Skip to content

feat(voyageai): rebrand to 'VoyageAI by MongoDB' and refresh model list - #1

Closed
fzowl wants to merge 1 commit into
mainfrom
feat/voyage-mongodb-rebrand-202609-weaviate
Closed

feat(voyageai): rebrand to 'VoyageAI by MongoDB' and refresh model list#1
fzowl wants to merge 1 commit into
mainfrom
feat/voyage-mongodb-rebrand-202609-weaviate

Conversation

@fzowl

@fzowl fzowl commented Sep 8, 2026

Copy link
Copy Markdown
Owner

What & why

VoyageAI is now part of MongoDB. This updates the VoyageAI-related parts of the client accordingly and refreshes the model catalog to VoyageAI's current offerings (verified against docs.voyageai.com, Sept 2026).

(1) Open-PR review

Reviewed all open PRs on the fork (none) and upstream weaviate/java-client (weaviate#598, weaviate#608, weaviate#609, weaviate#610, weaviate#612, weaviate#613, weaviate#616 — all by @dudanogueira). No fzowl/fzoll fork-PRs exist, so nothing to carry over.

(2) Rebrand — 'VoyageAI by MongoDB'

Updated user-visible prose only:

  • Reranker.voyageai() / voyageai(fn) Javadoc.
  • Added class-level Javadoc to Text2VecVoyageAiVectorizer, Multi2VecVoyageAiVectorizer, VoyageAiReranker naming the provider.

Left untouched (not user-visible text, changing them breaks the API/wire protocol): class/method/enum identifiers and module strings such as text2vec-voyageai, multi2vec-voyageai, reranker-voyageai. There is no website/UI/README copy in this repo mentioning VoyageAI.

(3) Refreshed model list

Added current models as constants, kept legacy ones for backward compatibility (removing public constants would be a breaking change):

  • text2vec: voyage-4-large, voyage-4, voyage-4-lite, voyage-code-4, voyage-finance-2, voyage-law-2, voyage-context-4
  • multi2vec: voyage-multimodal-3.5 (legacy voyage-multimodal-3)
  • reranker: rerank-3, rerank-3-lite, rerank-2.5, rerank-2.5-lite

(4) Contextual models — scope note

The contextualized_embed runtime API (inputs: list[str], enable_auto_chunking=true, chunk_size=32000) is part of the VoyageAI Python SDK, not this Java client. In Weaviate, embeddings run server-side in the text2vec-voyageai module; the client only declares the module config (model, baseURL, truncate, dimensions). There is no client-side embed call and the module config has no enable_auto_chunking/chunk_size fields, so those parameters are not applicable here. The contextual model itself (voyage-context-4) is now exposed as a constant. Fabricating unsupported wire fields was deliberately avoided.

Validation

  • mvn compile — clean.
  • mvn -Dtest=JSONTest test — 222 passed, 0 failures (serialization of these vectorizers/rerankers).
  • git diff --stat touches only the 4 VoyageAI files.

- Rebrand user-facing VoyageAI references to 'VoyageAI by MongoDB' in
  Javadoc (reranker factory methods, vectorizer/reranker class docs).
- Add current VoyageAI models as constants alongside legacy ones:
  text2vec: voyage-4-large, voyage-4, voyage-4-lite, voyage-code-4,
  voyage-finance-2, voyage-law-2, voyage-context-4;
  multi2vec: voyage-multimodal-3.5 (legacy voyage-multimodal-3);
  reranker: rerank-3, rerank-3-lite, rerank-2.5, rerank-2.5-lite.
- Legacy constants kept for backward compatibility.
@fzoll

fzoll commented Sep 8, 2026

Copy link
Copy Markdown

Visszavonva (Fődi kérése): a kampányt egyenkénti, ellenőrzött feldolgozásra állítjuk. A branch megmarad.

@fzowl

fzowl commented Sep 8, 2026

Copy link
Copy Markdown
Owner Author

Visszavonva (tulaj kérése): a kampány egyenkénti, ellenőrzött feldolgozásra állt át. A branch megmarad.

@fzowl fzowl closed this Sep 8, 2026
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