Razvan Dinu

This quick fix was included in another PR.

I can give more context here. For security reasons, the paths for configs need to be normalized, and we need to make sure they are within the original server config...

Hi @msalhab96 ! The way to achieve this is to define an input rail flow which changes the `$user_message` variable, not `$last_user_message`. See a quick example in this test: https://github.com/NVIDIA/NeMo-Guardrails/blob/develop/tests/test_input_ouput_rails_no_dialog.py...

@jackchan0528 : did you test with the provided notebook? (I've just tried it again and it works correctly)

You could set a context variable in your flow, and use that to conditionally trigger the logic in the input flow: ``` define flow your flow user some intent $mark_in_your_flow...

@jackchan0528 : this is most likely because you've changed the model to `gpt-3.5-turbo-0125` (from `gpt-3.5-turbo-instruct`). Can you confirm that use the instruct version works correctly on your end? cc @trebedea

On my end, running the notebook as-is yields the correct result:

Hi @analyticalpicasso! It seems that somehow you're mixing an older installed version with the latest code. If you've pulled the latest changes, make sure you also do `pip install -e...

Can you try to install from scratch version `0.8.0`, which is now published, in a fresh venv? And if it's still reproducible, do share the steps with a minimal configuration...

@pradeepdev-1995: The syntax is incorrect, inside the YAML file. The `models` key should be an array. You're missing a dash: ```yaml models: - type: main engine: azure model: parameters: azure_endpoint:...