AgentGraph / backend /services /reconstruction_service.py
wu981526092's picture
🚀 Deploy AgentGraph: Complete agent monitoring and knowledge graph system
c2ea5ed
"""
Reconstruction Service
This service handles all database operations for prompt reconstruction,
providing a clean interface between the database layer and the pure
reconstruction functions in agentgraph.reconstruction.
"""
import logging
from typing import Dict, List, Any, Optional
from sqlalchemy.orm import Session
from datetime import datetime, timezone
import traceback
from backend.database.models import PromptReconstruction, KnowledgeGraph
from backend.database.utils import get_knowledge_graph_by_id, get_knowledge_graph
from backend.database import get_db
from backend.services.task_service import update_task_status
logger = logging.getLogger(__name__)
class ReconstructionService:
"""
Service for orchestrating prompt reconstruction with database operations.
This service fetches data from the database, calls pure reconstruction functions
from agentgraph.reconstruction, and saves the results back to the database.
"""
def __init__(self, session: Session):
self.session = session
def fetch_reconstruction_data(self, kg_identifier: str) -> Dict[str, Any]:
"""
Fetch knowledge graph data needed for reconstruction from the database.
Args:
kg_identifier: Knowledge graph identifier (ID or filename)
Returns:
Dictionary containing knowledge graph data for reconstruction
"""
try:
# Try to load by ID first (if numeric), then by filename
kg = None
if str(kg_identifier).isdigit():
kg = get_knowledge_graph_by_id(self.session, kg_identifier)
if not kg:
kg = get_knowledge_graph(self.session, kg_identifier)
if not kg:
return {"error": f"Knowledge graph with identifier {kg_identifier} not found"}
# Extract the actual graph data
if hasattr(kg, 'graph_data'):
kg_data = kg.graph_data
else:
kg_data = kg
# Ensure the graph data has entities and relations
if not kg_data or 'entities' not in kg_data or 'relations' not in kg_data:
return {"error": f"Invalid knowledge graph data for {kg_identifier}"}
reconstruction_data = {
"knowledge_graph": kg_data,
"knowledge_graph_id": kg.id,
"knowledge_graph_filename": getattr(kg, 'filename', str(kg_identifier)),
"entities": {entity["id"]: entity for entity in kg_data["entities"]},
"relations": {relation["id"]: relation for relation in kg_data["relations"]}
}
logger.info(f"Successfully loaded knowledge graph {kg_identifier} with {len(kg_data['entities'])} entities and {len(kg_data['relations'])} relations")
return reconstruction_data
except Exception as e:
logger.error(f"Error loading knowledge graph {kg_identifier}: {repr(e)}")
return {"error": f"Failed to load knowledge graph: {repr(e)}"}
def save_prompt_reconstructions(
self,
kg_identifier: str,
reconstructed_relations: List[Dict[str, Any]]
) -> Dict[str, Any]:
"""
Save reconstructed prompts to the database.
Args:
kg_identifier: Knowledge graph identifier
reconstructed_relations: List of relations with reconstructed prompts
Returns:
Dictionary with save results and metadata
"""
try:
logger.info(f"Saving prompt reconstructions for knowledge graph: {kg_identifier}")
# Find the knowledge graph
kg = None
# First try direct lookup by ID
if str(kg_identifier).isdigit():
logger.info(f"Trying to find knowledge graph by ID: {kg_identifier}")
kg = self.session.query(KnowledgeGraph).filter_by(id=kg_identifier).first()
if kg:
logger.info(f"Found knowledge graph by ID {kg_identifier}")
# If not found by ID, try by filename
if not kg:
logger.info(f"Trying to find knowledge graph by filename: {kg_identifier}")
kg = self.session.query(KnowledgeGraph).filter_by(filename=kg_identifier).first()
if kg:
logger.info(f"Found knowledge graph by filename {kg_identifier}")
if not kg:
error_msg = f"Knowledge graph with identifier {kg_identifier} not found"
logger.error(error_msg)
return {"error": error_msg}
logger.info(f"Found knowledge graph: ID={kg.id}, filename={kg.filename}, status={kg.status}")
# Delete existing prompt reconstructions for this knowledge graph (if any)
existing_count = self.session.query(PromptReconstruction).filter_by(
knowledge_graph_id=kg.id
).count()
if existing_count > 0:
logger.info(f"Deleting {existing_count} existing prompt reconstructions")
self.session.query(PromptReconstruction).filter_by(
knowledge_graph_id=kg.id
).delete()
# Save new prompt reconstructions
saved_count = 0
for relation in reconstructed_relations:
pr = PromptReconstruction(
knowledge_graph_id=kg.id,
relation_id=relation["id"],
reconstructed_prompt=relation.get("prompt", ""),
dependencies=relation.get("dependencies", {}),
)
self.session.add(pr)
saved_count += 1
# Update the knowledge graph status to 'enriched'
kg.status = "enriched"
kg.updated_at = datetime.utcnow()
# Commit all changes
self.session.commit()
logger.info(f"Successfully saved {saved_count} prompt reconstructions for KG {kg_identifier}")
return {
"status": "success",
"knowledge_graph_id": kg.id,
"knowledge_graph_filename": kg.filename,
"saved_reconstructions": saved_count,
"replaced_existing": existing_count
}
except Exception as e:
logger.error(f"Error saving prompt reconstructions: {repr(e)}")
self.session.rollback()
return {"error": f"Failed to save prompt reconstructions: {repr(e)}"}
def run_prompt_reconstruction(
self,
kg_identifier: str,
output_identifier: str = None
) -> Dict[str, Any]:
"""
Run prompt reconstruction with database operations.
