Logical issues on graphrag_neo4j/graphrag.py

[cm8908]

From graphrag_neo4j/graphrag.py:
in function ingest_to_qdrant

it seems like the embedding being upserted does not correspond to the node_id.

Each embedding is generated from each line of raw_data, while node_id is generated by the LLM in an arbitrary order.
That is, the node_id and embedding pair that are being upserted are not related to each other.

Am I getting it right?

[Giwoong-Han]

I have the same idea.

When I tested this code, it seemed to work well. However, because of the "zip" function, it works even if the pairs are not of the same length, so it seems necessary to match the number of documents going into the embedding with the number of nodes.