feat: Support dimensions param for Cohere embed v4

[amirzaushnizer]

No description provided.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Feb 1, 2026 4:10pm

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

Greptile Overview

Greptile Summary

Added support for the dimensions parameter for Cohere embed v4 models on Bedrock. The implementation maps the OpenAI-compatible dimensions parameter to Cohere's output_dimension parameter, consistent with the existing Cohere embedding transformation logic. The change extends model detection in utils.py to recognize cohere.embed-v4 models alongside existing v3 models.

Confidence Score: 5/5

• This PR is safe to merge with minimal risk
• The changes are minimal, well-structured, and follow existing patterns in the codebase. The implementation correctly maps OpenAI parameters to Cohere API parameters, consistent with the standard Cohere embedding transformation. No performance concerns or custom rule violations detected.
• No files require special attention

Important Files Changed

Filename	Overview
litellm/llms/bedrock/embed/cohere_transformation.py	Added dimensions parameter support, mapping it to output_dimension for Cohere v4
litellm/types/llms/bedrock.py	Added output_dimension field to CohereEmbeddingRequest TypedDict
litellm/utils.py	Extended model detection to include cohere.embed-v4 alongside cohere.embed-multilingual-v3

Sequence Diagram

sequenceDiagram
    participant User
    participant LiteLLM
    participant Utils
    participant BedrockCohereConfig
    participant BedrockAPI

    User->>LiteLLM: embedding(model="bedrock/cohere.embed-v4:0", input=text, dimensions=512)
    LiteLLM->>Utils: get_optional_params_embeddings()
    Utils->>Utils: Check model contains "cohere.embed-v4"
    Utils->>BedrockCohereConfig: Instantiate BedrockCohereEmbeddingConfig()
    Utils->>BedrockCohereConfig: get_supported_openai_params()
    BedrockCohereConfig-->>Utils: ["encoding_format", "dimensions"]
    Utils->>BedrockCohereConfig: map_openai_params({"dimensions": 512})
    BedrockCohereConfig->>BedrockCohereConfig: Map "dimensions" to "output_dimension"
    BedrockCohereConfig-->>Utils: {"output_dimension": 512}
    Utils-->>LiteLLM: optional_params
    LiteLLM->>BedrockAPI: Invoke with output_dimension=512
    BedrockAPI-->>LiteLLM: Embedding response
    LiteLLM-->>User: EmbeddingResponse

Loading

[greptile-apps]

_{3 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[Sameerlite]

LGTM