llamaIndex - ✅(Solved) Fix [Bug]: `VectorStoreIndex` delete helpers use a wrong delete API and sync call in async path [2 pull requests, 1 participants]

llamaIndex2026-03-17 17:12:09

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

run-llama/llama_index#21049•Fetched 2026-04-08 00:53:04

View on GitHub

Comments

Participants

Timeline

Reactions

Author

gautamvarmadatla

Participants

gautamvarmadatla

Timeline (top)

cross-referenced ×2labeled ×2closed ×1

Root Cause

_delete_from_index_struct() and _adelete_from_index_struct() both call self._vector_store.delete(node_id), passing a node ID to a ref-doc deletion API. This is semantically incorrect imo, since node-wise deletion has separate delete_nodes() / adelete_nodes() APIs. It is also unnecessary, because delete_ref_doc() already calls delete(ref_doc_id), and adelete_ref_doc() already schedules adelete(ref_doc_id). In addition, _adelete_from_index_struct() makes this extra delete call synchronously instead of using the async delete path.

Code Example

### Relevant Logs/Tracbacks

RAW_BUFFERClick to expand / collapse

Bug Description

Version

0.14.18

Steps to Reproduce


import asyncio
from typing import Any

from llama_index.core import VectorStoreIndex, MockEmbedding
from llama_index.core.schema import Document
from llama_index.core.storage.storage_context import StorageContext
from llama_index.core.vector_stores.simple import SimpleVectorStore

delete_log = []


class TrackingVectorStore(SimpleVectorStore):
    def delete(self, ref_doc_id: str, **kwargs: Any) -> None:
        delete_log.append(("SYNC", ref_doc_id))
        super().delete(ref_doc_id, **kwargs)

    async def adelete(self, ref_doc_id: str, **kwargs: Any) -> None:
        delete_log.append(("ASYNC", ref_doc_id))
        await super().adelete(ref_doc_id, **kwargs)

    @property
    def client(self) -> Any:
        return None


store = TrackingVectorStore()
index = VectorStoreIndex.from_documents(
    [Document(text="hello world", doc_id="my-doc-id")],
    storage_context=StorageContext.from_defaults(vector_store=store),
    embed_model=MockEmbedding(embed_dim=8),
)

index.delete_ref_doc("my-doc-id")
print("Sync delete calls:", delete_log)

delete_log.clear()

store2 = TrackingVectorStore()
index2 = VectorStoreIndex.from_documents(
    [Document(text="hello world", doc_id="my-doc-id")],
    storage_context=StorageContext.from_defaults(vector_store=store2),
    embed_model=MockEmbedding(embed_dim=8),
)

await index2.adelete_ref_doc("my-doc-id")
print("Async delete calls:", delete_log)

Relevant Logs/Tracbacks

Sync delete calls: [('SYNC', 'my-doc-id'), ('SYNC', 'afc5df02-1d56-40bc-bda2-fdd222f2fd83')]
Async delete calls: [('ASYNC', 'my-doc-id'), ('SYNC', 'my-doc-id'), ('SYNC', 'b1aa8b5e-981f-4494-a9c0-ee137533f749')]

Note: with SimpleVectorStore, the extra ('SYNC', 'my-doc-id') in the async case comes from the default adelete() fallback to delete(). The final ('SYNC', '<node-id>') is the helper bug.

extent analysis

Fix Plan

To fix the issue, we need to remove the unnecessary self._vector_store.delete(node_id) calls from _delete_from_index_struct() and _adelete_from_index_struct().

Here are the steps:

Remove the self._vector_store.delete(node_id) call from _delete_from_index_struct().
Remove the self._vector_store.delete(node_id) call from _adelete_from_index_struct() and use the async delete path instead.

Example code:

def _delete_from_index_struct(self, node_id: str) -> None:
    # Remove the following line
    # self._vector_store.delete(node_id)
    # Rest of the method remains the same

async def _adelete_from_index_struct(self, node_id: str) -> None:
    # Replace the following line with the async delete call
    # self._vector_store.delete(node_id)
    await self._vector_store.adelete(node_id)
    # Rest of the method remains the same

Verification

To verify the fix, run the provided test code and check the delete_log output. It should no longer contain the extra ('SYNC', '<node-id>') calls.

Extra Tips

Make sure to test the fix with both sync and async delete methods to ensure the issue is fully resolved.
Review the code to ensure that the delete and adelete methods are being used correctly throughout the project.

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

#api #ssr #installation #tensor shape #autograd error #configuration error #environment variable #network issue #logging issue

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

llamaIndex - ✅(Solved) Fix [Bug]: `VectorStoreIndex` delete helpers use a wrong delete API and sync call in async path [2 pull requests, 1 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Fix Action

Fixed

PR fix notes

PR #21050: fix(core): remove incorrect per-node delete calls in index helpers

Description (problem / solution / changelog)

Description

New Package?

Version Bump?

Type of Change

How Has This Been Tested?

Suggested Checklist:

Changed files

PR #21055: fix: use async adelete in _adelete_from_index_struct