upport for compatible OpenAI models or Ollama.

[ydogg]

Very valuable project, thank you for your work. Currently the model only supports GEMINI, and there is no support for compatible OpenAI models or privately deployed models like Ollama. I wonder if there are any plans for supporting these in the future? Sorry, I'm not very familiar with Rust.

[xiaoniaoyouhuajiang]

I used fastembed-rs combined with a local embedding to replace original one to ensure that the codebase_index part works properly. However, I'm not entirely sure if this is a "reasonable" modification. My code is roughly as follows:

pub async fn embed_entity_local_async(text: String) -> Result<Vec<f64>> {
    let mut model = TextEmbedding::try_new(
        InitOptions::new(EmbeddingModel::BGELargeENV15Q).with_show_download_progress(true),
    )?;
    let text_container = vec![text];
    let embeddings = model.embed(text_container, None)?;
    let result = embeddings[0].clone().into_iter().map(|x| x as f64 ).collect();
    Ok(result)
}

@xav-db Could you please help me confirm whether this approach is reasonable? If so, I would be willing to merge it into a new command-line option and contribute.

[xav-db]

@Gahnxd @putt-t would you be able to take a look at this!

[xav-db]

I used fastembed-rs combined with a local embedding to replace original one to ensure that the codebase_index part works properly. However, I'm not entirely sure if this is a "reasonable" modification.

I think this seems reasonable