ValueOn Lead to Offer durchgespielt, bugfixes in Dateigenerierung und ai nodes

modules/aicore/aicorePluginAnthropic.py

@@ -351,6 +351,7 @@ class AiAnthropic(BaseConnectorAi):
             
             # Parse response
             anthropicResponse = response.json()
+            stop_reason = anthropicResponse.get("stop_reason")
             
             # Extract content and tool_use blocks from response
             content = ""
@@ -374,9 +375,25 @@ class AiAnthropic(BaseConnectorAi):
             
             if not content and not toolCalls:
                 logger.warning(f"Anthropic API returned empty content. Full response: {anthropicResponse}")
-                content = "[Anthropic API returned empty response]"
+                err = (
+                    "Anthropic refused the request (content policy) — try another model or adjust the prompt."
+                    if stop_reason == "refusal"
+                    else f"Anthropic returned no assistant text (stop_reason={stop_reason or 'unknown'})."
+                )
+                return AiModelResponse(
+                    content="",
+                    success=False,
+                    error=err,
+                    modelId=model.name,
+                    metadata={
+                        "response_id": anthropicResponse.get("id", ""),
+                        "stop_reason": stop_reason,
+                    },
+                )
             
             metadata = {"response_id": anthropicResponse.get("id", "")}
+            if stop_reason:
+                metadata["stop_reason"] = stop_reason
             if toolCalls:
                 metadata["toolCalls"] = toolCalls
             
@@ -492,6 +509,19 @@ class AiAnthropic(BaseConnectorAi):
                     f"Anthropic stream returned empty response: model={model.name}, "
                     f"stopReason={stopReason}"
                 )
+                err = (
+                    "Anthropic refused the request (content policy) — try another model or adjust the prompt."
+                    if stopReason == "refusal"
+                    else f"Anthropic returned no assistant text (stop_reason={stopReason or 'unknown'})."
+                )
+                yield AiModelResponse(
+                    content="",
+                    success=False,
+                    error=err,
+                    modelId=model.name,
+                    metadata={"stopReason": stopReason} if stopReason else {},
+                )
+                return
 
             metadata: Dict[str, Any] = {}
             if stopReason:

modules/connectors/connectorDbPostgre.py

@@ -834,7 +834,10 @@ class DatabaseConnector:
             createdTs = record.get("sysCreatedAt")
             if createdTs is None or createdTs == 0 or createdTs == 0.0:
                 record["sysCreatedAt"] = currentTime
-                if effective_user_id:
+                # Do not wipe caller-provided sysCreatedBy (e.g. FileItem from createFile with
+                # real user). ContextVar can be "system" for the DB pool while the business
+                # user is set on the record from model_dump().
+                if effective_user_id and not record.get("sysCreatedBy"):
                     record["sysCreatedBy"] = effective_user_id
             elif not record.get("sysCreatedBy"):
                 if effective_user_id:
@@ -1531,7 +1534,7 @@ class DatabaseConnector:
             createdTs = rec.get("sysCreatedAt")
             if createdTs is None or createdTs == 0 or createdTs == 0.0:
                 rec["sysCreatedAt"] = currentTime
-                if effectiveUserId:
+                if effectiveUserId and not rec.get("sysCreatedBy"):
                     rec["sysCreatedBy"] = effectiveUserId
             elif not rec.get("sysCreatedBy") and effectiveUserId:
                 rec["sysCreatedBy"] = effectiveUserId

modules/features/graphicalEditor/nodeDefinitions/ai.py

@@ -132,6 +132,14 @@ AI_NODES = [
         "parameters": [
             {"name": "prompt", "type": "str", "required": True, "frontendType": "textarea",
              "description": t("Generierungs-Prompt")},
+            {"name": "outputFormat", "type": "str", "required": False, "frontendType": "select",
+             "frontendOptions": {"options": ["docx", "pdf", "txt", "html", "md"]},
+             "description": t("Ausgabeformat"), "default": "docx"},
+            {"name": "title", "type": "str", "required": False, "frontendType": "text",
+             "description": t("Dokumenttitel (Metadaten / Dateiname)"), "default": ""},
+            {"name": "documentType", "type": "str", "required": False, "frontendType": "select",
+             "frontendOptions": {"options": ["letter", "memo", "proposal", "contract", "report", "email"]},
+             "description": t("Dokumentart (Inhaltshinweis fuer die KI)"), "default": "proposal"},
             {"name": "context", "type": "Any", "required": False, "frontendType": "contextBuilder",
              "description": t("Daten aus vorherigen Schritten"), "default": ""},
             {"name": "documentList", "type": "DocumentList", "required": False, "frontendType": "hidden",

modules/features/graphicalEditor/nodeDefinitions/file.py

@@ -28,7 +28,7 @@ FILE_NODES = [
         ],
         "inputs": 1,
         "outputs": 1,
-        "inputPorts": {0: {"accepts": ["AiResult", "TextResult", "Transit"]}},
+        "inputPorts": {0: {"accepts": ["AiResult", "TextResult", "Transit", "FormPayload"]}},
         "outputPorts": {0: {"schema": "DocumentList"}},
         "meta": {"icon": "mdi-file-plus-outline", "color": "#2196F3", "usesAi": False},
         "_method": "file",

modules/features/graphicalEditor/portTypes.py

@@ -221,9 +221,9 @@ PORT_TYPE_CATALOG: Dict[str, PortSchema] = {
         PortField(name="prompt", type="str",
                   description="Prompt"),
         PortField(name="response", type="str",
-                  description="Antworttext"),
+                  description="Antworttext", recommended=True),
         PortField(name="responseData", type="Dict", required=False,
-                  description="Strukturierte Antwort"),
+                  description="Strukturierte Antwort (nur bei JSON-Ausgabe)"),
         PortField(name="context", type="str",
                   description="Kontext"),
         PortField(name="documents", type="List[Document]",
@@ -660,8 +660,11 @@ def normalizeToSchema(raw: Any, schemaName: str) -> Dict[str, Any]:
     if not schema or schemaName == "Transit":
         return result
 
