An/add gemini embedding 2

[amiamaya]

Summary

Adds Google's new multimodal embedding model gemini-embedding-2 and extends the cost schema with input_image / input_video fields so per-modality input pricing can be represented.

Changes

New model

• providers/google/models/gemini-embedding-2.toml — Google's first multimodal embedding model. Maps text, images, video, audio, and PDFs into a unified 3072-dim embedding space.
  • Pricing (per 1M tokens): text $0.20, image $0.45, audio $6.50, video $12.00
  • Context: 8,192 input tokens, 3,072 output dimensions
  • Modalities: text, image, video, audio, pdf → text
• providers/vercel/models/google/gemini-embedding-2.toml — wrapper using [extends] from the canonical Google entry.

Schema

• packages/core/src/schema.ts — added optional cost.input_image and cost.input_video fields to the Cost schema, mirroring the existing input_audio pattern.
• README.md — documented the new fields and added them to the example.

Web rendering

• packages/web/src/render.tsx — added two new sortable columns: Image Input Cost and Video Input Cost (placed after Audio Output Cost).
• packages/web/src/index.css — updated nth-child selectors to include the new cost columns in the dark-text/mono-font rules, and shifted the Weights-column reference from col 18 → col 24.

Source

Pricing and capabilities sourced from:

• https://ai.google.dev/gemini-api/docs/pricing#gemini-embedding-2
• https://ai.google.dev/gemini-api/docs/embeddings

Notes

• Google publishes image/video pricing in both per-token and per-unit forms ($0.00012/image, $0.00079/frame). Only the per-token rates are encoded since the Cost schema is token-based.
• Batch tier pricing is not available for this model yet.
• Existing gemini-embedding-001 files in wrapper provider dirs were left as-is (full duplicates rather than extends) to keep this PR scoped.

[vladknysh]

Any ETA on this?