# RAG Document Deletion and Tombstone Handling
Status: public
Confidence: medium (0.725) (verified)
Last verified: 2026-06-02
Generation: ai_structured


## TL;DR

RAG deletion handling keeps removed documents from being retrieved, cited, cached, or answered from stale chunks.

## Core Explanation

RAG pipelines often split one source document into many chunks. A delete event therefore has to map from source document ID to every derived chunk, vector, metadata record, and cached answer that depends on it.

Agents should inspect deletion logs, tombstone records, vector store delete status, source connector checkpoints, and cache invalidation before saying that a document is gone. Missing this step can leave sensitive or obsolete content available through retrieval.

## Source-Mapped Facts

- Qdrant documentation describes deleting points by point IDs or by a filter. ([source](https://qdrant.tech/documentation/concepts/points/))
- Weaviate documentation describes deleting objects by ID and by filter. ([source](https://docs.weaviate.io/weaviate/manage-objects/delete))
- Elasticsearch documentation describes the Delete API as removing a JSON document from an index. ([source](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-delete))

## Further Reading

- [Qdrant Points](https://qdrant.tech/documentation/concepts/points/)
- [Weaviate Delete Objects](https://docs.weaviate.io/weaviate/manage-objects/delete)
- [Elasticsearch Delete API](https://www.elastic.co/docs/api/doc/elasticsearch/operation/operation-delete)