDPOINTS +# ============================================================================= + + +@router.get("/{instanceId}/pending-edits") +@limiter.limit("30/minute") +async def getPendingEdits( + request: Request, + instanceId: str = Path(...), + context: RequestContext = Depends(getRequestContext), +): + """Return all pending file edit proposals for this workspace instance.""" + _validateInstanceAccess(instanceId, context) + editList = [e.model_dump() for e in _pendingEditsStore.forInstance(instanceId).getPending()] + return JSONResponse({"edits": editList}) + + +@router.post("/{instanceId}/edit/{editId}/accept") +@limiter.limit("30/minute") +async def acceptEdit( + request: Request, + instanceId: str = Path(...), + editId: str = Path(...), + context: RequestContext = Depends(getRequestContext), +): + """Accept a proposed file edit -- applies the new content to the file.""" + _validateInstanceAccess(instanceId, context) + edit = _pendingEditsStore.forInstance(instanceId).get(editId) + if not edit: + raise HTTPException(status_code=404, detail=f"Edit proposal {editId} not found") + if edit.status != "pending": + raise HTTPException(status_code=409, detail=f"Edit proposal is already {edit.status}") + + dbMgmt = _getDbManagement(context, instanceId) + try: + success = dbMgmt.updateFileData(edit.fileId, edit.newContent.encode("utf-8")) + if not success: + raise HTTPException(status_code=500, detail="Failed to update file data") + except HTTPException: + raise + except Exception as e: + logger.error(f"Failed to apply edit {editId}: {e}") + raise HTTPException(status_code=500, detail=f"Failed to apply edit: {str(e)}") + + edit.status = "accepted" + logger.info(f"Edit {editId} accepted for file '{edit.fileName}' in instance {instanceId}") + return JSONResponse({ + "success": True, + "editId": editId, + "fileId": edit.fileId, + "fileName": edit.fileName, + }) + + +@router.post("/{instanceId}/edit/{editId}/reject") +@limiter.limit("30/minute") +async def rejectEdit( + request: Request, + instanceId: str = Path(...), + editId: str = Path(...), + context: RequestContext = Depends(getRequestContext), +): + """Reject a proposed file edit -- discards the change.""" + _validateInstanceAccess(instanceId, context) + edit = _pendingEditsStore.forInstance(instanceId).get(editId) + if not edit: + raise HTTPException(status_code=404, detail=f"Edit proposal {editId} not found") + if edit.status != "pending": + raise HTTPException(status_code=409, detail=f"Edit proposal is already {edit.status}") + + edit.status = "rejected" + logger.info(f"Edit {editId} rejected for file '{edit.fileName}' in instance {instanceId}") + return JSONResponse({ + "success": True, + "editId": editId, + "fileId": edit.fileId, + "fileName": edit.fileName, + }) + + +@router.post("/{instanceId}/edit/accept-all") +@limiter.limit("10/minute") +async def acceptAllEdits( + request: Request, + instanceId: str = Path(...), + context: RequestContext = Depends(getRequestContext), +): + """Accept all pending file edit proposals for this instance.""" + _validateInstanceAccess(instanceId, context) + instanceEdits = _pendingEditsStore.forInstance(instanceId) + dbMgmt = _getDbManagement(context, instanceId) + accepted = [] + errors = [] + + for editId, edit in instanceEdits.items(): + if edit.status != "pending": + continue + try: + success = dbMgmt.updateFileData(edit.fileId, edit.newContent.encode("utf-8")) + if success: + edit.status = "accepted" + accepted.append(editId) + else: + errors.append({"editId": editId, "error": "updateFileData returned False"}) + except Exception as e: + errors.append({"editId": editId, "error": str(e)}) + + logger.info(f"Accepted {len(accepted)} edits for instance {instanceId}, {len(errors)} errors") + return JSONResponse({"accepted": accepted, "errors": errors}) + + +@router.post("/{instanceId}/edit/reject-all") +@limiter.limit("10/minute") +async def rejectAllEdits( + request: Request, + instanceId: str = Path(...), + context: RequestContext = Depends(getRequestContext), +): + """Reject all pending file edit proposals for this instance.""" + _validateInstanceAccess(instanceId, context) + instanceEdits = _pendingEditsStore.forInstance(instanceId) + rejected = [] + + for editId, edit in instanceEdits.items(): + if edit.status != "pending": + continue + edit.status = "rejected" + rejected.append(editId) + + logger.info(f"Rejected {len(rejected)} edits for instance {instanceId}") + return JSONResponse({"rejected": rejected}) diff --git a/modules/interfaces/interfaceAiObjects.py b/modules/interfaces/interfaceAiObjects.py index d53c9b5a..981a6b46 100644 --- a/modules/interfaces/interfaceAiObjects.py +++ b/modules/interfaces/interfaceAiObjects.py @@ -12,6 +12,7 @@ logger = logging.getLogger(__name__) from modules.aicore.aicoreModelRegistry import modelRegistry from modules.aicore.aicoreModelSelector import modelSelector +from modules.aicore.aicoreBase import RateLimitExceededException from modules.datamodels.datamodelAi import ( AiModel, AiCallOptions, @@ -97,15 +98,18 @@ class AiObjects: # Get failover models for this operation type availableModels = modelRegistry.getAvailableModels() - # Filter by allowedProviders if specified (from workflow config) allowedProviders = getattr(options, 'allowedProviders', None) if options else None if allowedProviders: filteredModels = [m for m in availableModels if m.connectorType in allowedProviders] if filteredModels: - logger.info(f"Filtered models by allowedProviders {allowedProviders}: {len(filteredModels)} models (from {len(availableModels)})") availableModels = filteredModels else: - logger.warning(f"No models match allowedProviders {allowedProviders}, using all {len(availableModels)} available models") + errorMsg = f"No models match allowedProviders {allowedProviders} for operation {options.operationType}" + logger.error(errorMsg) + return AiCallResponse( + content=errorMsg, modelName="error", priceCHF=0.0, + processingTime=0.0, bytesSent=0, bytesReceived=0, errorCount=1, + ) failoverModelList = modelSelector.getFailoverModelList(prompt, context, options, availableModels) @@ -122,7 +126,8 @@ class AiObjects: errorCount=1 ) - # Try each model in failover sequence + _MAX_SHORT_RETRY = 15.0 + lastError = None for attempt, model in enumerate(failoverModelList): try: @@ -135,7 +140,32 @@ class AiObjects: logger.info(f"AI call successful with model: {model.name}") return response - + + except RateLimitExceededException as rle: + retryAfter = rle.retryAfterSeconds + lastError = rle + if 0 < retryAfter <= _MAX_SHORT_RETRY: + logger.info(f"Rate limit on {model.name}, waiting {retryAfter:.1f}s before retry") + await asyncio.sleep(retryAfter + 0.5) + try: + if request.messages: + response = await self._callWithMessages(model, request.messages, options, request.tools) + else: + response = await self._callWithModel(model, prompt, context, options) + logger.info(f"AI call successful with {model.name} after rate-limit retry") + return response + except Exception as retryErr: + lastError = retryErr + logger.warning(f"Retry after rate-limit wait also failed for {model.name}: {retryErr}") + else: + logger.warning(f"Rate limit on {model.name} (retryAfter={retryAfter:.1f}s), failing over") + cooldown = max(retryAfter, 10.0) if retryAfter > 0 else 0.0 + modelSelector.reportFailure(model.name, cooldownSeconds=cooldown) + if attempt < len(failoverModelList) - 1: + continue + logger.error(f"All {len(failoverModelList)} models failed for operation {options.operationType}") + break + except Exception as e: lastError = e logger.warning(f"AI call failed with model {model.name}: {str(e)}") @@ -323,6 +353,13 @@ class AiObjects: filtered = [m for m in availableModels if m.connectorType in allowedProviders] if filtered: availableModels = filtered + else: + yield AiCallResponse( + content=f"No models match allowedProviders {allowedProviders} for operation {options.operationType}", + modelName="error", priceCHF=0.0, processingTime=0.0, + bytesSent=0, bytesReceived=0, errorCount=1, + ) + return failoverModelList = modelSelector.getFailoverModelList( request.prompt, request.context or "", options, availableModels @@ -335,6 +372,8 @@ class AiObjects: ) return + _MAX_SHORT_RETRY = 15.0 + lastError = None for attempt, model in enumerate(failoverModelList): try: @@ -342,6 +381,28 @@ class AiObjects: async for chunk in self._callWithMessagesStream(model, request.messages, options, request.tools): yield chunk return + + except RateLimitExceededException as rle: + retryAfter = rle.retryAfterSeconds + lastError = rle + if 0 < retryAfter <= _MAX_SHORT_RETRY: + logger.info(f"Rate limit on {model.name}, waiting {retryAfter:.1f}s before retry") + await asyncio.sleep(retryAfter + 0.5) + try: + async for chunk in self._callWithMessagesStream(model, request.messages, options, request.tools): + yield chunk + return + except Exception as retryErr: + lastError = retryErr + logger.warning(f"Retry after rate-limit wait also failed for {model.name}: {retryErr}") + else: + logger.warning(f"Rate limit on {model.name} (retryAfter={retryAfter:.1f}s), failing over") + cooldown = max(retryAfter, 10.0) if retryAfter > 0 else 0.0 + modelSelector.reportFailure(model.name, cooldownSeconds=cooldown) + if attempt < len(failoverModelList) - 1: + continue + break + except Exception as e: lastError = e logger.warning(f"Streaming AI call failed with {model.name}: {e}") @@ -421,12 +482,18 @@ class AiObjects: async def callEmbedding(self, texts: List[str], options: AiCallOptions = None) -> AiCallResponse: """Generate embeddings for a list of texts using the best available embedding model. - Uses the standard model selector with OperationTypeEnum.EMBEDDING to pick the model. - Failover across providers (OpenAI → Mistral) works identically to chat models. + Token-aware batching: splits the texts list into batches that respect the + model's contextLength (with 10% safety margin). Each batch is sent as a + separate API call; the resulting embeddings are merged in order. + + Failover across providers (OpenAI -> Mistral) works identically to chat models, + but ContextLengthExceededException is NOT retried via failover (same limits). Returns: AiCallResponse with metadata["embeddings"] containing the vectors. """ + from modules.aicore.aicoreBase import ContextLengthExceededException as _CtxExc + if options is None: options = AiCallOptions(operationType=OperationTypeEnum.EMBEDDING) else: @@ -434,6 +501,15 @@ class AiObjects: combinedText = " ".join(texts[:3])[:500] availableModels = modelRegistry.getAvailableModels() + + allowedProviders = getattr(options, 'allowedProviders', None) if options else None + if allowedProviders: + filtered = [m for m in availableModels if m.connectorType in allowedProviders] + if filtered: + availableModels = filtered + else: + logger.warning(f"No embedding models match allowedProviders {allowedProviders}") + failoverModelList = modelSelector.getFailoverModelList( combinedText, "", options, availableModels ) @@ -451,23 +527,39 @@ class AiObjects: inputBytes = sum(len(t.encode("utf-8")) for t in texts) startTime = time.time() - modelCall = AiModelCall( - model=model, options=options, embeddingInput=texts + batches = _buildEmbeddingBatches(texts, model.contextLength) + logger.info( + f"Embedding: {len(texts)} texts -> {len(batches)} batch(es), " + f"model contextLength={model.contextLength}" ) - modelResponse = await model.functionCall(modelCall) - if not modelResponse.success: - raise ValueError(f"Embedding call failed: {modelResponse.error}") + allEmbeddings: List[List[float]] = [] + totalPriceCHF = 0.0 + + for batchIdx, batch in enumerate(batches): + modelCall = AiModelCall( + model=model, options=options, embeddingInput=batch + ) + modelResponse = await model.functionCall(modelCall) + + if not modelResponse.success: + raise ValueError(f"Embedding batch {batchIdx + 1} failed: {modelResponse.error}") + + batchEmbeddings = (modelResponse.metadata