Use compression when saving knowledge graphs

[lryan599]

Describe the Feature
I found that ragas directly dumps node and relationships directly when saving the knowledge graph, and the relationships holds all the information about the relevant node, with a very large amount of redundancy, especially in the summary_embedding field. So when saving relationships, wouldn't it be better to just save the ID of the related node instead?

Also, there are some compression methods that can be used when saving embeddings. For instances, some RAG pipelines use base64 encoding to save these embeddings.

Why is the feature important for you?
Generating test sets takes a lot of time, and it is necessary to save the knowledge graphs generated in between to be able to load the knowledge graphs directly in case of an anomaly instead of starting all over again.

[lryan599]

ragas/src/ragas/testset/graph.py

Line 88 in cb63a82

class Relationship(BaseModel):

It would be fine to add this code to class Relationship

    @field_serializer("source", "target")
    def serialize_node(self, node: Node):
        return node.id