Description

I would like to use Google's Gemini embedding models (e.g., gemini-embedding-001) with OpenEvolve. Currently, the EmbeddingClient in openevolve/embedding.py only supports OpenAI and Azure OpenAI models.

Motivation

Gemini embedding models provide a cost-effective (often free) alternative to OpenAI's embeddings. Enabling support for them would allow users to run the framework with lower costs.

Proposed Solution

Modify openevolve/embedding.py to:

1. Recognize Gemini model names (e.g., gemini-embedding-001).
2. When a Gemini model is selected, configure the openai client to point to Google's OpenAI-compatible endpoint: https://generativelanguage.googleapis.com/v1beta/openai/.
3. Use GEMINI_API_KEY or GOOGLE_API_KEY for authentication.

Example Configuration

Users should be able to set:

database:
  embedding_model: "gemini-embedding-001"

And have it work seamlessly provided the API key is set.