+    # Only default **required** fields. Optional fields stay absent so DataRefs / context
+    # resolution never pick a synthetic `{}` or `[]` (e.g. AiResult.responseData when the
+    # model returned plain text only).
     for field in schema.fields:
-        if field.name not in result:
+        if field.name not in result and field.required:
             result[field.name] = _defaultForType(field.type)
 
     return result

modules/interfaces/interfaceDbManagement.py

@@ -114,7 +114,15 @@ class ComponentObjects:
         
         # Update database context
         self.db.updateContext(self.userId)
-    
+
+    def _effective_user_id(self) -> Optional[str]:
+        """User id for audit + FileData writes; singleton hub may unset userId but keep currentUser."""
+        if self.userId:
+            return self.userId
+        if self.currentUser is not None:
+            return getattr(self.currentUser, "id", None)
+        return None
+
     def __del__(self):
         """Cleanup method to close database connection."""
         if hasattr(self, 'db') and self.db is not None:
@@ -1379,10 +1387,31 @@ class ComponentObjects:
             fileSize=fileSize,
             fileHash=fileHash,
         )
-        
+        # Ensure audit user is always stored: workflow/singleton contexts sometimes leave
+        # the connector without _current_user_id, so _saveRecord skips sysCreatedBy →
+        # getFile/createFileData RBAC then breaks (None != self.userId).
+        uid = self._effective_user_id()
+        if uid:
+            fileItem = fileItem.model_copy(update={"sysCreatedBy": str(uid)})
+
         # Store in database
         self.db.recordCreate(FileItem, fileItem)
-        
+        verify = self.db.getRecordset(FileItem, recordFilter={"id": fileItem.id})
+        verify_creator = (verify[0].get("sysCreatedBy") if verify else None)
+        logger.info(
+            "createFile: id=%s name=%s scope=%s model_sysCreatedBy=%r db_sysCreatedBy=%r mandateId=%r featureInstanceId=%r "
+            "verify_rows=%s db=%s",
+            fileItem.id,
+            uniqueName,
+            fileItem.scope,
+            getattr(fileItem, "sysCreatedBy", None),
+            verify_creator,
+            mandateId or None,
+            featureInstanceId if featureInstanceId else None,
+            len(verify) if verify else 0,
+            getattr(self.db, "dbDatabase", "?"),
+        )
+
         return fileItem
     
     def _isFileOwner(self, file) -> bool:
@@ -1579,14 +1608,134 @@ class ComponentObjects:
         return success
 
     # FileData methods - data operations
-    
+
+    def _getFileItemForDataWrite(self, fileId: str) -> Optional[FileItem]:
+        """Resolve FileItem for storing FileData: RBAC-aware getFile, then same-user row fallback.
+
+        createFile() can insert a row that getFile() still hides (e.g. scope NULL vs GROUP rules,
+        or connector / context edge cases). The creator must still be allowed to attach blob data.
+        """
+        logger.info(
+            "[FileData] resolve start fileId=%s iface_userId=%r effective_uid=%r mandateId=%r featureInstanceId=%r db=%s",
+            fileId,
+            self.userId,
+            self._effective_user_id(),
+            self.mandateId,
+            self.featureInstanceId,
+            getattr(self.db, "dbDatabase", "?"),
+        )
+        file = self.getFile(fileId)
+        if file:
+            logger.info("[FileData] getFile OK fileId=%s", fileId)
+            return file
+        uid = self._effective_user_id()
+        if not uid:
+            logger.error(
+                "[FileData] FAIL no user id fileId=%s userId=%r hasCurrentUser=%s",
+                fileId,
+                self.userId,
+                self.currentUser is not None,
+            )
+            return None
+        uid_s = str(uid)
+        rows = self.db.getRecordset(FileItem, recordFilter={"id": fileId})
+        if not rows:
+            logger.error(
+                "[FileData] FAIL no FileItem row fileId=%s (createFile committed to same db? db=%s)",
+                fileId,
+                getattr(self.db, "dbDatabase", "?"),
+            )
+            return None
+        row = dict(rows[0])
+        creator = row.get("sysCreatedBy")
+        creator_s = str(creator) if creator is not None else None
+        if creator_s != uid_s:
+            if not creator_s:
+                try:
+                    self.db.recordModify(FileItem, fileId, {"sysCreatedBy": uid_s})
+                    row["sysCreatedBy"] = uid_s
+                    logger.warning(
+                        "[FileData] patched NULL sysCreatedBy fileId=%s -> %s",
+                        fileId,
+                        uid_s,
+                    )
+                except Exception as e:
+                    logger.error(
+                        "[FileData] FAIL patch sysCreatedBy fileId=%s: %s",
+                        fileId,
+                        e,
+                        exc_info=True,
+                    )
+                    return None
+            else:
+                # _saveRecord used to overwrite explicit creators with contextvar "system"
+                if creator_s == "system":
+                    try:
+                        self.db.recordModify(FileItem, fileId, {"sysCreatedBy": uid_s})
+                        row["sysCreatedBy"] = uid_s
+                        logger.warning(
+                            "[FileData] patched sysCreatedBy system→user fileId=%s -> %s",
+                            fileId,
+                            uid_s,
+                        )
+                    except Exception as e:
+                        logger.error(
+                            "[FileData] FAIL patch system sysCreatedBy fileId=%s: %s",
+                            fileId,
+                            e,
+                            exc_info=True,
+                        )
+                        return None
+                else:
+                    logger.error(
+                        "[FileData] FAIL creator mismatch fileId=%s row.sysCreatedBy=%r (%s) effective_uid=%r (%s) scope=%r",
+                        fileId,
+                        creator,
+                        type(creator).__name__,
+                        uid,
+                        type(uid).__name__,
+                        row.get("scope"),
+                    )
+                    return None
+        logger.info(
+            "[FileData] RBAC miss, owner fallback OK fileId=%s scope=%r sysCreatedBy=%r",
+            fileId,
+            row.get("scope"),
+            row.get("sysCreatedBy"),
+        )
+        try:
+            if row.get("sysCreatedAt") is None or row.get("sysCreatedAt") in (0, 0.0):
+                row["sysCreatedAt"] = getUtcTimestamp()
+            if row.get("scope") is None:
+                row["scope"] = "personal"
+            if row.get("neutralize") is None:
+                row["neutralize"] = False
+            return FileItem(**row)
+        except Exception as e:
+            logger.error(
+                "[FileData] FAIL FileItem(**row) fileId=%s keys=%s err=%s",
+                fileId,
+                list(row.keys()),
+                e,
+                exc_info=True,
+            )
+            return None
+
     def createFileData(self, fileId: str, data: bytes) -> bool:
         """Stores the binary data of a file in the database."""
         try:
+            logger.info(
+                "[FileData] createFileData enter fileId=%s bytes=%s",
+                fileId,
+                len(data) if data is not None else 0,
+            )
             # Check file access
-            file = self.getFile(fileId)
+            file = self._getFileItemForDataWrite(fileId)
             if not file:
-                logger.error(f"File with ID {fileId} not found when storing data")
+                logger.error(
+                    "[FileData] FAIL _getFileItemForDataWrite returned None fileId=%s",
+                    fileId,
+                )
                 return False
                 
             # Determine if this is a text-based format
@@ -1630,13 +1779,11 @@ class ComponentObjects:
             }
             
             self.db.recordCreate(FileData, fileDataObj)
-            
-            # Clear cache to ensure fresh data
-            
-            logger.debug(f"Successfully stored data for file {fileId} (base64Encoded: {base64Encoded})")
+
+            logger.info("[FileData] recordCreate OK fileId=%s base64Encoded=%s", fileId, base64Encoded)
             return True
         except Exception as e:
-            logger.error(f"Error storing data for file {fileId}: {str(e)}")
+            logger.error("Error storing data for file %s: %s", fileId, e, exc_info=True)
             return False
 
     def getFileData(self, fileId: str) -> Optional[bytes]:

modules/interfaces/interfaceRbac.py

@@ -747,6 +747,7 @@ def buildFilesScopeWhereClause(
       Only own files: sysCreatedBy = currentUser
 
     WITH instance context (Instanz-Seiten):
+      - scope = 'personal' AND sysCreatedBy = me (creator's personal files; e.g. workflow outputs)
       - sysCreatedBy = me  AND featureInstanceId = X   (own personal files of this instance)
       - scope = 'featureInstance' AND featureInstanceId = X
       - scope = 'mandate'         AND mandateId = M     (M = mandate of the instance)
@@ -780,6 +781,15 @@ def buildFilesScopeWhereClause(
     scopeParts: List[str] = []
     scopeValues: List = []
 
+    # Personal files created by this user must remain visible even when the request
+    # carries mandate/instance context (GROUP reads use this clause). Otherwise
+    # createFile → createFileData → getFile fails and workflow outputs vanish from /files.
+    # Also treat scope IS NULL as legacy/personal for the owner (column default not applied).
+    scopeParts.append(
+        '(("scope" = \'personal\' OR "scope" IS NULL) AND "sysCreatedBy" = %s)'
+    )
+    scopeValues.append(currentUser.id)
+
     if featureInstanceId:
         # 1) Own personal files of this specific instance
         scopeParts.append('("sysCreatedBy" = %s AND "featureInstanceId" = %s)')

modules/serviceCenter/services/serviceAi/subAiCallLooping.py

@@ -142,6 +142,8 @@ class AiCallLooper:
         MAX_MERGE_FAILS = 3
         mergeFailCount = 0  # Global counter for merge failures across entire loop
         lastValidCompletePart = None  # Store last successfully parsed completePart for fallback
+        MAX_CONSECUTIVE_EMPTY_RESPONSES = 3
+        consecutive_empty_responses = 0
         
         # Get parent operation ID for iteration operations (parentId should be operationId, not log entry ID)
         parentOperationId = operationId  # Use the parent's operationId directly
@@ -284,8 +286,26 @@ class AiCallLooper:
                     break
                 
                 if not result or not result.strip():
-                    logger.warning(f"Iteration {iteration}: Empty response, stopping")
-                    break
+                    consecutive_empty_responses += 1
+                    logger.warning(
+                        "Iteration %s: Empty AI response (consecutive %s/%s) modelName=%s errorCount=%s",
+                        iteration,
+                        consecutive_empty_responses,
+                        MAX_CONSECUTIVE_EMPTY_RESPONSES,
+                        getattr(response, "modelName", None),
+                        getattr(response, "errorCount", None),
+                    )
+                    if iterationOperationId:
+                        self.services.chat.progressLogFinish(iterationOperationId, False)
+                    if consecutive_empty_responses >= MAX_CONSECUTIVE_EMPTY_RESPONSES:
+                        logger.error(
+                            "Stopping loop: %s consecutive empty responses from model",
+                            consecutive_empty_responses,
+                        )
+                        break
+                    continue
+                
+                consecutive_empty_responses = 0
                 