Args:
kg_identifier: Knowledge graph identifier to reconstruct
output_identifier: Optional output identifier (defaults to kg_identifier)
Returns:
Dictionary containing reconstruction results
"""
# Fetch data from database
reconstruction_data = self.fetch_reconstruction_data(kg_identifier)
if "error" in reconstruction_data:
return reconstruction_data
try:
# Import and call pure reconstruction function
from agentgraph.reconstruction import reconstruct_prompts_from_knowledge_graph
reconstructed_relations = reconstruct_prompts_from_knowledge_graph(
reconstruction_data["knowledge_graph"]
)
reconstruction_results = {
"reconstructed_relations": reconstructed_relations,
"metadata": {
"total_relations": len(reconstructed_relations),
"timestamp": datetime.utcnow().isoformat()
}
}
if "error" in reconstruction_results:
return reconstruction_results
# Save results to database
save_identifier = output_identifier or kg_identifier
save_results = self.save_prompt_reconstructions(
kg_identifier=save_identifier,
reconstructed_relations=reconstruction_results["reconstructed_relations"]
)
if "error" in save_results:
return save_results
# Combine results
final_results = {
"status": "success",
"knowledge_graph_id": reconstruction_data["knowledge_graph_id"],
"reconstruction_metadata": reconstruction_results["metadata"],
"save_results": save_results,
"total_relations_processed": len(reconstruction_results["reconstructed_relations"])
}
return final_results
except Exception as e:
logger.error(f"Error during prompt reconstruction: {repr(e)}")
return {"error": f"Reconstruction failed: {repr(e)}"}
def get_prompt_reconstructions(
self,
kg_identifier: str
) -> List[Dict[str, Any]]:
"""
Get saved prompt reconstructions from database.
Args:
kg_identifier: Knowledge graph identifier
Returns:
List of prompt reconstruction dictionaries
"""
try:
# Find the knowledge graph
kg = None
if str(kg_identifier).isdigit():
kg = self.session.query(KnowledgeGraph).filter_by(id=kg_identifier).first()
if not kg:
kg = self.session.query(KnowledgeGraph).filter_by(filename=kg_identifier).first()
if not kg:
logger.error(f"Knowledge graph with identifier {kg_identifier} not found")
return []
# Get prompt reconstructions
reconstructions = self.session.query(PromptReconstruction).filter_by(
knowledge_graph_id=kg.id
).all()
return [
{
"id": pr.id,
"relation_id": pr.relation_id,
"reconstructed_prompt": pr.reconstructed_prompt,
"dependencies": pr.dependencies,
"created_at": pr.created_at.isoformat() if pr.created_at else None
}
for pr in reconstructions
]
except Exception as e:
logger.error(f"Error retrieving prompt reconstructions: {repr(e)}")
return []
def get_reconstruction_summary(self, kg_identifier: str) -> Dict[str, Any]:
"""
Get a summary of reconstruction results for a knowledge graph.
Args:
kg_identifier: Knowledge graph identifier
Returns:
Dictionary containing summary statistics
"""
try:
reconstructions = self.get_prompt_reconstructions(kg_identifier)
if not reconstructions:
return {
"total_reconstructions": 0,
"knowledge_graph_identifier": kg_identifier,
"status": "no_reconstructions_found"
}
# Calculate summary statistics
total_reconstructions = len(reconstructions)
# Get relation types from dependencies if available
relation_types = set()
entity_count = 0
for recon in reconstructions:
deps = recon.get("dependencies", {})
entities = deps.get("entities", [])
relations = deps.get("relations", [])
entity_count += len(entities)
# This would require additional query to get relation types
# For now, just count unique relation IDs
relation_types.update(relations)
return {
"total_reconstructions": total_reconstructions,
"unique_relations_referenced": len(relation_types),
"total_entity_references": entity_count,
"knowledge_graph_identifier": kg_identifier,
"status": "reconstructions_available",
"latest_reconstruction": max(recon.get("created_at", "") for recon in reconstructions) if reconstructions else None
}
except Exception as e:
logger.error(f"Error generating reconstruction summary: {repr(e)}")
return {"error": f"Failed to generate summary: {repr(e)}"}
def reconstruct_single_relation(
self,
kg_identifier: str,
relation_id: str
) -> Dict[str, Any]:
"""
Reconstruct prompt for a single relation.
