# Multiple embeddings per ID

**URL:** <https://community.pinecone.io/t/multiple-embeddings-per-id/478>\
**Category:** Support\
**Created:** [October 5, 2022, 1:19pm UTC](https://community.pinecone.io/t/multiple-embeddings-per-id/478 "2022-10-05T13:19:15Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![etherealmachine](https://avatars.discourse-cdn.com/v4/letter/e/ce7236/32.png) [@etherealmachine](https://community.pinecone.io/u/etherealmachine)\
**Post date:** [October 5, 2022, 1:19pm UTC](https://community.pinecone.io/t/multiple-embeddings-per-id/478/1 "2022-10-05T13:19:15Z")

</div>

Does Pinecone support multiple embeddings per ID? It would appear that given the upsert commands of:

```auto
index.upsert([
    ("A", [0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1]),
    ("A", [0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2]),
])

```

The second embedding would overwrite the first one.

The task I’m trying here is embedding multiple paragraph vectors for a single document (“A” being the document, in the example). If multiple embeddings per ID isn’t possible, what’s the recommended way to approach this? I can thing of encode the paragraph in the id, e.g. “A-0, A-1, A-2” and possibly in the metadata, but this doesn’t seem ideal, like if I want to delete or change all paragraphs for a document I wouldn’t have a consistent index of the IDs associated with a given doc.

---

<div class="post-metadata">

**Author:** ![Cory\_Pinecone](https://sea2.discourse-cdn.com/flex020/user_avatar/community.pinecone.io/cory_pinecone/32/1712_2.png) [@Cory\_Pinecone](https://community.pinecone.io/u/Cory_Pinecone)\
**Post date:** [October 5, 2022, 1:33pm UTC](https://community.pinecone.io/t/multiple-embeddings-per-id/478/2 "2022-10-05T13:33:05Z")

</div>

Hi @etherealmachine,

You’re right, Pinecone doesn’t support lists of embeddings in a single vector. The recommended way to have separate embeddings for each paragraph in a document, and then associate them all together, would be to use metadata to store a unique key for the document itself. You can then filter by metadata to return or delete all of the vectors at once.

See [this page on metadata filtering](https://www.pinecone.io/docs/metadata-filtering/#metadata-filtering) for more details.
