Problem

I'm currently using a local embedding service with a custom base URL via `LLM_OPENAI_COMPATIBLE_BASE_URL`. However, the embedding model ID is currently hardcoded in `src/embedding_client.py`:

• `openai`: `text-embedding-3-small`
• `openrouter`: `openai/text-embedding-3-small`
• `gemini`: `gemini-embedding-001`

This means even when I point to my own embedding endpoint, I cannot specify which model to use. I had to manually modify the source code to change the model (e.g., to `openai/nomic-embed-text`), which isn't ideal for maintenance.

Suggested Solution

Would it be possible to add a new configuration parameter, perhaps something like:

```bash
LLM_EMBEDDING_MODEL=text-embedding-3-large
```

This would allow users to override the default model without modifying the codebase.

Benefits

• Greater flexibility for self-hosted deployments
• Easier to switch between different embedding models
• No code changes required when upgrading/changing models

Thanks for considering this! Happy to provide more context if needed.