                 # Check if this is a text response (not document generation)
                 # Text responses don't need JSON parsing - return immediately after first successful response
@@ -535,7 +555,12 @@ class AiCallLooper:
         # This code path should never be reached because all registered use cases
         # return early when JSON is complete. This would only execute for use cases that
         # require section extraction, but no such use cases are currently registered.
-        logger.error(f"Unexpected code path: reached end of loop without return for use case '{useCaseId}'")
+        logger.error(
+            "End of callAiWithLooping without success for use case %r (iterations=%s, lastResultLen=%s)",
+            useCaseId,
+            iteration,
+            len(result) if isinstance(result, str) else 0,
+        )
         return result if result else ""
     
     def _isJsonStringIncomplete(self, jsonString: str) -> bool:

modules/serviceCenter/services/serviceAi/subStructureGeneration.py

@@ -90,8 +90,7 @@ class StructureGenerator:
         )
         
         try:
-            # Baue Chapter-Struktur-Prompt mit Content-Index
-            structurePrompt = self._buildChapterStructurePrompt(
+            structurePrompt, templateStructure = self._buildChapterStructurePrompt(
                 userPrompt=userPrompt,
                 contentParts=contentParts,
                 outputFormat=outputFormat
@@ -108,12 +107,6 @@ class StructureGenerator:
                 resultFormat="json"
             )
             
-            structurePrompt, templateStructure = self._buildChapterStructurePrompt(
-                userPrompt=userPrompt,
-                contentParts=contentParts,
-                outputFormat=outputFormat
-            )
-            
             # Create prompt builder for continuation support
             async def buildChapterStructurePromptWithContinuation(
                 continuationContext: Any,
@@ -196,6 +189,13 @@ CRITICAL:
                 contentParts=None  # Do not pass ContentParts - only metadata needed, not content extraction
             )
             
+            if not isinstance(aiResponseJson, str) or not aiResponseJson.strip():
+                raise ValueError(
+                    "Structure generation returned no JSON text from the model (empty response after retries). "
+                    "Check the AI provider, allowed models, billing, and debug artifact "
+                    "'chapter_structure_generation_response'."
+                )
+            
             # Parse the complete JSON response (looping system already handles completion)
             extractedJson = self.services.utils.jsonExtractString(aiResponseJson)
             parsedJson, parseError, cleanedJson = self.services.utils.jsonTryParse(extractedJson)
@@ -215,7 +215,12 @@ CRITICAL:
                         raise ValueError(f"Failed to parse JSON structure after repair: {str(parseError)}")
                 else:
                     logger.error(f"Failed to repair JSON. Parse error: {str(parseError)}")
-                    logger.error(f"Cleaned JSON preview (first 500 chars): {cleanedJson[:500]}")
+                    raw_preview = (extractedJson or "")[:500]
+                    logger.error(
+                        "Raw extract preview (first 500 chars): %r",
+                        raw_preview,
+                    )
+                    logger.error(f"Cleaned JSON preview (first 500 chars): {cleanedJson[:500]!r}")
                     raise ValueError(f"Failed to parse JSON structure: {str(parseError)}")
             else:
                 structure = parsedJson

modules/serviceCenter/services/serviceChat/mainServiceChat.py

@@ -23,7 +23,11 @@ class ChatService:
         from modules.interfaces.interfaceDbManagement import getInterface as getComponentInterface
         from modules.interfaces.interfaceDbChat import getInterface as getChatInterface
         self.interfaceDbApp = getAppInterface(context.user, mandateId=context.mandate_id)
-        self.interfaceDbComponent = getComponentInterface(context.user, mandateId=context.mandate_id)
+        self.interfaceDbComponent = getComponentInterface(
+            context.user,
+            mandateId=context.mandate_id,
+            featureInstanceId=context.feature_instance_id,
+        )
         self.interfaceDbChat = getChatInterface(
             context.user,
             mandateId=context.mandate_id,

modules/serviceCenter/services/serviceGeneration/renderers/rendererCodeCsv.py

@@ -79,7 +79,15 @@ class RendererCodeCsv(BaseCodeRenderer):
         
         return renderedDocs
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """
         Render method for document generation compatibility.
         Delegates to document renderer if needed, or handles code files directly.
@@ -94,7 +102,7 @@ class RendererCodeCsv(BaseCodeRenderer):
             # Document generation path - delegate to document renderer
             from .rendererCsv import RendererCsv
             documentRenderer = RendererCsv(self.services)
-            return await documentRenderer.render(extractedContent, title, userPrompt, aiService)
+            return await documentRenderer.render(extractedContent, title, userPrompt, aiService, style=style)
     
     def _validateAndFixCsv(self, content: str) -> str:
         """Validate CSV structure and fix common issues."""

modules/serviceCenter/services/serviceGeneration/renderers/rendererCodeJson.py

@@ -91,7 +91,15 @@ class RendererCodeJson(BaseCodeRenderer):
         
         return renderedDocs
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """
         Render method for document generation compatibility.
         Delegates to document renderer if needed, or handles code files directly.
@@ -107,7 +115,7 @@ class RendererCodeJson(BaseCodeRenderer):
             # Import here to avoid circular dependency
             from .rendererJson import RendererJson
             documentRenderer = RendererJson(self.services)
-            return await documentRenderer.render(extractedContent, title, userPrompt, aiService)
+            return await documentRenderer.render(extractedContent, title, userPrompt, aiService, style=style)
     
     def _extractJsonStatistics(self, parsed: Any) -> Dict[str, Any]:
         """Extract JSON statistics for validation (object count, array count, key count)."""

modules/serviceCenter/services/serviceGeneration/renderers/rendererCodeXml.py

@@ -78,11 +78,20 @@ class RendererCodeXml(BaseCodeRenderer):
         
         return renderedDocs
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """
         Render method for document generation compatibility.
         For XML, we only support code generation (no document renderer exists yet).
         """
+        _ = style
         # Check if this is code generation (has files array)
         if "files" in extractedContent:
             # Code generation path - use renderCodeFiles

modules/serviceCenter/services/serviceGeneration/renderers/rendererCsv.py

@@ -39,8 +39,17 @@ class RendererCsv(BaseRenderer):
         """
         return ["table", "code_block"]
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """Render extracted JSON content to CSV format. Produces one CSV file per table section."""
+        _ = style
         try:
             # Validate JSON structure
             if not self._validateJsonStructure(extractedContent):

modules/serviceCenter/services/serviceGeneration/renderers/rendererImage.py

@@ -43,8 +43,17 @@ class RendererImage(BaseRenderer):
         """
         return ["image"]
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """Render extracted JSON content to image format using AI image generation."""
+        _ = style
         try:
             # Generate AI image from content
             imageContent = await self._generateAiImage(extractedContent, title, userPrompt, aiService)

modules/serviceCenter/services/serviceGeneration/renderers/rendererJson.py

@@ -42,8 +42,17 @@ class RendererJson(BaseRenderer):
         # Return all types except image
         return [st for st in supportedSectionTypes if st != "image"]
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """Render extracted JSON content to JSON format."""
+        _ = style
         try:
             # The extracted content should already be JSON from the AI
             # Just validate and format it

modules/serviceCenter/services/serviceGeneration/renderers/rendererMarkdown.py

@@ -40,8 +40,17 @@ class RendererMarkdown(BaseRenderer):
         from modules.datamodels.datamodelJson import supportedSectionTypes
         return [st for st in supportedSectionTypes if st != "image"]
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """Render extracted JSON content to Markdown format."""
+        _ = style
         try:
             # Generate markdown from JSON structure
             markdownContent = self._generateMarkdownFromJson(extractedContent, title)

modules/serviceCenter/services/serviceGeneration/renderers/rendererText.py

@@ -8,7 +8,7 @@ import re
 
 from .documentRendererBaseTemplate import BaseRenderer
 from modules.datamodels.datamodelDocument import RenderedDocument
-from typing import Dict, Any, List, Optional
+from typing import Dict, Any, List, Optional, Union
 
 class RendererText(BaseRenderer):
     """Renders content to plain text format with format-specific extraction."""
@@ -76,8 +76,17 @@ class RendererText(BaseRenderer):
         # Text renderer accepts all types except images
         return [st for st in supportedSectionTypes if st != "image"]
     
-    async def render(self, extractedContent: Dict[str, Any], title: str, userPrompt: str = None, aiService=None, *, style: Dict[str, Any] = None) -> List[RenderedDocument]:
+    async def render(
+        self,
+        extractedContent: Dict[str, Any],
+        title: str,
+        userPrompt: str = None,
+        aiService=None,
+        *,
+        style: Dict[str, Any] = None,
+    ) -> List[RenderedDocument]:
         """Render extracted JSON content to plain text format."""
+        _ = style  # unified style from renderReport; plain text ignores formatting hints
         try:
             # Generate text from JSON structure
             textContent = self._generateTextFromJson(extractedContent, title)
@@ -263,16 +272,16 @@ class RendererText(BaseRenderer):
             textParts = []
             
             # Create table header
-            headerLine = " | ".join(str(header) for header in headers)
+            headerLine = " | ".join(self._tableCellToPlainText(h) for h in headers)
             textParts.append(headerLine)
             
             # Add separator line
-            separatorLine = " | ".join("-" * len(str(header)) for header in headers)
+            separatorLine = " | ".join("-" * len(self._tableCellToPlainText(h)) for h in headers)
             textParts.append(separatorLine)
             
             # Add data rows
             for row in rows:
-                rowLine = " | ".join(str(cellData) for cellData in row)
+                rowLine = " | ".join(self._tableCellToPlainText(cellData) for cellData in row)
                 textParts.append(rowLine)
             
             return '\n'.join(textParts)
@@ -299,6 +308,9 @@ class RendererText(BaseRenderer):
                     textParts.append(f"- {self._stripMarkdownForPlainText(item)}")
                 elif isinstance(item, dict) and "text" in item:
                     textParts.append(f"- {self._stripMarkdownForPlainText(item['text'])}")
+                elif isinstance(item, list):
+                    # markdownToDocumentJson: each item is List[InlineRun]
+                    textParts.append(f"- {self._inlineRunsToPlainText(item)}")
             
             return '\n'.join(textParts)
             
@@ -345,12 +357,54 @@ class RendererText(BaseRenderer):
         text = re.sub(r'`([^`]+)`', r'\1', text)
         return text.strip()
 
+    def _inlineRunsToPlainText(self, runs: Union[List[Any], Any]) -> str:
+        """Flatten InlineRun dicts (from markdownToDocumentJson) to a single string."""
+        if runs is None:
+            return ""
+        if isinstance(runs, dict):
+            runs = [runs]
+        if not isinstance(runs, list):
+            return self._stripMarkdownForPlainText(str(runs))
+        parts: List[str] = []
+        for run in runs:
+            if not isinstance(run, dict):
+                parts.append(str(run))
+                continue
+            t = run.get("type") or "text"
+            val = run.get("value", "")
+            if t == "text":
+                parts.append(str(val))
+            elif t in ("bold", "italic", "code"):
+                parts.append(str(val))
+            elif t == "link":
+                parts.append(str(val))
+            elif t == "image":
+                parts.append(f"[{val}]")
+            else:
+                parts.append(str(val))
+        return "".join(parts)
+
+    def _tableCellToPlainText(self, cell: Any) -> str:
+        """Table header/cell: plain str, legacy dict, or List[InlineRun]."""
+        if cell is None:
+            return ""
+        if isinstance(cell, str):
+            return self._stripMarkdownForPlainText(cell)
+        if isinstance(cell, list):
+            return self._inlineRunsToPlainText(cell)
+        if isinstance(cell, dict) and "text" in cell:
+            return self._stripMarkdownForPlainText(str(cell["text"]))
+        return self._stripMarkdownForPlainText(str(cell))
+
     def _renderJsonParagraph(self, paragraphData: Dict[str, Any]) -> str:
         """Render a JSON paragraph to text. Strips markdown for plain text output."""
         try:
             # Extract from nested content structure
             content = paragraphData.get("content", {})
             if isinstance(content, dict):
+                runs = self._inlineRunsFromContent(content)
+                if runs:
+                    return self._stripMarkdownForPlainText(self._inlineRunsToPlainText(runs))
                 text = content.get("text", "")
             elif isinstance(content, str):
                 text = content

modules/workflows/automation2/executors/actionNodeExecutor.py

@@ -326,11 +326,25 @@ class ActionNodeExecutor:
             if isinstance(dumped, dict) and isinstance(rawData, bytes) and len(rawData) > 0:
                 try:
                     from modules.interfaces.interfaceDbManagement import getInterface as _getMgmtInterface
+                    from modules.interfaces.interfaceDbApp import getInterface as _getAppInterface
                     from modules.security.rootAccess import getRootUser
                     _userId = context.get("userId")
                     _mandateId = context.get("mandateId")
                     _instanceId = context.get("instanceId")
-                    _mgmt = _getMgmtInterface(getRootUser(), mandateId=_mandateId, featureInstanceId=_instanceId)
+                    _owner = None
+                    if _userId:
+                        try:
+                            _umap = _getAppInterface(getRootUser()).getUsersByIds([str(_userId)])
+                            _owner = _umap.get(str(_userId))
+                        except Exception as _ue:
+                            logger.warning("Could not resolve workflow user for file persistence: %s", _ue)
+                    if _owner is None:
+                        _owner = getRootUser()
+                        logger.debug(
+                            "Persisting workflow document as root user (no resolved owner userId=%r)",
+                            _userId,
+                        )
+                    _mgmt = _getMgmtInterface(_owner, mandateId=_mandateId, featureInstanceId=_instanceId)
                     _docName = dumped.get("documentName") or f"workflow-result-{nodeId}.bin"
                     _mimeType = dumped.get("mimeType") or "application/octet-stream"
                     _fileItem = _mgmt.createFile(_docName, _mimeType, rawData)
@@ -345,6 +359,20 @@ class ActionNodeExecutor:
                 dumped["_hasBinaryData"] = True
             docsList.append(dumped)
 
+        # Clean DocumentList shape for document nodes (match file.create: documents + count, no AiResult fields)
+        if outputSchema == "DocumentList" and nodeType in ("ai.generateDocument", "ai.convertDocument"):
+            if not result.success:
+                return _normalizeError(
+                    RuntimeError(str(result.error or "document action failed")),
+                    outputSchema,
+                )
+            list_out: Dict[str, Any] = {
+                "documents": docsList,
+                "count": len(docsList),
+            }
+            _attachConnectionProvenance(list_out, resolvedParams, outputSchema, chatService, self.services)
+            return normalizeToSchema(list_out, outputSchema)
+
         extractedContext = ""
         if result.documents:
             doc = result.documents[0]

modules/workflows/automation2/graphUtils.py

@@ -7,6 +7,50 @@ from typing import Dict, List, Any, Tuple, Set, Optional
 logger = logging.getLogger(__name__)
 
