Make the embedding field name configurable for the ElasticSearchVectorStore

[dev-jonghoonpark]

Related issue: #2082

Motivation:

On #2082, We found embedding field name is not configurable.

Modification:

Added a configuration option for the embedding field name in the ElasticSearchVectorStore

Result:

Users can now specify a custom embedding field name through configuration.
Close #2082 issue.

Further discussion:

Is an ElasticSearchDocument record really necessary? If not, what about just using the Map class?

[dev-jonghoonpark]

Fixed conflict with #2176

[ilayaperumalg]

Is an ElasticSearchDocument record really necessary? If not, what about just using the Map class?

@dev-jonghoonpark We wanted to go with a specific type instead of Map. Do you see any downside of using the Record type for this?

[dev-jonghoonpark]

@ilayaperumalg
No, I was just using the Map to make the field names configurable, with no other intentions.
In the code I wrote, I had to use conditional branching to configure the field names. This approach feels a bit unsatisfactory, so I wanted to share my thoughts with this comment.

private Object getDocument(Document document, float[] embedding, String fieldName) {
    Assert.notNull(document.getText(), "document's text must not be null");
    if (fieldName.equals("embedding")) {
	    return new ElasticSearchDocument(document.getId(), document.getText(), document.getMetadata(), embedding);
    }
    
    return Map.of("id", document.getId(), "content", document.getText(), "metadata", document.getMetadata(),
		    fieldName, embedding);
}

Could there have been a better way?

[dev-jonghoonpark]

@ilayaperumalg
I've updated it based on your feedback.

[dev-jonghoonpark]

Fixed conflict with #2209

[dev-jonghoonpark]

Hi @ilayaperumalg, I wanted to give a quick reminder about this PR in case it was forgotten.
I've resolved the conflicts.
Please take a look when you have time.
Thanks!

[ilayaperumalg]

With this change, you can delete ElasticSearchDocument now

[dev-jonghoonpark]

I deleted the ElasticSearchDocument record. Thanks for your feedback!

[ilayaperumalg]

@dev-jonghoonpark Thank you for the PR making the embedding field name configurable in ElasticSearch vector store. Rebased and merged as 4f4da30