gateway/modules/workflows/processing/shared/placeholderFactory.py
2025-10-14 14:47:50 +02:00

412 lines
20 KiB
Python

"""
Placeholder Factory
Centralized placeholder extraction functions for all workflow modes.
Each function corresponds to a {{KEY:PLACEHOLDER_NAME}} in prompt templates.
NAMING CONVENTION:
- All functions follow pattern: extract{PlaceholderName}()
- Placeholder names are in UPPER_CASE with underscores
- Function names are in camelCase
MAPPING TABLE (keys → function) with usage [taskplan | actionplan | react]:
{{KEY:USER_PROMPT}} -> extractUserPrompt() [taskplan, actionplan, react]
{{KEY:USER_LANGUAGE}} -> extractUserLanguage() [actionplan, react]
{{KEY:WORKFLOW_HISTORY}} -> extractWorkflowHistory() [taskplan, actionplan, react]
{{KEY:AVAILABLE_CONNECTIONS_INDEX}} -> extractAvailableConnectionsIndex() [actionplan, react]
{{KEY:AVAILABLE_CONNECTIONS_SUMMARY}} -> extractAvailableConnectionsSummary() []
{{KEY:AVAILABLE_DOCUMENTS_SUMMARY}} -> extractAvailableDocumentsSummary() [taskplan, actionplan, react]
{{KEY:AVAILABLE_DOCUMENTS_INDEX}} -> extractAvailableDocumentsIndex() [react]
{{KEY:AVAILABLE_METHODS}} -> extractAvailableMethods() [actionplan, react]
{{KEY:REVIEW_CONTENT}} -> extractReviewContent() [actionplan, react]
{{KEY:PREVIOUS_ACTION_RESULTS}} -> extractPreviousActionResults() [react]
{{KEY:LEARNINGS_AND_IMPROVEMENTS}} -> extractLearningsAndImprovements() [react]
{{KEY:LATEST_REFINEMENT_FEEDBACK}} -> extractLatestRefinementFeedback() [react]
Following placeholders are populated directly by prompt builders with according context in promptGenerationActionsReact module:
- ACTION_OBJECTIVE,
- SELECTED_ACTION,
- ACTION_SIGNATURE
"""
import json
import logging
from typing import Dict, Any, List
from modules.datamodels.datamodelChat import ChatDocument
logger = logging.getLogger(__name__)
from modules.workflows.processing.shared.methodDiscovery import (methods, discoverMethods)
def extractUserPrompt(context: Any) -> str:
"""Extract user prompt from context. Maps to {{KEY:USER_PROMPT}}.
Prefer the cleaned intent stored on the services object if available via context.
Fallback to the task_step objective.
"""
try:
# Prefer services.currentUserPrompt when accessible through context
services = getattr(context, 'services', None)
if services and getattr(services, 'currentUserPrompt', None):
return services.currentUserPrompt
except Exception:
pass
if hasattr(context, 'task_step') and context.task_step:
return context.task_step.objective or 'No request specified'
return 'No request specified'
def extractWorkflowHistory(service: Any, context: Any) -> str:
"""Extract workflow history from context. Maps to {{KEY:WORKFLOW_HISTORY}}
Reverse-chronological, enriched with message summaries and document labels.
"""
# Prefer explicit workflow on context; else fall back to services.workflow
workflow = None
try:
if hasattr(context, 'workflow') and context.workflow:
workflow = context.workflow
elif hasattr(service, 'workflow') and service.workflow:
workflow = service.workflow
except Exception:
workflow = None
if workflow:
history = getPreviousRoundContext(service, workflow)
return history or "No previous workflow rounds available"
return "No previous workflow rounds available"
def extractAvailableMethods(service: Any) -> str:
"""Extract available methods for action planning. Maps to {{KEY:AVAILABLE_METHODS}}"""
try:
# Get the methods dictionary directly from the global methods variable
if not methods:
discoverMethods(service)
# Create a flat JSON format with compound action names for better AI parsing
available_actions_json = {}
for methodName, methodInfo in methods.items():
# Convert MethodAi -> ai, MethodDocument -> document, etc.
shortName = methodName.replace('Method', '').lower()
for actionName, actionInfo in methodInfo['actions'].items():
# Create compound action name: method.action
compoundActionName = f"{shortName}.{actionName}"
# Get the action description
action_description = actionInfo.get('description', f"Execute {actionName} action")
available_actions_json[compoundActionName] = action_description
return json.dumps(available_actions_json, indent=2, ensure_ascii=False)
except Exception as e:
logger.error(f"Error extracting available methods: {str(e)}")
return json.dumps({}, indent=2, ensure_ascii=False)
def extractUserLanguage(service: Any) -> str:
"""Extract user language from service. Maps to {{KEY:USER_LANGUAGE}}"""
return service.user.language if service and service.user else 'en'
def _computeMessageSummary(msg) -> str:
"""Create a concise summary for a ChatMessage with documents only.
Fields: documentCount, roundNumber, documentsLabel, document names, message (full), success flag.
"""
try:
docs = getattr(msg, 'documents', []) or []
if not docs:
return "" # Only summarize messages that contain documents
document_count = len(docs)
round_number = getattr(msg, 'roundNumber', None) or 0
label = getattr(msg, 'documentsLabel', None) or ""
# Collect ALL document names (supports ChatDocument objects and dicts)
doc_names = []
for d in docs:
name = None
try:
if isinstance(d, dict):
# For dict objects, try multiple possible field names
name = d.get('fileName') or d.get('documentName') or d.get('name') or d.get('filename')
else:
# For ChatDocument objects, use fileName field
name = getattr(d, 'fileName', None) or getattr(d, 'documentName', None) or getattr(d, 'name', None) or getattr(d, 'filename', None)
except Exception:
name = None
doc_names.append(name or "(unnamed)")
# Format document names in brackets
if doc_names:
names_part = f"({', '.join(doc_names)})"
else:
names_part = "(no documents)"
# Don't truncate the message - show full content
user_message = (getattr(msg, 'message', '') or '').strip().replace("\n", " ")
# Read success from ChatMessage.success field
success_flag = getattr(msg, 'success', None)
success_text = "success=True" if success_flag is True else ("success=False" if success_flag is False else "success=Unknown")
label_part = f" label='{label}'" if label else ""
# Add learning/feedback if available
learning_part = ""
if hasattr(msg, 'summary') and msg.summary and 'learnings' in msg.summary.lower():
learning_part = " | learnings available"
return f"Round {round_number}: {document_count} docs {names_part}{label_part} | {success_text}{learning_part} | msg='{user_message}'"
except Exception:
return ""
def getMessageSummary(msg) -> str:
"""Return existing ChatMessage.summary or compute, set, and return it (documents only)."""
try:
if getattr(msg, 'summary', None):
return msg.summary
summary = _computeMessageSummary(msg)
# Persist in-memory only; caller can store if desired
if summary:
try:
setattr(msg, 'summary', summary)
except Exception:
pass
return summary
except Exception:
return ""
def getPreviousRoundContext(services, workflow: Any) -> str:
"""Get enriched context:
- Reverse-chronological ordering
- Current round first (newest → oldest), then older rounds
- Only messages with documents summarized
- Include available documents snapshot at end
"""
try:
if not workflow:
return "No previous round context available"
lines: List[str] = []
# Reverse-chronological, current round first
try:
msgs = getattr(workflow, 'messages', []) or []
current_round = getattr(workflow, 'currentRound', None)
current_round_msgs: List[Any] = []
previous_round_msgs: List[Any] = []
for m in msgs:
if current_round is not None and getattr(m, 'roundNumber', None) == current_round:
current_round_msgs.append(m)
else:
previous_round_msgs.append(m)
for m in reversed(current_round_msgs):
s = getMessageSummary(m)
if s:
lines.append(f"- {s}")
for m in reversed(previous_round_msgs):
s = getMessageSummary(m)
if s:
lines.append(f"- {s}")
except Exception:
pass
# Include available documents snapshot at end
try:
if hasattr(services, 'workflow'):
docs_index = services.workflow.getAvailableDocuments(workflow)
if docs_index and docs_index != "No documents available":
