Skip to content

LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows

fix/Prompt-Tool-delete-fix (#399)

Athul39e803a78a8e
diff --git a/backend/prompt_studio/prompt_studio_index_manager/models.py b/backend/prompt_studio/prompt_studio_index_manager/models.py
index df249bb74..9129b9cd9 100644
--- a/backend/prompt_studio/prompt_studio_index_manager/models.py
+++ b/backend/prompt_studio/prompt_studio_index_manager/models.py
@@ -4,13 +4,12 @@ import uuid
from account.models import User
from django.db import connection, models
-from django.db.models.signals import post_delete
+from django.db.models.signals import pre_delete
from django.dispatch import receiver
from prompt_studio.prompt_profile_manager.models import ProfileManager
from prompt_studio.prompt_studio_core.prompt_ide_base_tool import PromptIdeBaseTool
from prompt_studio.prompt_studio_document_manager.models import DocumentManager
from unstract.sdk.constants import LogLevel
-from unstract.sdk.exceptions import SdkError
from unstract.sdk.vector_db import VectorDB
from utils.models.base_model import BaseModel
@@ -90,31 +89,38 @@ class IndexManager(BaseModel):
]
+def delete_from_vector_db(index_ids_history, vector_db_instance_id):
+ org_schema = connection.tenant.schema_name
+ util = PromptIdeBaseTool(log_level=LogLevel.INFO, org_id=org_schema)
+ vector_db = VectorDB(
+ tool=util,
+ adapter_instance_id=vector_db_instance_id,
+ )
+ for index_id in index_ids_history:
+ logger.debug(f"Deleting from VectorDB - index id: {index_id}")
+ try:
+ vector_db.delete(ref_doc_id=index_id)
+ except Exception as e:
+ # Log error and continue with the next index id
+ logger.error(f"Error deleting index: {index_id} - {e}")
+
+
# Function will be executed every time an instance of IndexManager is deleted.
-@receiver(post_delete, sender=IndexManager)
+@receiver(pre_delete, sender=IndexManager)
def perform_vector_db_cleanup(sender, instance, **kwargs):
"""Signal to perform vector db cleanup."""
logger.info("Performing vector db cleanup")
- logger.debug("Document tool id: %s", instance.document_manager.tool_id)
+ logger.debug(f"Document tool id: {instance.document_manager.tool_id}")
try:
# Get the index_ids_history to clean up from the vector db
index_ids_history = json.loads(instance.index_ids_history)
vector_db_instance_id = str(instance.profile_manager.vector_store.id)
- # Generate a run_id
- org_schema = connection.tenant.schema_name
- util = PromptIdeBaseTool(log_level=LogLevel.INFO, org_id=org_schema)
- vector_db = VectorDB(
- tool=util,
- adapter_instance_id=vector_db_instance_id,
+ delete_from_vector_db(index_ids_history, vector_db_instance_id)
+ except Exception as e:
+ logger.warning(
+ "Error during vector DB cleanup for deleted document "
+ "in prompt studio tool %s: %s",
+ instance.document_manager.tool_id,
+ e,
+ exc_info=True, # For additional stack trace
)
- for index_id in index_ids_history:
- logger.debug(f"Deleting from VectorDB - index id: {index_id}")
- try:
- vector_db.delete(ref_doc_id=index_id)
- except Exception as e:
- # Log error and continue with the next index id
- logger.error(f"Error deleting index: {index_id} - {e}")
- # Not raising any exception.
- # Cleanup should not fail the deletion of the index manager.
- except SdkError as e:
- logger.error(f"Error while performing vector db cleanup: {e}")