Args:
kg_identifier: Knowledge graph identifier
relation_id: Specific relation ID to reconstruct
Returns:
Dictionary containing reconstruction result for the single relation
"""
# Fetch data from database
reconstruction_data = self.fetch_reconstruction_data(kg_identifier)
if "error" in reconstruction_data:
return reconstruction_data
try:
# Import and call pure reconstruction function
from agentgraph.reconstruction import reconstruct_single_relation_prompt
result = reconstruct_single_relation_prompt(
knowledge_graph_data=reconstruction_data["knowledge_graph"],
relation_id=relation_id
)
return result
except Exception as e:
logger.error(f"Error reconstructing single relation {relation_id}: {repr(e)}")
return {"error": f"Single relation reconstruction failed: {repr(e)}"}
def enrich_knowledge_graph_with_prompts(
self,
kg_identifier: str,
output_identifier: str = None
) -> Dict[str, Any]:
"""
Enrich a knowledge graph with reconstructed prompts and save to database.
Args:
kg_identifier: Knowledge graph identifier to enrich
output_identifier: Optional output identifier
Returns:
Dictionary containing enrichment results
"""
logger.info(f"Starting knowledge graph enrichment for {kg_identifier}")
# Run the prompt reconstruction
results = self.run_prompt_reconstruction(kg_identifier, output_identifier)
if "error" in results:
return results
return {
"status": "enriched",
"knowledge_graph": results.get("knowledge_graph_id"),
"reconstruction_results": results
}
def save_reconstructions(self, kg_id: str, reconstructions: list):
"""Saves prompt reconstructions to the database and updates KG status."""
session = next(get_db())
try:
for recon in reconstructions:
pr = PromptReconstruction(
knowledge_graph_id=kg_id,
relation_id=recon["relation_id"],
reconstructed_prompt=recon["reconstructed_prompt"],
dependencies=recon.get("dependencies", {})
)
session.add(pr)
kg = get_knowledge_graph_by_id(session, kg_id)
if kg:
kg.status = "enriched"
kg.update_timestamp = datetime.now(timezone.utc)
session.commit()
finally:
session.close()
async def enrich_knowledge_graph_task(kg_id: str, task_id: str) -> bool:
"""
Background task for enriching a knowledge graph using PromptReconstructor.
Returns True if successful, False otherwise.
"""
logger.info(f"Starting knowledge graph enrichment task {task_id} for KG {kg_id}")
update_task_status(task_id, "RUNNING", "Enriching knowledge graph")
try:
session = next(get_db())
try:
from agentgraph.reconstruction import PromptReconstructor
# First get the knowledge graph to ensure it exists and to get its filename
kg = get_knowledge_graph_by_id(session, kg_id)
if not kg:
logger.error(f"Knowledge graph with ID {kg_id} not found")
update_task_status(task_id, "FAILED", f"Knowledge graph with ID {kg_id} not found")
return False
# Use the actual filename as the output identifier to ensure it's found properly
# If filename is None, use the ID as a fallback
output_identifier = kg.filename if kg.filename else str(kg_id)
# Log the KG details and output_identifier for debugging
logger.info(f"Knowledge graph found: ID={kg.id}, filename={kg.filename}, status={kg.status}")
logger.info(f"Using output_identifier: {output_identifier}")
# Use the new pure function approach
from agentgraph.reconstruction import reconstruct_prompts_from_knowledge_graph
from backend.database.models import PromptReconstruction
# Get the knowledge graph data
kg_data = kg.graph_data
# Use pure function to reconstruct prompts
reconstructed_relations = reconstruct_prompts_from_knowledge_graph(kg_data)
# Save the prompt reconstructions to the database
for relation in reconstructed_relations:
# Check if prompt reconstruction already exists
existing_pr = session.query(PromptReconstruction).filter_by(
knowledge_graph_id=kg.id,
relation_id=relation["id"]
).first()
if existing_pr:
# Update existing
existing_pr.reconstructed_prompt = relation["prompt"]
existing_pr.dependencies = relation.get("dependencies", {})
else:
# Create new
pr = PromptReconstruction(
knowledge_graph_id=kg.id,
relation_id=relation["id"],
reconstructed_prompt=relation["prompt"],
dependencies=relation.get("dependencies", {}),
)
session.add(pr)
# Update the knowledge graph status
kg.status = "enriched"
kg.update_timestamp = datetime.now(timezone.utc)
session.commit()
update_task_status(task_id, "COMPLETED", "Knowledge graph enriched successfully")
logger.info(f"Knowledge graph {kg_id} enriched successfully")
return True
finally:
session.close()
except Exception as e:
error_message = f"Error enriching knowledge graph: {str(e)}"
logger.error(error_message)
# Log the full traceback for easier debugging
logger.error(traceback.format_exc())
update_task_status(task_id, "FAILED", error_message)
return False