doc_count = docs_index.count("docItem:") # Only count actual documents, not document list labels
lines.append(f"Available documents: {doc_count}")
except Exception:
pass
if not lines:
return "No previous round context available"
return "\n".join(lines)
except Exception as e:
logger.error(f"Error getting previous round context: {str(e)}")
return "Error retrieving previous round context"
def extractReviewContent(context: Any) -> str:
"""Extract review content for result validation. Maps to {{KEY:REVIEW_CONTENT}}"""
try:
if hasattr(context, 'action_results') and context.action_results:
# Build result summary
result_summary = ""
for i, result in enumerate(context.action_results):
result_summary += f"\nRESULT {i+1}:\n"
result_summary += f" Success: {result.success}\n"
if result.error:
result_summary += f" Error: {result.error}\n"
if result.documents:
result_summary += f" Documents: {len(result.documents)} document(s)\n"
for doc in result.documents:
# Extract all available metadata without content
doc_metadata = {
"name": getattr(doc, 'fileName', None) or getattr(doc, 'documentName', 'Unknown'),
"mimeType": getattr(doc, 'mimeType', 'Unknown'),
"size": getattr(doc, 'size', 'Unknown'),
"created": getattr(doc, 'created', 'Unknown'),
"modified": getattr(doc, 'modified', 'Unknown'),
"typeGroup": getattr(doc, 'typeGroup', 'Unknown'),
"documentId": getattr(doc, 'documentId', 'Unknown'),
"reference": getattr(doc, 'reference', 'Unknown')
}
# Remove 'Unknown' values to keep it clean
doc_metadata = {k: v for k, v in doc_metadata.items() if v != 'Unknown'}
result_summary += f" - {json.dumps(doc_metadata, indent=6, ensure_ascii=False)}\n"
else:
result_summary += f" Documents: None\n"
return result_summary
elif hasattr(context, 'observation') and context.observation:
# For observation data, show full content but handle documents specially
if isinstance(context.observation, dict):
# Create a copy to modify
obs_copy = context.observation.copy()
# If there are previews with documents, show only metadata
if 'previews' in obs_copy and isinstance(obs_copy['previews'], list):
for preview in obs_copy['previews']:
if isinstance(preview, dict) and 'snippet' in preview:
# Replace snippet with metadata indicator
preview['snippet'] = f"[Content: {len(preview.get('snippet', ''))} characters]"
return json.dumps(obs_copy, indent=2, ensure_ascii=False)
else:
return json.dumps(context.observation, ensure_ascii=False)
elif hasattr(context, 'step_result') and context.step_result and 'observation' in context.step_result:
# For observation data in step_result, show full content but handle documents specially
observation = context.step_result['observation']
if isinstance(observation, dict):
# Create a copy to modify
obs_copy = observation.copy()
# If there are previews with documents, show only metadata
if 'previews' in obs_copy and isinstance(obs_copy['previews'], list):
for preview in obs_copy['previews']:
if isinstance(preview, dict) and 'snippet' in preview:
# Replace snippet with metadata indicator
preview['snippet'] = f"[Content: {len(preview.get('snippet', ''))} characters]"
return json.dumps(obs_copy, indent=2, ensure_ascii=False)
else:
return json.dumps(observation, ensure_ascii=False)
else:
return "No review content available"
except Exception as e:
logger.error(f"Error extracting review content: {str(e)}")
return "No review content available"
def extractPreviousActionResults(context: Any) -> str:
"""Extract previous action results for learning context. Maps to {{KEY:PREVIOUS_ACTION_RESULTS}}"""
try:
if not hasattr(context, 'previous_action_results') or not context.previous_action_results:
return "No previous actions executed yet"
results = []
for i, result in enumerate(context.previous_action_results[-5:], 1): # Last 5 results
if hasattr(result, 'resultLabel') and hasattr(result, 'status'):
status = "SUCCESS" if result.status == "completed" else "FAILED"
results.append(f"Action {i}: {result.resultLabel} - {status}")
if hasattr(result, 'error') and result.error:
results.append(f" Error: {result.error}")
return "\n".join(results) if results else "No previous actions executed yet"
except Exception as e:
logger.error(f"Error extracting previous action results: {str(e)}")