 
+def _ai_result_text_from_documents(d: Dict[str, Any]) -> Optional[str]:
+    """Extract plain-text body from AiResult-style ``documents[0].documentData``."""
+    docs = d.get("documents")
+    if not isinstance(docs, list) or not docs:
+        return None
+    d0 = docs[0]
+    raw: Any = None
+    if isinstance(d0, dict):
+        raw = d0.get("documentData")
+    elif d0 is not None:
+        raw = getattr(d0, "documentData", None)
+    if raw is None:
+        return None
+    if isinstance(raw, bytes):
+        try:
+            t = raw.decode("utf-8").strip()
+            return t or None
+        except (UnicodeDecodeError, ValueError):
+            return None
+    if isinstance(raw, str):
+        s = raw.strip()
+        return s or None
+    return None
+
+
+def _ref_coalesce_empty_ai_result_text(data: Any, path: List[Any], resolved: Any) -> Any:
+    """If a ref targets AiResult text fields but resolves empty/missing, fall back to documents.
+
+    Needed when: optional ``responseData`` is absent (no synthetic ``{}``), ``response`` is
+    still empty but ``documents`` hold the model output, or legacy graphs bind responseData only.
+    """
+    if resolved not in (None, ""):
+        return resolved
+    if not isinstance(data, dict) or not path:
+        return resolved
+    head = path[0]
+    if head not in ("response", "responseData", "context"):
+        return resolved
+    if head == "context" and len(path) != 1:
+        return resolved
+    fb = _ai_result_text_from_documents(data)
+    return fb if fb is not None else resolved
+
+
 def parseGraph(graph: Dict[str, Any]) -> Tuple[List[Dict], List[Dict], Set[str]]:
     """
     Parse graph into nodes, connections, and node IDs.
@@ -356,14 +400,15 @@ def resolveParameterReferences(value: Any, nodeOutputs: Dict[str, Any]) -> Any:
                     data = data.get("data", data)
                 plist = list(path)
                 resolved = _get_by_path(data, plist)
-                if (
-                    resolved is None
-                    and isinstance(data, dict)
-                    and plist
-                    and plist[0] == "payload"
-                    and len(plist) > 1
-                ):
-                    resolved = _get_by_path(data, plist[1:])
+                if resolved is None and isinstance(data, dict) and plist:
+                    if plist[0] == "payload" and len(plist) > 1:
+                        # Strip explicit "payload" prefix (legacy DataPicker paths)
+                        resolved = _get_by_path(data, plist[1:])
+                    elif "payload" in data and isinstance(data["payload"], dict):
+                        # Form nodes store fields under {"payload": {fieldName: …}}.
+                        # DataPicker emits bare field paths like ["url"]; try under payload.
+                        resolved = _get_by_path(data["payload"], plist)
+                resolved = _ref_coalesce_empty_ai_result_text(data, plist, resolved)
                 return resolveParameterReferences(resolved, nodeOutputs)
             return value
         if value.get("type") == "value":
@@ -386,16 +431,27 @@ def resolveParameterReferences(value: Any, nodeOutputs: Dict[str, Any]) -> Any:
             if len(parts) < 2:
                 return json.dumps(data) if isinstance(data, (dict, list)) else str(data)
             rest = ".".join(parts[1:])
-            if data is None:
+
+            def _walk(root, keys):
+                cur = root
+                for k in keys:
+                    if isinstance(cur, dict) and k in cur:
+                        cur = cur[k]
+                    elif isinstance(cur, (list, tuple)) and k.isdigit():
+                        cur = cur[int(k)]
+                    else:
+                        return None
+                return cur
+
+            keys = rest.split(".")
+            result = _walk(data, keys)
+            # Form nodes store fields under {"payload": {field: …}}.
+            # Fall back to looking under "payload" when the direct path misses.
+            if result is None and isinstance(data, dict) and "payload" in data:
+                result = _walk(data["payload"], keys)
+            if result is None:
                 return m.group(0)
-            for k in rest.split("."):
-                if isinstance(data, dict) and k in data:
-                    data = data[k]
-                elif isinstance(data, (list, tuple)) and k.isdigit():
-                    data = data[int(k)]
-                else:
-                    return m.group(0)
-            return str(data) if data is not None else m.group(0)
+            return str(result) if not isinstance(result, (dict, list)) else json.dumps(result, ensure_ascii=False)
         return re.sub(r"\{\{\s*([^}]+)\s*\}\}", repl, value)
     if isinstance(value, list):
         # contextBuilder: list where every item is a `{"type":"ref",...}` envelope.

modules/workflows/methods/methodAi/_common.py

@@ -11,12 +11,15 @@ def serialize_context(val: Any) -> str:
     """Convert any context value to a readable string for use in AI prompts.
 
     - None / empty string → ""
+    - empty dict (no keys) → ""  (avoids literal "{}" in file.create / prompts)
     - str → as-is
     - dict / list → pretty-printed JSON
     - anything else → str()
     """
     if val is None or val == "" or val == []:
         return ""
+    if isinstance(val, dict) and len(val) == 0:
+        return ""
     if isinstance(val, str):
         return val.strip()
     try:

modules/workflows/methods/methodAi/actions/generateDocument.py

@@ -23,8 +23,10 @@ async def generateDocument(self, parameters: Dict[str, Any]) -> ActionResult:
     
     documentList = parameters.get("documentList", [])
     documentType = parameters.get("documentType")
-    # Optional: if omitted, formats determined from prompt by AI
-    resultType = parameters.get("resultType")
+    # Prefer explicit outputFormat (flow UI); resultType remains for legacy / API callers.
+    resultType = parameters.get("outputFormat") or parameters.get("resultType")
+    if isinstance(resultType, str):
+        resultType = resultType.strip().lstrip(".").lower() or None
     
     if not resultType:
         logger.debug("resultType not provided - formats will be determined from prompt by AI")
@@ -49,8 +51,12 @@ async def generateDocument(self, parameters: Dict[str, Any]) -> ActionResult:
             else:
                 docRefList = DocumentReferenceList(references=[])
         
-        # Prepare title
-        title = parameters.get("documentType") or "Generated Document"
+        title_raw = parameters.get("title")
+        title = (title_raw.strip() if isinstance(title_raw, str) else "") or None
+        if not title and isinstance(documentType, str) and documentType.strip():
+            title = documentType.strip()
+        if not title:
+            title = "Generated Document"
         
         # Call AI service for document generation
         # callAiContent handles documentList internally via Phases 5A-5E
@@ -98,6 +104,8 @@ async def generateDocument(self, parameters: Dict[str, Any]) -> ActionResult:
                         "actionType": "ai.generateDocument",
                         "documentType": documentType,
                         "resultType": resultType,
+                        "outputFormat": resultType,
+                        "title": title,
                     }
                 ))
         
@@ -119,14 +127,15 @@ async def generateDocument(self, parameters: Dict[str, Any]) -> ActionResult:
                         docName = sanitized
             
             # Determine mime type
+            rt = resultTypeFallback
             mimeType = "text/plain"
-            if resultType == "html":
+            if rt == "html":
                 mimeType = "text/html"
-            elif resultType == "json":
+            elif rt == "json":
                 mimeType = "application/json"
-            elif resultType == "pdf":
+            elif rt == "pdf":
                 mimeType = "application/pdf"
-            elif resultType == "md":
+            elif rt == "md":
                 mimeType = "text/markdown"
             
             documents.append(ActionDocument(
@@ -137,6 +146,8 @@ async def generateDocument(self, parameters: Dict[str, Any]) -> ActionResult:
                     "actionType": "ai.generateDocument",
                     "documentType": documentType,
                     "resultType": resultType,
+                    "outputFormat": resultType,
+                    "title": title,
                 }
             ))
         

modules/workflows/methods/methodAi/methodAi.py

@@ -289,6 +289,30 @@ class MethodAi(MethodBase):
                         required=True,
                         description="Description of the document to generate"
                     ),
+                    "outputFormat": WorkflowActionParameter(
+                        name="outputFormat",
+                        type="str",
+                        frontendType=FrontendType.SELECT,
+                        frontendOptions=["docx", "pdf", "txt", "html", "md"],
+                        required=False,
+                        default="docx",
+                        description="Rendered output format (same choices as file.create). If omitted alongside resultType, the model may infer format from the prompt."
+                    ),
+                    "title": WorkflowActionParameter(
+                        name="title",
+                        type="str",
+                        frontendType=FrontendType.TEXT,
+                        required=False,
+                        description="Document title / metadata (optional); used as generation title and for file naming hints."
+                    ),
+                    "context": WorkflowActionParameter(
+                        name="context",
+                        type="Any",
+                        frontendType=FrontendType.TEXTAREA,
+                        required=False,
+                        default="",
+                        description="Additional structured or text context from upstream steps; serialized into the prompt."
+                    ),
                     "documentList": WorkflowActionParameter(
                         name="documentList",
                         type="DocumentList",
@@ -302,16 +326,15 @@ class MethodAi(MethodBase):
                         frontendType=FrontendType.SELECT,
                         frontendOptions=["letter", "memo", "proposal", "contract", "report", "email"],
                         required=False,
-                        description="Type of document"
+                        description="Type of document (content hint for the model); used as title fallback when title is empty."
                     ),
                     "resultType": WorkflowActionParameter(
                         name="resultType",
                         type="str",
                         frontendType=FrontendType.TEXT,
                         required=False,
-                        default="txt",
-                        description="Output format (e.g., txt, html, pdf, docx, md, json, csv, xlsx, pptx, png, jpg). Optional: if omitted, formats are determined from prompt by AI. Default \"txt\" is validation fallback only. With per-document format determination, AI can determine different formats for different documents based on prompt."
-                    )
+                        description="Legacy/API output format extension (e.g. txt, docx). Ignored when outputFormat is set."
+                    ),
                 },
                 execute=generateDocument.__get__(self, self.__class__)
             ),

modules/workflows/methods/methodFile/actions/create.py

@@ -35,6 +35,12 @@ def _persistDocumentsToUserFiles(
             return
     if not mgmt:
         return
+    logger.info(
+        "file.create persist: mgmt=%s id(mgmt)=%s has_createFileData=%s",
+        type(mgmt).__name__,
+        id(mgmt),
+        hasattr(mgmt, "createFileData"),
+    )
     for doc in action_documents:
         try:
             doc_data = doc.documentData if hasattr(doc, "documentData") else doc.get("documentData")
@@ -54,8 +60,15 @@ def _persistDocumentsToUserFiles(
                 or doc.get("mimeType")
                 or "application/octet-stream"
             )
+            logger.info(
+                "file.create persist: calling createFile name=%s bytes=%s",
+                doc_name,
+                len(content),
+            )
             file_item = mgmt.createFile(doc_name, mime, content)
-            mgmt.createFileData(file_item.id, content)
+            logger.info("file.create persist: createFile returned id=%s", file_item.id)
+            ok = mgmt.createFileData(file_item.id, content)
+            logger.info("file.create persist: createFileData returned %s for id=%s", ok, file_item.id)
             meta = getattr(doc, "validationMetadata", None) or doc.get("validationMetadata") or {}
             if isinstance(meta, dict):
                 meta["fileId"] = file_item.id
@@ -79,6 +92,10 @@ async def create(self, parameters: Dict[str, Any]) -> ActionResult:
     context = serialize_context(raw_context)
 
     if not context:
+        logger.warning(
+            "file.create: context empty after resolve — check DataRefs (e.g. Antworttext / "
+            "documents[0].documentData from the AI step)."
+        )
         return ActionResult.isFailure(error="context is required (connect an AI node or provide text)")
 
     outputFormat = (parameters.get("outputFormat") or "docx").strip().lower().lstrip(".")