The update operation in the Pinecone node needs to reference to an existing ID in the vector data base.
Example:
Unfortunately updating a vector database is not as straightforward as updating a row from a csv. Vector databases are more useful when working with unstructured data.
Since csv is a highly structured data format, you might be better off using a different structured database (e.g. google sheets, airtable, etc). You can then identify an UUID to update against matching records.
Alternatively, you can keep inserting (embedding) the new CSV file to include the new data. This however can lead to data duplication.
You might wanna check out this tutorial on Upserting Vector Stores: