gateway/modules/services/serviceAi/subDocumentGeneration.py
2025-10-18 02:29:03 +02:00

933 lines
45 KiB
Python

import logging
from typing import Dict, Any, List, Optional, Tuple, Union
from modules.datamodels.datamodelChat import ChatDocument
from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationType
logger = logging.getLogger(__name__)
class SubDocumentGeneration:
"""Document generation operations including single-file and multi-file generation."""
def __init__(self, services, aiObjects, documentProcessor):
"""Initialize document generation service.
Args:
services: Service center instance for accessing other services
aiObjects: Initialized AiObjects instance
documentProcessor: Document processing service instance
"""
self.services = services
self.aiObjects = aiObjects
self.documentProcessor = documentProcessor
async def callAiWithDocumentGeneration(
self,
prompt: str,
documents: Optional[List[ChatDocument]],
options: AiCallOptions,
outputFormat: str,
title: Optional[str]
) -> Dict[str, Any]:
"""
Handle AI calls with document generation in specific output format.
Now supports both single-file and multi-file generation.
Args:
prompt: The main prompt for the AI call
documents: Optional list of documents to process
options: AI call configuration options
outputFormat: Target output format (html, pdf, docx, txt, md, json, csv, xlsx)
title: Optional title for generated documents
Returns:
Dict with generated documents and metadata
"""
try:
# Use AI to analyze prompt intent
prompt_analysis = await self._analyzePromptIntent(prompt, self)
logger.info(f"Prompt analysis result: {prompt_analysis}")
if prompt_analysis.get("is_multi_file", False):
return await self._callAiWithMultiFileGeneration(
prompt, documents, options, outputFormat, title, prompt_analysis
)
else:
return await self._callAiWithSingleFileGeneration(
prompt, documents, options, outputFormat, title
)
except Exception as e:
logger.error(f"Error in document generation: {str(e)}")
return {
"success": False,
"error": str(e),
"content": "",
"rendered_content": "",
"mime_type": "text/plain",
"filename": f"error_{outputFormat}",
"format": outputFormat,
"title": title or "Error",
"documents": []
}
async def _callAiWithSingleFileGeneration(
self,
prompt: str,
documents: Optional[List[ChatDocument]],
options: AiCallOptions,
outputFormat: str,
title: Optional[str],
generationPrompt: Optional[str] = None
) -> Dict[str, Any]:
"""Handle single-file document generation (existing functionality)."""
import time
# Create progress logger
workflow = self.services.currentWorkflow
progressLogger = self.services.workflow.createProgressLogger(workflow)
operationId = f"docGenSingle_{workflow.id}_{int(time.time())}"
try:
# Start progress tracking
progressLogger.startOperation(
operationId,
"Generate",
"Single-file Generation",
f"Processing {len(documents) if documents else 0} documents"
)
# Get format-specific extraction prompt from generation service
from modules.services.serviceGeneration.mainServiceGeneration import GenerationService
generation_service = GenerationService(self.services)
# Use default title if not provided
if not title:
title = "AI Generated Document"
# Update progress - generating extraction prompt
progressLogger.updateProgress(operationId, 0.1, "Generating prompt")
# Get format-specific extraction prompt
extractionPrompt = await generation_service.getExtractionPrompt(
outputFormat=outputFormat,
userPrompt=prompt,
title=title,
aiService=self
)
# Update progress - starting AI processing
progressLogger.updateProgress(operationId, 0.3, "AI processing")
# Process documents with format-specific prompt using JSON mode
# This ensures structured JSON output instead of text
aiResponseJson = await self._callAiJson(extractionPrompt, documents, options)
# Update progress - AI processing completed
progressLogger.updateProgress(operationId, 0.6, "Processing done")
# Validate JSON response
if not isinstance(aiResponseJson, dict) or "sections" not in aiResponseJson:
raise Exception("AI response is not valid JSON document structure")
# Emit raw extracted data as a chat message attachment before rendering
try:
await self._postRawDataChatMessage(aiResponseJson, label="raw_extraction_single")
except Exception:
logger.warning("Failed to emit raw extraction chat message (single-file)")
# Generate filename from document metadata
parsedFilename = None
try:
if aiResponseJson.get("metadata", {}).get("title"):
title = aiResponseJson["metadata"]["title"]
# Clean title for filename
import re
parsed = re.sub(r"[^a-zA-Z0-9._-]", "-", title)
parsed = re.sub(r"-+", "-", parsed).strip('-')
if parsed:
parsedFilename = f"{parsed}.{outputFormat}"
except Exception:
parsedFilename = None
# Use AI generation to enhance the extracted JSON before rendering
enhancedContent = aiResponseJson # Default to original
if prompt:
try:
from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationType
# Get generation prompt
generationPrompt = await generation_service.getGenerationPrompt(
outputFormat=outputFormat,
userPrompt=prompt,
title=title,
aiService=self
)
# Prepare the AI call
request_options = AiCallOptions()
request_options.operationType = OperationType.GENERAL
# Create context with the extracted JSON content
import json
context = f"Extracted JSON content:\n{json.dumps(aiResponseJson, indent=2)}"
request = AiCallRequest(
prompt=generationPrompt,
context=context,
options=request_options
)
# Call AI to enhance the content
response = await self.aiObjects.call(request)
# Save generation prompt and response to debug
try:
from modules.shared.debugLogger import writeDebugFile
debugData = {
"output_format": outputFormat,
"title": title,
"context_length": len(context),
"extracted_content_keys": list(aiResponseJson.keys()) if isinstance(aiResponseJson, dict) else []
}
writeDebugFile(generationPrompt, "generation_single", debugData)
writeDebugFile(response.content or '', "generation_single_response")
except Exception:
pass
if response and response.content:
# Parse the AI response as JSON
try:
import re
result = response.content.strip()
# Check if result is empty after stripping
if not result:
logger.warning("AI generation returned empty content after stripping, using original content")
enhancedContent = aiResponseJson
else:
# Extract JSON from markdown if present
json_match = re.search(r'```json\s*\n(.*?)\n```', result, re.DOTALL)
if json_match:
result = json_match.group(1).strip()
elif result.startswith('```json'):
result = re.sub(r'^```json\s*', '', result)
result = re.sub(r'\s*```$', '', result)
elif result.startswith('```'):
result = re.sub(r'^```\s*', '', result)
result = re.sub(r'\s*```$', '', result)
# Check if result is still empty after markdown extraction
if not result:
logger.warning("AI generation returned empty content after markdown extraction, using original content")
enhancedContent = aiResponseJson
else:
# Try to parse JSON with better error handling
try:
enhancedContent = json.loads(result)
logger.info(f"AI enhanced JSON content successfully")
except json.JSONDecodeError as jsonError:
# Try to fix common JSON issues
fixed_result = self._attemptJsonFix(result)
if fixed_result != result:
try:
enhancedContent = json.loads(fixed_result)
logger.info(f"AI enhanced JSON content successfully after fixing")
except json.JSONDecodeError:
logger.warning(f"AI generation returned invalid JSON even after fixing: {str(jsonError)}, using original content")
enhancedContent = aiResponseJson
else:
logger.warning(f"AI generation returned invalid JSON: {str(jsonError)}, using original content")
enhancedContent = aiResponseJson
except json.JSONDecodeError as e:
logger.warning(f"AI generation returned invalid JSON: {str(e)}, using original content")
enhancedContent = aiResponseJson
else:
logger.warning("AI generation returned empty response, using original content")
enhancedContent = aiResponseJson
except Exception as e:
logger.warning(f"AI generation failed: {str(e)}, using original content")
enhancedContent = aiResponseJson
# Render the enhanced JSON content
renderedContent, mimeType = await generation_service.renderReport(
extractedContent=enhancedContent,
outputFormat=outputFormat,
title=title,
userPrompt=prompt,
aiService=self
)
# Generate meaningful filename (use AI-provided if valid, else fallback)
from datetime import datetime, UTC
timestamp = datetime.now(UTC).strftime("%Y%m%d-%H%M%S")
if parsedFilename and parsedFilename.lower().endswith(f".{outputFormat.lower()}"):
filename = parsedFilename
else:
safeTitle = ''.join(c if c.isalnum() else '-' for c in (title or 'document')).strip('-')
filename = f"{safeTitle or 'document'}-{timestamp}.{outputFormat}"
# Update progress - generation completed
progressLogger.updateProgress(operationId, 0.9, "Rendering")
result = {
"success": True,
"content": aiResponseJson, # Structured JSON document
"rendered_content": renderedContent, # Formatted content
"mime_type": mimeType,
"filename": filename,
"format": outputFormat,
"title": title,
"documents": [{
"documentName": filename,
"documentData": renderedContent,
"mimeType": mimeType
}],
"is_multi_file": False
}
# Complete progress tracking
progressLogger.completeOperation(operationId, True)
return result
except Exception as e:
logger.error(f"Error in single-file document generation: {str(e)}")
# Complete progress tracking with failure
progressLogger.completeOperation(operationId, False)
raise
async def _callAiWithMultiFileGeneration(
self,
prompt: str,
documents: Optional[List[ChatDocument]],
options: AiCallOptions,
outputFormat: str,
title: Optional[str],
prompt_analysis: Dict[str, Any]
) -> Dict[str, Any]:
"""Handle multi-file document generation using AI analysis."""
import time
# Create progress logger
workflow = self.services.currentWorkflow
progressLogger = self.services.workflow.createProgressLogger(workflow)
operationId = f"docGen_{workflow.id}_{int(time.time())}"
try:
# Start progress tracking
progressLogger.startOperation(
operationId,
"Generate",
"Multi-file Generation",
f"Processing {len(documents) if documents else 0} documents"
)
# Get multi-file extraction prompt based on AI analysis
from modules.services.serviceGeneration.mainServiceGeneration import GenerationService
generation_service = GenerationService(self.services)
# Use default title if not provided
if not title:
title = "AI Generated Documents"
# Update progress - generating extraction prompt
progressLogger.updateProgress(operationId, 0.1, "Generating prompt")
# Get adaptive extraction prompt
extraction_prompt = await generation_service.getAdaptiveExtractionPrompt(
outputFormat=outputFormat,
userPrompt=prompt,
title=title,
promptAnalysis=prompt_analysis,
aiService=self
)
logger.info(f"Adaptive extraction prompt length: {len(extraction_prompt)} characters")
logger.debug(f"Adaptive extraction prompt preview: {extraction_prompt[:500]}...")
# Update progress - starting document processing
progressLogger.updateProgress(operationId, 0.2, "Processing docs")
# Process with adaptive JSON schema - use the existing pipeline but with adaptive prompt
logger.info(f"Using adaptive prompt with existing pipeline: {len(extraction_prompt)} chars")
logger.debug(f"Processing documents: {len(documents) if documents else 0} documents")
# Use the existing pipeline but replace the prompt with our adaptive one
# This ensures proper document processing while using the multi-file prompt
ai_response = await self.documentProcessor.processDocumentsPerChunkJsonWithPrompt(documents, extraction_prompt, options)
logger.info(f"AI response type: {type(ai_response)}")
logger.info(f"AI response keys: {list(ai_response.keys()) if isinstance(ai_response, dict) else 'Not a dict'}")
logger.debug(f"AI response preview: {str(ai_response)[:500]}...")
# Validate response structure
if not self._validateResponseStructure(ai_response, prompt_analysis):
# Fallback to single-file if multi-file fails
logger.warning(f"Multi-file processing failed - Invalid response structure. Expected multi-file but got: {list(ai_response.keys()) if isinstance(ai_response, dict) else type(ai_response)}")
logger.warning(f"Prompt analysis: {prompt_analysis}")
logger.warning("Falling back to single-file generation")
return await self._callAiWithSingleFileGeneration(
prompt, documents, options, outputFormat, title
)
# Emit raw extracted data as a chat message attachment before transformation/rendering
try:
await self._postRawDataChatMessage(ai_response, label="raw_extraction_multi")
except Exception:
logger.warning("Failed to emit raw extraction chat message (multi-file)")
# Process multiple documents
generated_documents = []
for i, doc_data in enumerate(ai_response.get("documents", [])):
# Transform AI-generated sections to renderer-compatible format
transformed_sections = []
for section in doc_data.get("sections", []):
# Convert AI format to renderer format
transformed_section = {
"id": section.get("id", f"section_{len(transformed_sections) + 1}"),
"content_type": section.get("content_type", "paragraph"),
"elements": section.get("elements", []),
"order": section.get("order", len(transformed_sections) + 1)
}
# Extract text from elements for simple text-based sections
if section.get("content_type") in ["paragraph", "heading"]:
text_parts = []
for element in section.get("elements", []):
if "text" in element:
text_parts.append(element["text"])
# Add text to the first element or create a new one
if transformed_section["elements"]:
transformed_section["elements"][0]["text"] = "\n".join(text_parts)
else:
transformed_section["elements"] = [{"text": "\n".join(text_parts)}]
transformed_sections.append(transformed_section)
# Create complete document structure for rendering
complete_document = {
"metadata": {
"title": doc_data["title"],
"source_document": "multi_file_generation",
"document_id": doc_data.get("id", f"doc_{i+1}"),
"filename": doc_data.get("filename", f"document_{i+1}"),
"split_strategy": prompt_analysis.get("strategy", "custom")
},
"sections": transformed_sections,
"summary": f"Generated document: {doc_data['title']}",
"tags": ["multi_file", "ai_generated"]
}
# Use AI generation to enhance the extracted JSON before rendering
enhancedContent = complete_document # Default to original
if prompt:
try:
from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationType
# Get generation prompt
generationPrompt = await generation_service.getGenerationPrompt(
outputFormat=outputFormat,
userPrompt=prompt,
title=doc_data["title"],
aiService=self
)
# Prepare the AI call
request_options = AiCallOptions()
request_options.operationType = OperationType.GENERAL
# Create context with the extracted JSON content
import json
context = f"Extracted JSON content:\n{json.dumps(complete_document, indent=2)}"
request = AiCallRequest(
prompt=generationPrompt,
context=context,
options=request_options
)
# Call AI to enhance the content
response = await self.aiObjects.call(request)
# Save generation prompt and response to debug
try:
from modules.shared.debugLogger import writeDebugFile
debugData = {
"output_format": outputFormat,
"title": doc_data["title"],
"document_index": i,
"context_length": len(context),
"extracted_content_keys": list(complete_document.keys()) if isinstance(complete_document, dict) else []
}
writeDebugFile(generationPrompt, f"generation_multi_doc_{i}", debugData)
writeDebugFile(response.content or '', f"generation_multi_doc_{i}_response")
except Exception:
pass
if response and response.content:
# Parse the AI response as JSON
try:
import re
result = response.content.strip()
# Extract JSON from markdown if present
json_match = re.search(r'```json\s*\n(.*?)\n```', result, re.DOTALL)
if json_match:
result = json_match.group(1).strip()
elif result.startswith('```json'):
result = re.sub(r'^```json\s*', '', result)
result = re.sub(r'\s*```$', '', result)
elif result.startswith('```'):
result = re.sub(r'^```\s*', '', result)
result = re.sub(r'\s*```$', '', result)
# Try to parse JSON
enhancedContent = json.loads(result)
logger.info(f"AI enhanced JSON content successfully")
except json.JSONDecodeError as e:
logger.warning(f"AI generation returned invalid JSON: {str(e)}, attempting to repair...")
# Try to repair common JSON issues
try:
repaired_result = self._repairJson(result)
enhancedContent = json.loads(repaired_result)
logger.info(f"Successfully repaired JSON content")
except (json.JSONDecodeError, Exception) as repair_error:
logger.warning(f"JSON repair failed: {str(repair_error)}, trying AI repair...")
# Try AI-powered JSON repair as last resort
try:
ai_repaired = await self._repairJsonWithAI(result)
enhancedContent = json.loads(ai_repaired)
logger.info(f"AI successfully repaired JSON content")
except Exception as ai_repair_error:
logger.warning(f"AI JSON repair also failed: {str(ai_repair_error)}, using original content")
enhancedContent = complete_document
else:
logger.warning("AI generation returned empty response, using original content")
enhancedContent = complete_document
except Exception as e:
logger.warning(f"AI generation failed: {str(e)}, using original content")
enhancedContent = complete_document
# Render the enhanced JSON content
rendered_content, mime_type = await generation_service.renderReport(
extractedContent=enhancedContent,
outputFormat=outputFormat,
title=doc_data["title"],
userPrompt=prompt,
aiService=self
)
# Generate proper filename with correct extension
base_filename = doc_data.get("filename", f"document_{i+1}")
# Remove any existing extension and add the correct one
if '.' in base_filename:
base_filename = base_filename.rsplit('.', 1)[0]
# Add proper extension based on output format
if outputFormat.lower() == "docx":
filename = f"{base_filename}.docx"
elif outputFormat.lower() == "pdf":
filename = f"{base_filename}.pdf"
elif outputFormat.lower() == "html":
filename = f"{base_filename}.html"
else:
filename = f"{base_filename}.{outputFormat}"
generated_documents.append({
"documentName": filename,
"documentData": rendered_content,
"mimeType": mime_type
})