return "No previous actions executed yet"
def extractLearningsAndImprovements(context: Any) -> str:
"""Extract learnings and improvements from previous actions. Maps to {{KEY:LEARNINGS_AND_IMPROVEMENTS}}"""
try:
learnings = []
# Get improvements from context
if hasattr(context, 'improvements') and context.improvements and isinstance(context.improvements, list):
learnings.append("IMPROVEMENTS:")
for improvement in context.improvements[-3:]: # Last 3 improvements
learnings.append(f"- {improvement}")
# Get failure patterns
if hasattr(context, 'failure_patterns') and context.failure_patterns and isinstance(context.failure_patterns, list):
learnings.append("FAILURE PATTERNS TO AVOID:")
for pattern in context.failure_patterns[-3:]: # Last 3 patterns
learnings.append(f"- {pattern}")
# Get successful actions
if hasattr(context, 'successful_actions') and context.successful_actions and isinstance(context.successful_actions, list):
learnings.append("SUCCESSFUL APPROACHES:")
for action in context.successful_actions[-3:]: # Last 3 successful
learnings.append(f"- {action}")
return "\n".join(learnings) if learnings else "No learnings available yet"
except Exception as e:
logger.error(f"Error extracting learnings and improvements: {str(e)}")
return "No learnings available yet"
def extractLatestRefinementFeedback(context: Any) -> str:
"""Extract the latest refinement feedback. Maps to {{KEY:LATEST_REFINEMENT_FEEDBACK}}"""
try:
if not hasattr(context, 'previous_review_result') or not context.previous_review_result or not isinstance(context.previous_review_result, list):
return "No previous refinement feedback available"
# Get the most recent refinement decision
latest_decision = context.previous_review_result[-1]
if not isinstance(latest_decision, dict):
return "No previous refinement feedback available"
feedback_parts = []
# Add decision and reason
decision = latest_decision.get('decision', 'unknown')
reason = latest_decision.get('reason', 'No reason provided')
feedback_parts.append(f"Latest Decision: {decision}")
feedback_parts.append(f"Reason: {reason}")
# Add any specific feedback or suggestions
if 'feedback' in latest_decision:
feedback_parts.append(f"Feedback: {latest_decision['feedback']}")
if 'suggestions' in latest_decision:
feedback_parts.append(f"Suggestions: {latest_decision['suggestions']}")
return "\n".join(feedback_parts)
except Exception as e:
logger.error(f"Error extracting latest refinement feedback: {str(e)}")
return "No previous refinement feedback available"
def extractAvailableDocumentsSummary(service: Any, context: Any) -> str:
"""Summary of available documents (count only)."""
try:
documents = service.workflow.getAvailableDocuments(context.workflow)
if documents and documents != "No documents available":
# Count only actual documents, not list labels
doc_count = documents.count("docItem:")
return f"{doc_count} documents available from previous tasks"
return "No documents available"
except Exception as e:
logger.error(f"Error getting document summary: {str(e)}")
return "No documents available"
def extractAvailableDocumentsIndex(service: Any, context: Any) -> str:
"""Index of available documents with detailed references for parameter generation."""
try:
return service.workflow.getAvailableDocuments(context.workflow)
except Exception as e:
logger.error(f"Error getting document index: {str(e)}")
return "No documents available"
def extractAvailableConnectionsSummary(service: Any) -> str:
"""Summary of available connections (count only)."""
try:
connections = service.workflow.getConnectionReferenceList()
if connections:
return f"{len(connections)} connections available"
return "No connections available"
except Exception as e:
logger.error(f"Error getting connection summary: {str(e)}")
return "No connections available"
def extractAvailableConnectionsIndex(service: Any) -> str:
"""Index of available connections with detailed references for parameter generation."""
try:
connections = service.workflow.getConnectionReferenceList()
if connections:
return '\n'.join(f"- {conn}" for conn in connections)
return "No connections available"
except Exception as e:
logger.error(f"Error getting connection index: {str(e)}")
return "No connections available"