Chat with non-embedding db

[CahidArda]

Fixes two functionality for usage with non-embedding Vector Index:

• It was only possible to upsert without data. Vectors without data are not usable in chat. Now it's possible to pass a text with your embedding:

await ragChat.context.addMany([
  {
    id: 1,
    type: "embedding",
    data: [1, 1, 0],
    text: "first embedding", // NEW
    options: { namespace },
  },
  {
    id: 2,
    type: "embedding",
    data: [1, 0, 1],
    text: "second embedding", // NEW
    options: { namespace },
  },
]);

text is added to the data field of the index.

• adding embedding option in chat to make it possible to query a non-embedding index with your own embedding

await ragChat.chat("hello world!", {
  embedding: [0, 0, 0.5], // NEW
  onContextFetched,
  namespace,
});

Fixes

index.addMany is fixed with Promise.all

[linear]

DX-1398 rag-chat: fix usage with non-embedding vector

[fahreddinozcan]

don't we have a batch option? Why do we do it 1 by 1

[CahidArda]

presumably it was done to avoid refactoring the save method which formats the AddContextPayload and makes the upsert. To batch in addMany, we would have to split the save method into two and call the first half in addMany.

If would prefer to do this another PR if that's okay with you