feat(pricing): gemini-embedding-2 GA cost map, blog, and test

[Sameerlite]

Summary

• Register gemini-embedding-2 (GA) in model_prices_and_context_window.json and backup for Gemini and Vertex-style keys.
• Add blog gemini_embedding_2_ga with proxy-only /v1/embeddings curl examples.
• Add test_gemini_embedding_2_ga_in_cost_map to lock multimodal unit pricing fields.

Test plan

poetry run pytest tests/test_litellm/test_utils.py::test_gemini_embedding_2_ga_in_cost_map -v

[veria-ai]

Low: No security issues found

This PR adds pricing/cost-map entries for a new Gemini embedding model, a blog post documenting the feature, and a test verifying the cost map entries. All changes are to documentation, static JSON configuration data, and tests — no runtime code is modified.

---

Status: 0 open
Risk: 1/10

Posted by Veria AI · 2026-04-24T06:28:48.450Z

[greptile-apps]

Greptile Summary

This PR registers the GA release of gemini-embedding-2 in the cost map (both primary and backup JSON), adds a blog post, and adds a test to lock the pricing fields. Three entries are added (gemini-embedding-2, vertex_ai/gemini-embedding-2, gemini/gemini-embedding-2) with pricing fields matching the preview model.

Prior review concerns are addressed: the bare-key gemini-embedding-2 entry now includes supports_multimodal: true, and uses_embed_content: true is correctly set on both Vertex AI entries. The gemini/ entry correctly omits uses_embed_content because that flag is only consumed in the Vertex AI code path (main.py:5242, common_utils.py:271) via custom_llm_provider=\"vertex_ai\".

Confidence Score: 5/5

Safe to merge — all three cost map entries are internally consistent and correct, prior concerns are resolved.

No P0 or P1 issues found. Both previously flagged concerns (missing supports_multimodal on bare key, uses_embed_content consistency) are addressed in this PR. The test correctly validates all pricing and routing fields without making real network calls.

No files require special attention.

Important Files Changed

Filename	Overview
model_prices_and_context_window.json	Adds three GA entries for gemini-embedding-2 (bare key, vertex_ai/, gemini/) with consistent pricing fields and correct uses_embed_content/supports_multimodal flags.
litellm/model_prices_and_context_window_backup.json	Backup JSON updated identically to the primary; all three GA entries are consistent with the primary file.
tests/test_litellm/test_utils.py	New test reads JSON from filesystem (no network calls) and validates pricing fields and routing flags for all three GA entries.
docs/my-website/blog/gemini_embedding_2_ga/index.md	New blog post documenting GA model name, proxy config, and SDK usage examples; no code changes.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["embedding(model=...)"] --> B{Parse provider}
    B -->|"gemini/gemini-embedding-2"| C["Gemini API path\n(generativelanguage.googleapis.com)"]
    B -->|"vertex_ai/gemini-embedding-2"| D["Vertex AI path"]
    B -->|"gemini-embedding-2 (bare key)"| E["vertex_ai-embedding-models path"]
    D --> F{uses_embed_content?}
    E --> F
    F -->|true| G["embedContent endpoint\n(main.py:5246)"]
    F -->|false| H["predict endpoint"]
    C --> I["Gemini embedContent\n(no flag check needed)"]

Loading

_{Reviews (2): Last reviewed commit: "Fix greptile reviews" | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!