You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Apr 30, 2026. It is now read-only.
I'm using PGVectorStore from llamaindex/postgres version 0.0.65. I noticed that the external_id column in the vector store table is always filled with an empty string.
I would have expected the column to contain the ID of the document in the document store from which the node was derived. This would allow me to identify the vectors belonging to a document.
To implement this identification at present, I would have to store the ID of the document in the document store in the node's metadata.
Questions
Is writing the external_id to the vector store planned for a future version?
Is there another way to identify vectors belonging to a document besides using metadata?
Hello,
I'm using PGVectorStore from llamaindex/postgres version 0.0.65. I noticed that the external_id column in the vector store table is always filled with an empty string.
From my perspective, this empty string is hard-coded in the getDataToInsert method of the PGVectorStore class. See https://github.com/run-llama/LlamaIndexTS/blob/main/packages/providers/storage/postgres/src/PGVectorStore.ts#L316.
I would have expected the column to contain the ID of the document in the document store from which the node was derived. This would allow me to identify the vectors belonging to a document.
To implement this identification at present, I would have to store the ID of the document in the document store in the node's metadata.
Questions
Is writing the external_id to the vector store planned for a future version?
Is there another way to identify vectors belonging to a document besides using metadata?
regards Christian