or {}).get("embeddings", []) + allEmbeddings.extend(batchEmbeddings) + + batchBytes = sum(len(t.encode("utf-8")) for t in batch) + totalPriceCHF += model.calculatepriceCHF(0, batchBytes, 0) processingTime = time.time() - startTime - priceCHF = model.calculatepriceCHF(processingTime, inputBytes, 0) - embeddings = (modelResponse.metadata or {}).get("embeddings", []) + if totalPriceCHF == 0.0: + totalPriceCHF = model.calculatepriceCHF(processingTime, inputBytes, 0) response = AiCallResponse( content="", modelName=model.name, provider=model.connectorType, - priceCHF=priceCHF, processingTime=processingTime, + priceCHF=totalPriceCHF, processingTime=processingTime, bytesSent=inputBytes, bytesReceived=0, errorCount=0, - metadata={"embeddings": embeddings} + metadata={"embeddings": allEmbeddings} ) if self.billingCallback: @@ -478,6 +570,23 @@ class AiObjects: return response + except _CtxExc as e: + logger.error(f"ContextLengthExceeded for {model.name} despite batching – aborting failover: {e}") + return AiCallResponse( + content=str(e), modelName=model.name, priceCHF=0.0, + processingTime=0.0, bytesSent=0, bytesReceived=0, errorCount=1 + ) + + except RateLimitExceededException as rle: + retryAfter = rle.retryAfterSeconds + lastError = rle + cooldown = max(retryAfter, 10.0) if retryAfter > 0 else 0.0 + logger.warning(f"Rate limit on {model.name} during embedding (retryAfter={retryAfter:.1f}s)") + modelSelector.reportFailure(model.name, cooldownSeconds=cooldown) + if attempt < len(failoverModelList) - 1: + continue + break + except Exception as e: lastError = e logger.warning(f"Embedding call failed with {model.name}: {str(e)}") @@ -514,4 +623,50 @@ class AiObjects: return [model.displayName for model in models] +# ============================================================================= +# Internal helpers +# ============================================================================= + +_CHARS_PER_TOKEN = 4 +_SAFETY_MARGIN = 0.90 + + +def _estimateTokens(text: str) -> int: + """Rough token estimate: 1 token ~ 4 characters.""" + return max(1, len(text) // _CHARS_PER_TOKEN) + + +def _buildEmbeddingBatches(texts: List[str], contextLength: int) -> List[List[str]]: + """Split a list of texts into batches whose total estimated token count + stays within the model's contextLength (with safety margin). + + Each individual text is assumed to already be within limits (enforced by + the chunking layer). If a single text exceeds the budget, it is placed + in its own batch as a last resort. + """ + if not texts: + return [] + if contextLength <= 0: + return [texts] + + maxTokensPerBatch = int(contextLength * _SAFETY_MARGIN) + batches: List[List[str]] = [] + currentBatch: List[str] = [] + currentTokens = 0 + + for text in texts: + textTokens = _estimateTokens(text) + if currentBatch and (currentTokens + textTokens) > maxTokensPerBatch: + batches.append(currentBatch) + currentBatch = [] + currentTokens = 0 + currentBatch.append(text) + currentTokens += textTokens + + if currentBatch: + batches.append(currentBatch) + + return batches + + diff --git a/modules/interfaces/interfaceBootstrap.py b/modules/interfaces/interfaceBootstrap.py index 609dd61b..63978165 100644 --- a/modules/interfaces/interfaceBootstrap.py +++ b/modules/interfaces/interfaceBootstrap.py @@ -96,6 +96,9 @@ def initBootstrap(db: DatabaseConnector) -> None: if mandateId: initRootMandateFeatures(db, mandateId) + # Remove feature instances for features that no longer exist in the codebase + _cleanupRemovedFeatureInstances(db) + # Initialize billing settings for root mandate if mandateId: initRootMandateBilling(mandateId) @@ -257,6 +260,33 @@ def initRootMandateFeatures(db: DatabaseConnector, mandateId: str) -> None: logger.info("Root mandate features initialization completed") +def _cleanupRemovedFeatureInstances(db: DatabaseConnector) -> None: + """Remove feature instances whose featureCode no longer exists in the codebase.""" + from modules.datamodels.datamodelFeatures import FeatureInstance + from modules.system.registry import loadFeatureMainModules + + mainModules = loadFeatureMainModules() + activeCodes = set() + for featureName, module in mainModules.items(): + if hasattr(module, "getFeatureDefinition"): + try: + featureDef = module.getFeatureDefinition() + activeCodes.add(featureDef.get("code", featureName)) + except Exception: + pass + + allInstances = db.getRecordset(FeatureInstance) + for inst in allInstances: + code = inst.get("featureCode") if isinstance(inst, dict) else getattr(inst, "featureCode", None) + instId = inst.get("id") if isinstance(inst, dict) else getattr(inst, "id", None) + if code and code not in activeCodes: + try: + db.recordDelete(FeatureInstance, str(instId)) + logger.info(f"Removed orphaned feature instance '{instId}' (featureCode='{code}')") + except Exception as e: + logger.warning(f"Could not remove orphaned feature instance '{instId}': {e}") + + def initRootMandate(db: DatabaseConnector) -> Optional[str]: """ Creates the Root mandate if it doesn't exist. @@ -443,7 +473,7 @@ def initRoles(db: DatabaseConnector) -> None: # Check specifically for system template roles: # mandateId=NULL, isSystemRole=True, featureCode=NULL - # Feature templates (e.g. chatplayground admin) share the same labels but have featureCode set! + # Feature templates (e.g. automation admin) share the same labels but have featureCode set! allTemplates = db.getRecordset( Role, recordFilter={"mandateId": None, "isSystemRole": True} @@ -475,7 +505,7 @@ def _deduplicateRoles(db: DatabaseConnector) -> None: # Group by (roleLabel, mandateId, featureInstanceId, featureCode) # featureCode is essential: system template ('admin', None, None, None) - # must NOT be grouped with feature template ('admin', None, None, 'chatplayground') + # must NOT be grouped with feature template ('admin', None, None, 'automation') groups: dict = {} for role in allRoles: key = (role.get("roleLabel"), role.get("mandateId"), role.get("featureInstanceId"), role.get("featureCode")) @@ -1931,8 +1961,6 @@ def _createStoreResourceRules(db: DatabaseConnector) -> None: """ storeResources = [ "resource.store.automation", - "resource.store.chatplayground", - "resource.store.codeeditor", "resource.store.teamsbot", ] diff --git a/modules/interfaces/interfaceDbManagement.py b/modules/interfaces/interfaceDbManagement.py index f439a7f1..4ddcb972 100644 --- a/modules/interfaces/interfaceDbManagement.py +++ b/modules/interfaces/interfaceDbManagement.py @@ -853,7 +853,9 @@ class ComponentObjects: "svg": "image/svg+xml", "py": "text/x-python", "js": "application/javascript", - "css": "text/css" + "css": "text/css", + "eml": "message/rfc822", + "msg": "application/vnd.ms-outlook", } return extensionToMime.get(ext.lower(), "application/octet-stream") diff --git a/modules/routes/routeSystem.py b/modules/routes/routeSystem.py index 95d90aa6..c2be23a6 100644 --- a/modules/routes/routeSystem.py +++ b/modules/routes/routeSystem.py @@ -102,12 +102,6 @@ def _getFeatureUiObjects(featureCode: str) -> List[Dict[str, Any]]: elif featureCode == "realestate": from modules.features.realEstate.mainRealEstate import UI_OBJECTS return UI_OBJECTS - elif featureCode == "chatplayground": - from modules.features.chatplayground.mainChatplayground import UI_OBJECTS - return UI_OBJECTS - elif featureCode == "codeeditor": - from modules.features.codeeditor.mainCodeeditor import UI_OBJECTS - return UI_OBJECTS elif featureCode == "automation": from modules.features.automation.mainAutomation import UI_OBJECTS return UI_OBJECTS @@ -127,7 +121,7 @@ def _getFeatureUiObjects(featureCode: str) -> List[Dict[str, Any]]: from modules.features.workspace.mainWorkspace import UI_OBJECTS return UI_OBJECTS else: - logger.warning(f"Unknown feature code: {featureCode}") + logger.debug(f"Skipping removed feature code: {featureCode}") return [] except ImportError as e: logger.error(f"Failed to import UI_OBJECTS for feature {featureCode}: {e}") diff --git a/modules/serviceCenter/services/serviceAgent/actionToolAdapter.py b/modules/serviceCenter/services/serviceAgent/actionToolAdapter.py index c80ffdeb..17e953fd 100644 --- a/modules/serviceCenter/services/serviceAgent/actionToolAdapter.py +++ b/modules/serviceCenter/services/serviceAgent/actionToolAdapter.py @@ -41,7 +41,7 @@ class ActionToolAdapter: if not actionDef or not getattr(actionDef, "dynamicMode", False): continue - compoundName = f"{shortName}.{actionName}" + compoundName = f"{shortName}_{actionName}" toolDef = _buildToolDefinition(compoundName, actionDef, actionInfo) handler = _createDispatchHandler(self._actionExecutor, shortName, actionName) @@ -120,16 +120,16 @@ def _createDispatchHandler(actionExecutor, methodName: str, actionName: str): data = _formatActionResult(result) return ToolResult( toolCallId="", - toolName=f"{methodName}.{actionName}", + toolName=f"{methodName}_{actionName}", success=result.success, data=data, error=result.error ) except Exception as e: - logger.error(f"ActionToolAdapter dispatch failed for {methodName}.{actionName}: {e}") + logger.error(f"ActionToolAdapter dispatch failed for {methodName}_{actionName}: {e}") return ToolResult( toolCallId="", - toolName=f"{methodName}.{actionName}", + toolName=f"{methodName}_{actionName}", success=False, error=str(e) ) diff --git a/modules/serviceCenter/services/serviceAgent/datamodelAgent.py b/modules/serviceCenter/services/serviceAgent/datamodelAgent.py index 93a8e966..c70d8344 100644 --- a/modules/serviceCenter/services/serviceAgent/datamodelAgent.py +++ b/modules/serviceCenter/services/serviceAgent/datamodelAgent.py @@ -26,6 +26,10 @@ class AgentEventTypeEnum(str, Enum): AGENT_PROGRESS = "agentProgress" AGENT_SUMMARY = "agentSummary" FILE_CREATED = "fileCreated" + FILE_UPDATED = "fileUpdated" + FILE_EDIT_PROPOSAL = "fileEditProposal" + FILE_VERSION = "fileVersion" + FILE_EDIT_REJECTED = "fileEditRejected" DATA_SOURCE_ACCESS = "dataSourceAccess" VOICE_RESPONSE = "voiceResponse" FINAL = "final" @@ -50,7 +54,7 @@ class ToolDefinition(BaseModel): ) toolSet: Optional[str] = Field( default=None, - description="Tool-set scope (None = available to all sets, e.g. 'core', 'codeeditor')" + description="Tool-set scope (None = available to all sets, e.g. 'core', 'workspace')" ) @@ -133,3 +137,16 @@ class AgentTrace(BaseModel): totalCostCHF: float = 0.0 abortReason: Optional[str] = None rounds: List[AgentRoundLog] = Field(default_factory=list) + + +class PendingFileEdit(BaseModel): + """A proposed file edit awaiting user approval.""" + id: str = Field(default_factory=lambda: str(uuid.uuid4())) + fileId: str + fileName: str + mimeType: str = "" + oldContent: str = "" + newContent: str = "" + status: str = Field(default="pending", description="pending | accepted | rejected") + toolCallId: str = "" + workflowId: str = "" diff --git a/modules/serviceCenter/services/serviceAgent/mainServiceAgent.py b/modules/serviceCenter/services/serviceAgent/mainServiceAgent.py index 79781804..c642a2fa 100644 --- a/modules/serviceCenter/services/serviceAgent/mainServiceAgent.py +++ b/modules/serviceCenter/services/serviceAgent/mainServiceAgent.py @@ -160,7 +160,8 @@ class AgentService: ): if event.type == AgentEventTypeEnum.AGENT_SUMMARY: await self._persistTrace(workflowId, event.data or {}) - logger.debug(f"runAgent yielding event type={event.type}") + if event.type != AgentEventTypeEnum.CHUNK: + logger.debug(f"runAgent yielding event type={event.type}") yield event logger.info(f"runAgent loop completed for workflow {workflowId}") @@ -366,6 +367,7 @@ def _getOrCreateTempFolder(chatService) -> Optional[str]: def _registerCoreTools(registry: ToolRegistry, services): """Register built-in core tools: file operations, search, and folder management.""" + import uuid as _uuid from modules.serviceCenter.services.serviceAgent.datamodelAgent import ToolResult # ---- Read-only tools ---- @@ -411,7 +413,8 @@ def _registerCoreTools(registry: ToolRegistry, services): _BINARY_TYPES = ("application/pdf", "image/", "application/vnd.", "application/zip", "application/x-zip", "application/x-tar", "application/x-7z", - "application/msword", "application/octet-stream") + "application/msword", "application/octet-stream", + "message/rfc822") isBinary = any(mimeType.startswith(t) for t in _BINARY_TYPES) rawBytes = chatService.getFileData(fileId) @@ -1013,16 +1016,28 @@ def _registerCoreTools(registry: ToolRegistry, services): toolCallId="", toolName="editFile", success=False, error=f"Cannot edit binary file ({file.mimeType}). Only text-based files are supported." ) - contentBytes = content.encode("utf-8") - success = dbMgmt.updateFileData(fileId, contentBytes) - if not success: - return ToolResult(toolCallId="", toolName="editFile", success=False, error="Failed to update file data") + oldContent = "" + oldData = dbMgmt.getFileData(fileId) + if oldData: + try: + oldContent = oldData.decode("utf-8") + except UnicodeDecodeError: + oldContent = "" + + editId = str(_uuid.uuid4()) return ToolResult( toolCallId="", toolName="editFile", success=True, - data=f"File '{file.fileName}' updated ({len(contentBytes)} bytes)", + data=f"Edit proposed for '{file.fileName}'. Waiting for user review.", sideEvents=[{ - "type": "fileUpdated", - "data": {"fileId": fileId, "fileName": file.fileName, "fileSize": len(contentBytes)}, + "type": "fileEditProposal", + "data": { + "id": editId, + "fileId": fileId, + "fileName": file.fileName, + "mimeType": file.mimeType, + "oldContent": oldContent, + "newContent": content, + }, }], ) except Exception as e: @@ -1088,8 +1103,9 @@ def _registerCoreTools(registry: ToolRegistry, services): registry.register( "editFile", _editFile, description=( - "Update the content of an existing text file. Only works for text-based files " - "(text/*, application/json, etc.). For binary files, create a new file instead." + "Propose an edit to an existing text file. The change is shown to the user " + "for review (accept/reject) before being applied. Only works for text-based " + "files (text/*, application/json, etc.). For binary files, create a new file instead." ), parameters={ "type": "object", @@ -1164,15 +1180,24 @@ def _registerCoreTools(registry: ToolRegistry, services): return ToolResult(toolCallId="", toolName="externalDownload", success=False, error="connectionId, service, and path are required") try: from modules.connectors.connectorResolver import ConnectorResolver + from modules.connectors.connectorProviderBase import DownloadResult as _DR resolver = ConnectorResolver( services.getService("security"), _buildResolverDb(), ) adapter = await resolver.resolveService(connectionId, service) - fileBytes = await adapter.download(path) + result = await adapter.download(path) + + if isinstance(result, _DR): + fileBytes = result.data + fileName = result.fileName or path.split("/")[-1] or "downloaded_file" + else: + fileBytes = result + fileName = path.split("/")[-1] or "downloaded_file" + if not fileBytes: return ToolResult(toolCallId="", toolName="externalDownload", success=False, error="Download returned empty") - fileName = path.split("/")[-1] or "downloaded_file" + chatService = services.chat fileItem, _ = chatService.interfaceDbComponent.saveUploadedFile(fileBytes, fileName) fid = fileItem.id if hasattr(fileItem, "id") else fileItem.get("id", "?") @@ -1183,7 +1208,7 @@ def _registerCoreTools(registry: ToolRegistry, services): if tempFolderId: chatService.interfaceDbComponent.updateFile(fid, {"folderId": tempFolderId}) ext = fileName.rsplit(".", 1)[-1].lower() if "." in fileName else "" - hint = "Use readFile to read text content." if ext in ("doc", "docx", "txt", "csv", "json", "xml", "html", "md", "rtf", "odt", "xls", "xlsx", "pptx") else "Use readFile to access the content." + hint = "Use readFile to read text content." if ext in ("doc", "docx", "txt", "csv", "json", "xml", "html", "md", "rtf", "odt", "xls", "xlsx", "pptx", "eml", "msg") else "Use readFile to access the content." return ToolResult( toolCallId="", toolName="externalDownload", success=True, data=f"Downloaded '{fileName}' ({len(fileBytes)} bytes) → local file id: {fid}. {hint}" @@ -1375,6 +1400,8 @@ def _registerCoreTools(registry: ToolRegistry, services): logger.info(f"Resolved DataSource '{dsId}' ({label}): sourceType={sourceType}, service={service}, connectionId={connectionId}, path={path[:80]}") return connectionId, service, path + _MAIL_SERVICES = {"outlook", "gmail"} + async def _browseDataSource(args: Dict[str, Any], context: Dict[str, Any]): dsId = args.get("dataSourceId", "") subPath = args.get("subPath", "") @@ -1403,7 +1430,10 @@ def _registerCoreTools(registry: ToolRegistry, services): prefix = "[DIR]" if e.isFolder else "[FILE]" sizeInfo = f" ({e.size} bytes)" if e.size else "" lines.append(f"- {prefix} {e.name}{sizeInfo} path: {e.path}") - return ToolResult(toolCallId="", toolName="browseDataSource", success=True, data="\n".join(lines)) + result = "\n".join(lines) + if service in _MAIL_SERVICES: + result += "\n\nIMPORTANT: These are email subjects only. To read the full email content, use downloadFromDataSource with the path, then readFile on the returned file ID." + return ToolResult(toolCallId="", toolName="browseDataSource", success=True, data=result) except Exception as e: return ToolResult(toolCallId="", toolName="browseDataSource", success=False, error=str(e)) @@ -1424,7 +1454,10 @@ def _registerCoreTools(registry: ToolRegistry, services): if not entries: return ToolResult(toolCallId="", toolName="searchDataSource", success=True, data="No results found.") lines = [f"- {e.name} (path: {e.path})" for e in entries] - return ToolResult(toolCallId="", toolName="searchDataSource", success=True, data="\n".join(lines)) + result = "\n".join(lines) + if service in _MAIL_SERVICES: + result += "\n\nIMPORTANT: These are email subjects only. To read the full email content, use downloadFromDataSource with the path, then readFile on the returned file ID." + return ToolResult(toolCallId="", toolName="searchDataSource", success=True, data=result) except Exception as e: return ToolResult(toolCallId="", toolName="searchDataSource", success=False, error=str(e)) @@ -1435,17 +1468,26 @@ def _registerCoreTools(registry: ToolRegistry, services): if not dsId or not filePath: return ToolResult(toolCallId="", toolName="downloadFromDataSource", success=False, error="dataSourceId and filePath are required") try: + from modules.connectors.connectorResolver import ConnectorResolver + from modules.connectors.connectorProviderBase import DownloadResult as _DR connectionId, service, basePath = await _resolveDataSource(dsId) fullPath = filePath if filePath.startswith("/") else f"{basePath.rstrip('/')}/{filePath}" - from modules.connectors.connectorResolver import ConnectorResolver resolver = ConnectorResolver( services.getService("security"), _buildResolverDb(), ) adapter = await resolver.resolveService(connectionId, service) - fileBytes = await adapter.download(fullPath) + result = await adapter.download(fullPath) + + if isinstance(result, _DR): + fileBytes = result.data + fileName = result.fileName or fileName + else: + fileBytes = result + if not fileBytes: return ToolResult(toolCallId="", toolName="downloadFromDataSource", success=False, error="Download returned empty") + if not fileName or "." not in fileName: pathSegment = fullPath.split("/")[-1] or "downloaded_file" fileName = fileName or pathSegment @@ -1460,16 +1502,20 @@ def _registerCoreTools(registry: ToolRegistry, services): except Exception: pass if "." not in fileName: - import mimetypes as _mt - guessed = _mt.guess_type(f"file.{_mt.guess_extension('application/octet-stream') or ''}")[0] - if not guessed and fileBytes[:4] == b"%PDF": + if fileBytes[:4] == b"%PDF": fileName = f"{fileName}.pdf" - elif not guessed and fileBytes[:2] == b"PK": + elif fileBytes[:2] == b"PK": fileName = f"{fileName}.zip" chatService = services.chat fileItem, _ = chatService.interfaceDbComponent.saveUploadedFile(fileBytes, fileName) + fiId = context.get("featureInstanceId") or (services.featureInstanceId if services else "") + if fiId: + chatService.interfaceDbComponent.updateFile(fileItem.id, {"featureInstanceId": fiId}) + tempFolderId = _getOrCreateTempFolder(chatService) + if tempFolderId: + chatService.interfaceDbComponent.updateFile(fileItem.id, {"folderId": tempFolderId}) ext = fileName.rsplit(".", 1)[-1].lower() if "." in fileName else "" - hint = "Use readFile to read the text content." if ext in ("doc", "docx", "txt", "csv", "json", "xml", "html", "md", "rtf", "odt", "xls", "xlsx", "pptx", "pdf") else "Use readFile to access the content." + hint = "Use readFile to read the text content." if ext in ("doc", "docx", "txt", "csv", "json", "xml", "html", "md", "rtf", "odt", "xls", "xlsx", "pptx", "pdf", "eml", "msg") else "Use readFile to access the content." return ToolResult( toolCallId="", toolName="downloadFromDataSource", success=True, data=f"Downloaded '{fileName}' ({len(fileBytes)} bytes) → local file id: {fileItem.id}. {hint}" @@ -1508,7 +1554,7 @@ def _registerCoreTools(registry: ToolRegistry, services): registry.register( "downloadFromDataSource", _downloadFromDataSource, - description="Download a file from an attached data source into local storage. Returns the local file ID which can then be read with readFile. Always provide the fileName if known from the browse results.", + description="Download a file or email message from an attached data source into local storage. Returns the local file ID which can then be read with readFile. For email sources (Outlook, Gmail), this downloads the full email content -- browse/search only return subjects. Always provide the fileName if known.", parameters={ "type": "object", "properties": { diff --git a/modules/serviceCenter/services/serviceAi/mainServiceAi.py b/modules/serviceCenter/services/serviceAi/mainServiceAi.py index a24af9a9..17ff2da2 100644 --- a/modules/serviceCenter/services/serviceAi/mainServiceAi.py +++ b/modules/serviceCenter/services/serviceAi/mainServiceAi.py @@ -7,7 +7,7 @@ import time import base64 from typing import Dict, Any, List, Optional, Tuple, Callable from modules.datamodels.datamodelChat import PromptPlaceholder, ChatDocument, WorkflowModeEnum -from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationTypeEnum, PriorityEnum, ProcessingModeEnum +from modules.datamodels.datamodelAi import AiCallRequest, AiCallResponse, AiCallOptions, OperationTypeEnum, PriorityEnum, ProcessingModeEnum from modules.datamodels.datamodelExtraction import ContentPart, DocumentIntent from modules.datamodels.datamodelWorkflow import AiResponse, AiResponseMetadata, DocumentData from modules.datamodels.datamodelDocument import RenderedDocument @@ -198,6 +198,19 @@ class AiService: finally: self.aiObjects.billingCallback = None + async def callEmbedding(self, texts: List[str]) -> AiCallResponse: + """Generate embeddings while respecting allowedProviders.""" + await self.ensureAiObjectsInitialized() + options = AiCallOptions(operationType=OperationTypeEnum.EMBEDDING) + effectiveProviders = self._calculateEffectiveProviders() + if effectiveProviders: + options.allowedProviders = effectiveProviders + self.aiObjects.billingCallback = self._createBillingCallback() + try: + return await self.aiObjects.callEmbedding(texts, options) + finally: + self.aiObjects.billingCallback = None + # ========================================================================= # SPEECH_TEAMS: Dedicated handler for Teams Meeting AI analysis # Bypasses standard model selection. Uses a fixed fast model. diff --git a/modules/serviceCenter/services/serviceBilling/mainServiceBilling.py b/modules/serviceCenter/services/serviceBilling/mainServiceBilling.py index 7a6951ae..ac475251 100644 --- a/modules/serviceCenter/services/serviceBilling/mainServiceBilling.py +++ b/modules/serviceCenter/services/serviceBilling/mainServiceBilling.py @@ -42,7 +42,7 @@ def getService(currentUser: User, mandateId: str, featureInstanceId: str = None, currentUser: Current user object mandateId: Mandate ID for context featureInstanceId: Optional feature instance ID - featureCode: Optional feature code (e.g., 'chatplayground', 'automation') + featureCode: Optional feature code (e.g., 'automation') Returns: BillingService instance diff --git a/modules/serviceCenter/services/serviceKnowledge/mainServiceKnowledge.py b/modules/serviceCenter/services/serviceKnowledge/mainServiceKnowledge.py index a2dabadf..91e85da4 100644 --- a/modules/serviceCenter/services/serviceKnowledge/mainServiceKnowledge.py +++ b/modules/serviceCenter/services/serviceKnowledge/mainServiceKnowledge.py @@ -3,6 +3,7 @@ """Knowledge service: 3-tier RAG with indexing, semantic search, and context building.""" import logging +import re from typing import Any, Callable, Dict, List, Optional from modules.datamodels.datamodelKnowledge import ( @@ -14,7 +15,8 @@ from modules.shared.timeUtils import getUtcTimestamp logger = logging.getLogger(__name__) -DEFAULT_CHUNK_SIZE = 512 +CHARS_PER_TOKEN = 4 +DEFAULT_CHUNK_TOKENS = 400 DEFAULT_CONTEXT_BUDGET = 8000 @@ -31,14 +33,9 @@ class KnowledgeService: # ========================================================================= async def _embed(self, texts: List[str]) -> List[List[float]]: - """Embed texts via the AI interface's generic embedding method.""" + """Embed texts via AiService (respects allowedProviders).""" aiService = self._getService("ai") - await aiService.ensureAiObjectsInitialized() - aiObjects = aiService.aiObjects - if aiObjects is None: - logger.warning("Embedding skipped: aiObjects not available") - return [] - response = await aiObjects.callEmbedding(texts) + response = await aiService.callEmbedding(texts) if response.errorCount > 0: logger.error(f"Embedding failed: {response.content}") return [] @@ -115,9 +112,16 @@ class KnowledgeService: textObjects = [o for o in contentObjects if o.get("contentType") == "text"] if textObjects: self._knowledgeDb.updateFileStatus(fileId, "embedding") - chunks = _chunkForEmbedding(textObjects, chunkSize=DEFAULT_CHUNK_SIZE) + chunks = _chunkForEmbedding(textObjects, maxTokens=DEFAULT_CHUNK_TOKENS) texts = [c["data"] for c in chunks] + totalChars = sum(len(t) for t in texts) + estTokens = totalChars // CHARS_PER_TOKEN + logger.info( + f"Embedding file {fileId}: {len(textObjects)} text objects -> " + f"{len(chunks)} chunks, ~{estTokens} tokens total" + ) + embeddings = await self._embed(texts) if texts else [] for i, chunk in enumerate(chunks): @@ -428,49 +432,77 @@ class KnowledgeService: # Internal helpers # ============================================================================= +def _estimateTokens(text: str) -> int: + """Estimate token count using character-based heuristic (1 token ~ 4 chars).""" + return max(1, len(text) // CHARS_PER_TOKEN) + + +def _splitSentences(text: str) -> List[str]: + """Split text into sentences at common boundaries (.!?) followed by whitespace.""" + parts = re.split(r'(?<=[.!?])\s+', text.replace("\n", " ").strip()) + return [p for p in parts if p.strip()] + + +def _hardSplitByTokens(text: str, maxTokens: int) -> List[str]: + """Force-split text into pieces that each fit within maxTokens. + + Used as safety net when sentence splitting produces oversized segments. + Splits at word boundaries where possible. + """ + maxChars = maxTokens * CHARS_PER_TOKEN + pieces = [] + while len(text) > maxChars: + splitAt = text.rfind(" ", 0, maxChars) + if splitAt <= 0: + splitAt = maxChars + pieces.append(text[:splitAt].strip()) + text = text[splitAt:].strip() + if text: + pieces.append(text) + return pieces + + def _chunkForEmbedding( - textObjects: List[Dict[str, Any]], chunkSize: int = 512 + textObjects: List[Dict[str, Any]], maxTokens: int = DEFAULT_CHUNK_TOKENS ) -> List[Dict[str, Any]]: - """Split text content objects into chunks suitable for embedding. + """Split text content objects into token-aware chunks suitable for embedding. Each chunk preserves the contextRef from its source object. - Long texts are split at sentence boundaries where possible. + Splits at sentence boundaries; applies hard-cap if a single sentence exceeds maxTokens. """ chunks = [] for obj in textObjects: - text = obj.get("data", "") + text = (obj.get("data", "") or "").strip() + if not text: + continue contentObjectId = obj.get("contentObjectId", "") contextRef = obj.get("contextRef", {}) - if len(text) <= chunkSize: - chunks.append({ - "data": text, - "contentObjectId": contentObjectId, - "contextRef": contextRef, - }) + if _estimateTokens(text) <= maxTokens: + chunks.append({"data": text, "contentObjectId": contentObjectId, "contextRef": contextRef}) continue - # Split at sentence boundaries - sentences = text.replace("\n", " ").split(". ") + sentences = _splitSentences(text) currentChunk = "" for sentence in sentences: - candidate = f"{currentChunk}. {sentence}" if currentChunk else sentence - if len(candidate) > chunkSize and currentChunk: - chunks.append({ - "data": currentChunk.strip(), - "contentObjectId": contentObjectId, - "contextRef": contextRef, - }) + if _estimateTokens(sentence) > maxTokens: + if currentChunk.strip(): + chunks.append({"data": currentChunk.strip(), "contentObjectId": contentObjectId, "contextRef": contextRef}) + currentChunk = "" + for piece in _hardSplitByTokens(sentence, maxTokens): + chunks.append({"data": piece, "contentObjectId": contentObjectId, "contextRef": contextRef}) + continue + + candidate = f"{currentChunk} {sentence}".strip() if currentChunk else sentence + if _estimateTokens(candidate) > maxTokens: + if currentChunk.strip(): + chunks.append({"data": currentChunk.strip(), "contentObjectId": contentObjectId, "contextRef": contextRef}) currentChunk = sentence else: currentChunk = candidate if currentChunk.strip(): - chunks.append({ - "data": currentChunk.strip(), - "contentObjectId": contentObjectId, - "contextRef": contextRef, - }) + chunks.append({"data": currentChunk.strip(), "contentObjectId": contentObjectId, "contextRef": contextRef}) return chunks diff --git a/modules/system/mainSystem.py b/modules/system/mainSystem.py index 3268ffa3..35f06ed1 100644 --- a/modules/system/mainSystem.py +++ b/modules/system/mainSystem.py @@ -437,16 +437,6 @@ RESOURCE_OBJECTS = [ "label": {"en": "Store: Automation", "de": "Store: Automation", "fr": "Store: Automatisation"}, "meta": {"category": "store", "featureCode": "automation"} }, - { - "objectKey": "resource.store.chatplayground", - "label": {"en": "Store: Chat Playground", "de": "Store: Chat Playground", "fr": "Store: Chat Playground"}, - "meta": {"category": "store", "featureCode": "chatplayground"} - }, - { - "objectKey": "resource.store.codeeditor", - "label": {"en": "Store: Code Editor", "de": "Store: Code Editor", "fr": "Store: Code Editor"}, - "meta": {"category": "store", "featureCode": "codeeditor"} - }, { "objectKey": "resource.store.teamsbot", "label": {"en": "Store: Teams Bot", "de": "Store: Teams Bot", "fr": "Store: Teams Bot"}, diff --git a/modules/workflows/automation/__init__.py b/modules/workflows/automation/__init__.py index ce459ef8..89022240 100644 --- a/modules/workflows/automation/__init__.py +++ b/modules/workflows/automation/__init__.py @@ -1,11 +1,7 @@ # Copyright (c) 2025 Patrick Motsch # All rights reserved. """ -Workflow feature - handles workflow execution, scheduling, and chat playground operations. - -Combines functionality from: -- automation: Automation workflow execution and scheduling -- chatPlayground: Chat playground workflow start/stop operations +Workflow feature - handles workflow execution, scheduling, and start/stop operations. """ from .mainWorkflow import chatStart, chatStop, executeAutomation, syncAutomationEvents, createAutomationEventHandler diff --git a/modules/workflows/automation/mainWorkflow.py b/modules/workflows/automation/mainWorkflow.py index 625384d7..19473c01 100644 --- a/modules/workflows/automation/mainWorkflow.py +++ b/modules/workflows/automation/mainWorkflow.py @@ -1,11 +1,7 @@ # Copyright (c) 2025 Patrick Motsch # All rights reserved. """ -Main workflow service - handles workflow execution, scheduling, and chat playground operations. - -Combines functionality from: -- mainAutomation.py: Automation workflow execution and scheduling -- mainChatPlayground.py: Chat playground workflow start/stop operations +Main workflow service - handles workflow execution, scheduling, and start/stop operations. """ import logging @@ -35,11 +31,11 @@ async def chatStart(currentUser: User, userInput: UserInputRequest, workflowMode workflowMode: Workflow mode (Dynamic, Automation, etc.) mandateId: Mandate ID (required for billing) featureInstanceId: Feature instance ID (required for billing) - featureCode: Feature code (e.g., 'chatplayground', 'automation') + featureCode: Feature code (e.g., 'automation') services: Pre-built service hub from the calling feature (required). Each feature must pass its own services. """ if services is None: - raise ValueError("services is required: each feature must pass its own service hub (e.g. getChatplaygroundServices, getAutomationServices)") + raise ValueError("services is required: each feature must pass its own service hub (e.g. getAutomationServices)") try: # Store allowedProviders in services context for model selection @@ -61,7 +57,7 @@ async def chatStart(currentUser: User, userInput: UserInputRequest, workflowMode async def chatStop(currentUser: User, workflowId: str, mandateId: Optional[str] = None, featureInstanceId: Optional[str] = None, featureCode: Optional[str] = None, services=None) -> ChatWorkflow: """Stops a running chat. Caller must pass services from the owning feature.""" if services is None: - raise ValueError("services is required: each feature must pass its own service hub (e.g. getChatplaygroundServices, getAutomationServices)") + raise ValueError("services is required: each feature must pass its own service hub (e.g. getAutomationServices)") try: if featureCode: services.featureCode = featureCode diff --git a/scripts/migrate_async_to_sync.py b/scripts/migrate_async_to_sync.py index d2b030a4..4e18b3ea 100644 --- a/scripts/migrate_async_to_sync.py +++ b/scripts/migrate_async_to_sync.py @@ -76,10 +76,6 @@ _MUST_STAY_ASYNC: Dict[str, Set[str]] = { "event_stream", # await request.is_disconnected(), await asyncio.wait_for(...) "stop_chatbot", # await event_manager.emit_event(...) }, - "modules/features/chatplayground/routeFeatureChatplayground.py": { - "start_workflow", # await chatStart(...) - "stop_workflow", # await chatStop(...) - }, "modules/features/neutralization/routeFeatureNeutralizer.py": { "process_sharepoint_files", # await service.processSharepointFiles(...) }, @@ -105,7 +101,6 @@ _SKIP_FILES: Set[str] = { # Helper functions that are fake-async (async def but no await inside) # These will be converted from async def -> def _FAKE_ASYNC_HELPERS: Dict[str, Set[str]] = { - "modules/features/chatplayground/routeFeatureChatplayground.py": {"_validateInstanceAccess"}, "modules/features/trustee/routeFeatureTrustee.py": {"_validateInstanceAccess", "_validateInstanceAdmin"}, "modules/features/realestate/routeFeatureRealEstate.py": {"_validateInstanceAccess"}, "modules/features/chatbot/routeFeatureChatbot.py": {"_validateInstanceAccess"}, diff --git a/tests/functional/test03_ai_operations.py b/tests/functional/test03_ai_operations.py index 169d3ba6..4df53c5d 100644 --- a/tests/functional/test03_ai_operations.py +++ b/tests/functional/test03_ai_operations.py @@ -96,7 +96,7 @@ class MethodAiOperationsTester: import logging logging.getLogger().setLevel(logging.DEBUG) - # Import and initialize services - use the same approach as routeChatPlayground + # Import and initialize services import modules.interfaces.interfaceDbChat as interfaceFeatureAiChat interfaceDbChat = interfaceDbChat.getInterface(self.testUser)