Skip to content

Knowledge Graph node IDs should be content hashes for idempotency #2036

Open
@ahgraber

Description

@ahgraber

Describe the Feature
Currently, IDs are UUIDs. it would improve idempotency of knowledge graph generation if the IDs were based on hashes of the text; that way the same text automatically receives the same ID. This is particularly valuable in the case of Splitters; assuming teh splitting algorithm is deterministic, the resulting chunks with a hash ID would mean that the same pipeline does not replicate the chunks (as I believe it currently does).

Why is the feature important for you?
Run knowledge graph generation twice - I believe you end up with 2x chunks because the chunk UUIDs do not recognize the chunk exists.

Additional context
Add any other context about the feature you want to share with us.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions