docs: VoyageAI by MongoDB rebrand + refreshed model list + contextualized_embed spec

[fzowl]

What

1. Rebrand — every visible `VoyageAI` / `Voyage AI` reference is now VoyageAI by MongoDB (MongoDB acquired VoyageAI): docs prose, `SUMMARY.md` nav, the integrations card in `platform/integrations/README.md`, and the `voyage-ai.md` page. Technical identifiers left untouched — model IDs (`voyageai/voyage-*`), the `X-VoyageAI-API-Key` header, and `docs.voyageai.com` links.
2. Refreshed model list — both embedding tables (`platform/integrations/voyage-ai.md`, `vector-database/embeddings.md`) now list the models currently available on docs.voyageai.com: voyage-4 family (`voyage-4-large`, `voyage-4`, `voyage-4-lite`, `voyage-4-nano`), `voyage-code-4`, `voyage-context-4`, plus current `multimodal-3`, `context-3`, `3.5`, `3.5-lite`, `3-large`, `code-3`, `finance-2`, `law-2`. Deprecated/duplicate entries removed (voyage-2/02, large-2, code-2, lite-02-instruct, …). Sample index now uses `voyage-3.5` instead of the retired `voyage-02`.
3. contextualized_embed — documented for the `voyage-context-4`/`voyage-context-3` models per the official spec: `inputs: Union[List[List[str]], List[str]]`, with examples for both supported input formats (nested per-document lists and flat string lists).

Why

Keep Epsilla's VoyageAI integration docs accurate after the MongoDB acquisition and after VoyageAI's model catalog moved to the v4 generation, and give correct contextual-embedding API usage.

Notes

• Model dimensions kept as they were in the repo where unchanged; `voyage-3.5-lite` = 512.
• Heading `## Voyage AI Embedding` renamed → `## VoyageAI by MongoDB Embedding`; the cross-link anchor in `voyage-ai.md` was updated to match (`#voyageai-by-mongodb-embedding`).

[fzowl]

VERDICT:APPROVED

Docs-only rebrand + model-table refresh + contextualized_embed spec. Reviewed correctness, conventions, breaking-change risk, and commit hygiene.

Correctness

• Heading rename ## Voyage AI Embedding → ## VoyageAI by MongoDB Embedding slugifies to voyageai-by-mongodb-embedding, and the cross-link in platform/integrations/voyage-ai.md was updated to that exact anchor. No dangling references.
• vector-database/embeddings.md table drops the pre-existing duplicate rows (voyage-code-2 / law-2 / finance-2 / multilingual-2 were each listed twice) — net cleanup.
• Sample index model updated from the retired voyage-02 to voyage-3.5 in both Python and JS snippets.
• contextualized_embed matches the official spec: inputs: Union[List[List[str]], List[str]], both formats shown, correct input_type="document".
• Technical identifiers correctly left untouched: voyageai/voyage-* model IDs, X-VoyageAI-API-Key header, docs.voyageai.com links.

Conventions / fit — matches existing GitBook doc patterns; voyage-context-4/voyage-context-3 are surfaced the same way the prior merged PR added voyage-context-3.

Breaking-change risk — none; documentation only, no schema/API changes.

Commit hygiene — CLEAN. Single commit, authored zoltan@voyageai.com, no Co-Authored-By, no 'Generated with' trailer, no .claude/ files.

Non-blocking nits

• Prose 'your VoyageAI by MongoDB API key' reads slightly heavy where repeated; purely stylistic.
• voyageai/voyage-3.5-lite = 512 is pre-existing (not introduced here); VoyageAI's default for 3.5-lite is 1024 (configurable). Worth a follow-up verify, not a blocker.
• Contextual models (voyage-context-4/3) appear in the standard embedding table where a reader sets "model": ... at index creation; the added note clarifying they route through contextualized_embed mitigates this.

Approving.

[fzoll]

Új kampány-követelmény (a következő fix körbe beépül): Ha az integráció nem a hivatalos voyageai Python vagy TypeScript csomagot használja, a MongoDB API redirection-t is implementálni kell — a voyageai-python get_default_base_url() logikája szerint: al- prefixű API-kulcs → base URL https://ai.mongodb.com/v1, egyébként https://api.voyageai.com/v1. (Ref: voyage-ai/voyageai-python@9aca465 voyageai/util.py L99.)