Razvan Dinu

Related to #495 . - add an `LLMRails.register_embedding_provider(name: str, cls: Type[EmbeddingModel])` method to `LLMRails`. - store it in a global `registered_embedding_providers` in https://github.com/NVIDIA/NeMo-Guardrails/blob/develop/nemoguardrails/embeddings/embedding_providers/__init__.py - update the logic here to use...

A set of fixes for various small issues.

Right now, the import fails silently and typically you get an "Action not found" error.

Allow new line characters inside predefined bot messages

3

If "\n" is used inside a predefined bot message, it will be returned as is as part of the message. It should be translated to an actual `\n` character.

If no main LLM model is configured, the following error might be triggered: ``` ERROR:nemoguardrails.actions.action_dispatcher:Error 'NoneType' object has no attribute 'agenerate_prompt' ``` This PR adds a clear message for this...

This is WIP. Needs more testing.

WIP. Still needs: - [x] configuration option to choose modes - [x] update all existing rails to comply - [ ] documentation - [ ] tests

Allow handling LLM generation exceptions

4

If a third party API is used, some external guardrails can block the LLM generation. If this happens, currently a `None` response is returned. There should be an option to...

If the output of a chain is a string, and it if the guardrails layer is blocking the input, or the output, the response format becomes a dict with the...

See #553 for context.