Add OpenAIEmbeddingCustom

[tirongi]

Similarly to the OpenAIChatModelCustom node,
this feature allow to select a customised modelName or the default one.

[HenryHengZJ]

Can we add an extra field for Model Name on the existing OpenAIEmbedding node instead of creating a new custom node?
Reason ChatOpenAICustom was created separately because the original ChatOpenAI node has different pre-coded dropdown options for different models, whereas OpenAIEmbedding has only 1 model, so there is no need for dropdown.

We can have the extra Model Name field added in, and when provided, it will be used: if (modelName) obj.modelName= modelName

[tirongi]

Can we add an extra field for Model Name on the existing OpenAIEmbedding node instead of creating a new custom node? Reason ChatOpenAICustom was created separately because the original ChatOpenAI node has different pre-coded dropdown options for different models, whereas OpenAIEmbedding has only 1 model, so there is no need for dropdown.

We can have the extra Model Name field added in, and when provided, it will be used: if (modelName) obj.modelName= modelName

Hi, thanks for your comment.

In the beginning I did what you said, but then I realised that would be better to use the same approach used in OpenAI chat model components for better extensibility beyond the reason of the analogy between chat and embedding components.
I thought the case eventually OpenAI in the future may release some other embedding model and we want to pre-code them in the "not-custom" model, in this way we will fall in the same case as in the chat model.

[HenryHengZJ]

Can we add an extra field for Model Name on the existing OpenAIEmbedding node instead of creating a new custom node? Reason ChatOpenAICustom was created separately because the original ChatOpenAI node has different pre-coded dropdown options for different models, whereas OpenAIEmbedding has only 1 model, so there is no need for dropdown.
We can have the extra Model Name field added in, and when provided, it will be used: if (modelName) obj.modelName= modelName

Hi, thanks for your comment.

In the beginning I did what you said, but then I realised that would be better to use the same approach used in OpenAI chat model components for better extensibility beyond the reason of the analogy between chat and embedding components. I thought the case eventually OpenAI in the future may release some other embedding model and we want to pre-code them in the "not-custom" model, in this way we will fall in the same case as in the chat model.

yeah got it, make sense. Dont have a strong preference though. Thanks!