# Save debug files for multi-file generation - only if debug enabled
debug_enabled = self.services.utils.configGet("APP_DEBUG_CHAT_WORKFLOW_ENABLED", False)
if debug_enabled:
try:
import os
from datetime import datetime, UTC
ts = datetime.now(UTC).strftime("%Y%m%d-%H%M%S")
debug_root = "./test-chat/ai"
debug_dir = os.path.join(debug_root, f"multifile_output_{ts}")
os.makedirs(debug_dir, exist_ok=True)
# Save metadata
with open(os.path.join(debug_dir, "metadata.txt"), "w", encoding="utf-8") as f:
f.write(f"title: {title}\n")
f.write(f"format: {outputFormat}\n")
f.write(f"documents_count: {len(generated_documents)}\n")
f.write(f"split_strategy: {prompt_analysis.get('strategy', 'custom')}\n")
f.write(f"prompt_analysis: {prompt_analysis}\n")
# Save each generated document
for i, doc in enumerate(generated_documents):
doc_filename = doc["documentName"]
doc_data = doc["documentData"]
doc_mime = doc["mimeType"]
# Determine file extension
if outputFormat.lower() == "docx":
file_ext = ".docx"
elif outputFormat.lower() == "pdf":
file_ext = ".pdf"
elif outputFormat.lower() == "html":
file_ext = ".html"
else:
file_ext = f".{outputFormat}"
# Save the rendered document
output_path = os.path.join(debug_dir, f"document_{i+1}_{doc_filename}")
if file_ext in ['.md', '.txt', '.html', '.json', '.csv']:
# Text-based formats
with open(output_path, 'w', encoding='utf-8') as f:
f.write(doc_data)
else:
# Binary formats - decode from base64 if needed
try:
import base64
doc_bytes = base64.b64decode(doc_data)
with open(output_path, 'wb') as f:
f.write(doc_bytes)
except Exception:
# If not base64, save as text
with open(output_path, 'w', encoding='utf-8') as f:
f.write(doc_data)
logger.info(f"💾 Debug: Saved multi-file document {i+1}: {output_path}")
logger.info(f"💾 Debug: Multi-file output saved to: {debug_dir}")
except Exception as e:
logger.warning(f"Failed to save multi-file debug output: {e}")
# Update progress - generation completed
progressLogger.updateProgress(operationId, 0.9, "Rendering")
result = {
"success": True,
"content": ai_response,
"rendered_content": None, # Not applicable for multi-file
"mime_type": None, # Not applicable for multi-file
"filename": None, # Not applicable for multi-file
"format": outputFormat,
"title": title,
"documents": generated_documents,
"is_multi_file": True,
"split_strategy": prompt_analysis.get("strategy", "custom")
}
# Complete progress tracking
progressLogger.completeOperation(operationId, True)
return result
except Exception as e:
logger.error(f"Error in multi-file document generation: {str(e)}")
# Complete progress tracking with failure
progressLogger.completeOperation(operationId, False)
# Fallback to single-file
return await self._callAiWithSingleFileGeneration(
prompt, documents, options, outputFormat, title
)
async def _callAiJson(
self,
prompt: str,
documents: Optional[List[ChatDocument]],
options: AiCallOptions
) -> Dict[str, Any]:
"""
Handle AI calls with document processing for JSON output.
Returns structured JSON document instead of text.
"""
# Process documents with JSON merging
return await self.documentProcessor.processDocumentsPerChunkJson(documents, prompt, options)
async def _analyzePromptIntent(self, prompt: str, ai_service=None) -> Dict[str, Any]:
"""Use AI to analyze user prompt and determine processing requirements."""
if not ai_service:
return {"is_multi_file": False, "strategy": "single", "criteria": None}
try:
analysis_prompt = f"""
Analyze this user request and determine if it requires multiple file output or single file output.
User request: "{prompt}"
Respond with JSON only in this exact format:
{{
"is_multi_file": true/false,
"strategy": "single|per_entity|by_section|by_criteria|custom",
"criteria": "description of how to split content",
"file_naming_pattern": "suggested pattern for filenames",
"reasoning": "brief explanation of the analysis"
}}
Consider:
- Does the user want separate files for different entities (customers, products, etc.)?
- Does the user want to split content into multiple documents?
- What would be the most logical way to organize the content?
- What language is the request in? (analyze in the original language)
Return only the JSON response.
"""
from modules.datamodels.datamodelAi import AiCallRequest, AiCallOptions, OperationType
request_options = AiCallOptions()
request_options.operationType = OperationType.GENERAL
request = AiCallRequest(prompt=analysis_prompt, context="", options=request_options)
response = await ai_service.aiObjects.call(request)
if response and response.content:
import json
import re
# Extract JSON from response
result = response.content.strip()
json_match = re.search(r'\{.*\}', result, re.DOTALL)
if json_match:
result = json_match.group(0)
analysis = json.loads(result)
return analysis
else:
return {"is_multi_file": False, "strategy": "single", "criteria": None}
except Exception as e:
logger.warning(f"AI prompt analysis failed: {str(e)}, defaulting to single file")
return {"is_multi_file": False, "strategy": "single", "criteria": None}
def _validateResponseStructure(self, response: Dict[str, Any], prompt_analysis: Dict[str, Any]) -> bool:
"""Validate that AI response matches the expected structure."""
try:
if not isinstance(response, dict):
logger.warning(f"Response validation failed: Response is not a dict, got {type(response)}")
return False
# Check for multi-file structure
if prompt_analysis.get("is_multi_file", False):
has_documents = "documents" in response
is_documents_list = isinstance(response.get("documents"), list)
logger.info(f"Multi-file validation: has_documents={has_documents}, is_documents_list={is_documents_list}")
if has_documents and is_documents_list:
logger.info(f"Multi-file validation passed: {len(response['documents'])} documents found")
else:
logger.warning(f"Multi-file validation failed: documents key present={has_documents}, documents is list={is_documents_list}")
logger.warning(f"Available keys: {list(response.keys())}")
return has_documents and is_documents_list
else:
has_sections = "sections" in response
is_sections_list = isinstance(response.get("sections"), list)
logger.info(f"Single-file validation: has_sections={has_sections}, is_sections_list={is_sections_list}")
return has_sections and is_sections_list
except Exception as e:
logger.warning(f"Response validation failed with exception: {str(e)}")
return False
async def _postRawDataChatMessage(self, payload: Any, label: str = "raw_extraction") -> None:
"""
Create a ChatMessage with the extracted raw JSON attached as a file so the user
has access to the data even if downstream processing fails.
"""
try:
services = self.services
workflow = services.currentWorkflow
# Serialize payload
import json as _json
from datetime import datetime, UTC
ts = datetime.now(UTC).strftime("%Y%m%d-%H%M%S")
content_text = _json.dumps(payload, ensure_ascii=False, indent=2)
content_bytes = content_text.encode('utf-8')
# Store as file via component storage
file_name = f"{label}_{ts}.json"
file_item = services.interfaceDbComponent.createFile(
name=file_name,
mimeType="application/json",
content=content_bytes
)
services.interfaceDbComponent.createFileData(file_item.id, content_bytes)
# Lookup file info for ChatDocument
file_info = services.workflow.getFileInfo(file_item.id)
doc = ChatDocument(
messageId="", # set after message creation
fileId=file_item.id,
fileName=file_info.get("fileName", file_name) if file_info else file_name,
fileSize=file_info.get("size", len(content_bytes)) if file_info else len(content_bytes),
mimeType=file_info.get("mimeType", "application/json") if file_info else "application/json"
)
# Create message referencing the file - include document in initial call
messageData = {
"workflowId": workflow.id,
"role": "assistant",
"message": "Raw extraction data saved",
"status": "data",
"sequenceNr": len(getattr(workflow, 'messages', []) or []) + 1,
"publishedAt": services.utils.getUtcTimestamp(),
"documentsLabel": label,
"documents": []
}
# Store message with document included from the start
services.workflow.storeMessageWithDocuments(services.workflow.workflow, messageData, [doc])
except Exception:
# Non-fatal; ignore if storage or chat creation fails
return
def _repairJson(self, json_string: str) -> str:
"""Repair common JSON syntax errors efficiently for large JSON."""
try:
import re
import json
# Remove any leading/trailing whitespace
json_string = json_string.strip()
# For large JSON, skip substring extraction and go straight to targeted repairs
logger.info(f"Attempting JSON repair for {len(json_string)} characters...")
# Try to parse first to see what specific error we get
try:
json.loads(json_string)
return json_string # Already valid
except json.JSONDecodeError as e:
error_msg = str(e)
logger.info(f"JSON error: {error_msg}")
# Apply targeted fixes based on the specific error
if "Expecting ',' delimiter" in error_msg:
# Fix missing commas between array elements
json_string = re.sub(r'\]\s*\[', '], [', json_string)
json_string = re.sub(r'\}\s*\{', '}, {', json_string)
# Fix missing commas between object properties
json_string = re.sub(r'("\s*:\s*[^,}]+)\s*(")', r'\1, \2', json_string)
if "Expecting value" in error_msg:
# Fix missing values (replace empty with null)
json_string = re.sub(r':\s*,', ': null,', json_string)
json_string = re.sub(r':\s*}', ': null}', json_string)
if "Expecting property name" in error_msg:
# Fix unquoted property names
json_string = re.sub(r'(\w+):', r'"\1":', json_string)
# Fix trailing commas before closing brackets/braces
json_string = re.sub(r',(\s*[}\]])', r'\1', json_string)
# Fix missing closing brackets/braces (only if reasonable)
open_braces = json_string.count('{')
close_braces = json_string.count('}')
open_brackets = json_string.count('[')
close_brackets = json_string.count(']')
# Only add missing brackets if the difference is small (avoid runaway)
if 0 < (open_braces - close_braces) <= 5:
missing_braces = open_braces - close_braces
json_string += '}' * missing_braces
if 0 < (open_brackets - close_brackets) <= 5:
missing_brackets = open_brackets - close_brackets
json_string += ']' * missing_brackets
# Try to parse again
try:
json.loads(json_string)
logger.info("JSON repair successful")
return json_string
except json.JSONDecodeError:
logger.warning("JSON repair failed - will try AI repair")
return json_string
except Exception as e:
logger.warning(f"JSON repair failed: {str(e)}")
return json_string
async def _repairJsonWithAI(self, malformed_json: str) -> str:
"""Use AI to repair malformed JSON efficiently for large files."""
try:
# Limit JSON size for AI processing (max 50KB to avoid token limits)
max_json_size = 50000
json_to_repair = malformed_json
if len(malformed_json) > max_json_size:
logger.warning(f"JSON too large ({len(malformed_json)} chars), truncating to {max_json_size} chars for AI repair")
# Try to find a good truncation point (end of a complete object/array)
truncate_at = max_json_size
for i in range(max_json_size, max(0, max_json_size - 1000), -1):
if malformed_json[i] in ['}', ']']:
truncate_at = i + 1
break
json_to_repair = malformed_json[:truncate_at] + "..."
repair_prompt = f"""
You are a JSON repair expert. Fix the following malformed JSON and return ONLY the corrected JSON, no explanations.
Malformed JSON:
{json_to_repair}
Return only the valid JSON:
"""
# Use AI to repair the JSON
repaired_json = await self.services.ai.callAi(
prompt=repair_prompt,
documents=None,
options={
"process_type": "text",
"operation_type": "generate_content",
"priority": "speed",
"max_cost": 0.01
}
)
# Clean up the response (remove any markdown formatting)
repaired_json = repaired_json.strip()
if repaired_json.startswith('```json'):
repaired_json = repaired_json[7:]
if repaired_json.endswith('```'):
repaired_json = repaired_json[:-3]
repaired_json = repaired_json.strip()
# Validate the repaired JSON
import json
json.loads(repaired_json)
logger.info("AI JSON repair successful")
return repaired_json
except Exception as e:
logger.warning(f"AI JSON repair failed: {str(e)}")
return malformed_json
def _attemptJsonFix(self, json_string: str) -> str:
"""Attempt to fix common JSON issues"""
try:
# Remove any trailing commas before closing braces/brackets
import re
fixed = re.sub(r',(\s*[}\]])', r'\1', json_string)
# Try to fix unterminated strings by adding quotes at the end
if '"' in fixed and not fixed.strip().endswith('"'):
# Count quotes to see if we have an odd number (unterminated string)
quote_count = fixed.count('"')
if quote_count % 2 == 1:
# Find the last quote and add a closing quote
last_quote_pos = fixed.rfind('"')
if last_quote_pos != -1:
# Check if there's content after the last quote that needs to be quoted
after_quote = fixed[last_quote_pos + 1:].strip()
if after_quote and not after_quote.startswith(','):
# Add closing quote before any trailing content
fixed = fixed[:last_quote_pos + 1] + '"' + after_quote
return fixed
except Exception:
return json_string