[Bug]: PineconeVectorStore.get_nodes() does not return embedding values

[LHFO94]

Bug Description

When trying to call .get_nodes() on a PineconeVectorStore the embedding values are not returned despite them being in the index.

Versions

llama-index-core == 0.12.11
llama-index-vector-stores-pinecone == 0.4.2

Version

0.12.11

Steps to Reproduce

from dotenv import load_dotenv
from llama_index.core.vector_stores import (
    MetadataFilter,
    MetadataFilters,
    FilterOperator,
)
from llama_index.vector_stores.pinecone import PineconeVectorStore

load_dotenv()

index = PineconeVectorStore.from_params(
    api_key=os.getenv("PINECONE_API_KEY"),
    index_name="documents",
    namespace="dails",
)

doc_id_filter = MetadataFilters(
    filters=[
        MetadataFilter(
            key="doc_id",
            value="02023R1803-20250101",
            operator=FilterOperator.EQ,
        )
    ],
)

nodes = index.get_nodes(filters=doc_id_filter)
for node in nodes:
    print(
        f"nodeid={node.node_id}, ref_doc_id={node.ref_doc_id}, embeddings={node.embedding}"
    )

The code above results in the following output

nodeid=0, ref_doc_id=02023R1803-20250101, embeddings=None
nodeid=1, ref_doc_id=02023R1803-20250101, embeddings=None
nodeid=2, ref_doc_id=02023R1803-20250101, embeddings=None
nodeid=3, ref_doc_id=02023R1803-20250101, embeddings=None
nodeid=4, ref_doc_id=02023R1803-20250101, embeddings=None

The issue lies in the metadata_dict_to_node function which doesn't correctly parse the response from the pinecone index query. If this issue is acknowledged as a bug I would to happy to provide a fix for this.

Relevant Logs/Tracbacks

[dosubot]

Hey @LHFO94! I'm here to help you with any bugs, questions, or contributions. Let me know how I can assist you!

I found a similar closed issue titled "[Bug]: Metadata is missing from vector store results" which was resolved by a pull request that fixed a variable name clash for legacy vector indexes ^[1].

In your case, the issue seems to be related to the metadata_dict_to_node function not correctly parsing the response from the Pinecone index query. The get_nodes method should assign the embedding values from the Pinecone query response to the node.embedding attribute. The embedding values are included in the match.values field of the response ^[2].

You might want to check if the metadata_dict_to_node function is correctly handling the embedding values and ensure that the get_nodes method is properly assigning these values to the nodes.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[LHFO94]

As stated in my PR the metadata_dict_to_node function does not work properly. The reference to the PR supplied is unrelated.

[logan-markewich]

Eh, it's a cross between a bug and a feature. Returning embeddings can eat up a ton of memory.

I'd prefer some kind of bool option

[LHFO94]

Something like this? #17635