Skip to content

Conversation

@jba
Copy link
Contributor

@jba jba commented Jul 3, 2024

  • An EmbedRequest takes a slice of Documents instead of a single Document.

  • An EmbedResponse contains embeddings for each document. The []float32
    containing the embedding is inside a struct, to accommodate future
    additions (and to match the JS).

  • The googleai embedder works on multiple documents sequentially.
    It should be changed to use the BatchEmbed RPC.

  • The vertexai embedder always handled multiple "instances". Now an
    instance is the concatenated text parts of a document; before it
    was one text part of the sole document. (This is the only behavioral
    change.)

There is one unrelated change: the prompt of a generation test
was changed because the previous prompt is now blocked for the
"recitation" reason.

- An EmbedRequest takes a slice of Documents instead of a single Document.

- An EmbedResponse contains embeddings for each document. The []float32
  containing the embedding is inside a struct, to accommodate future
  additions (and to match the JS).

- The googleai embedder works on multiple documents sequentially.
  It should be changed to use the BatchEmbed RPC.

- The vertexai embedder always handled multiple "instances". Now an
  instance is the concatenated text parts of a document; before it
  was one text part of the sole document. (This is the only behavioral
  change.)

There is one unrelated change: the prompt of a generation test
was changed because the previous prompt is now blocked for the
"recitation" reason.
@jba jba merged commit 1546e28 into main Jul 7, 2024
@jba jba deleted the jba/multiembed branch July 7, 2024 15:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants