RAG ChatApp issue

[anguslou]

Hi,

I modified a RAG based on 5:2.5 Retrieval Augmented Generation ChatApp. It works fine until the RAG ChatApp node with an error

WARN Vector Store Retriever 3:1774:0:21 The selected credentials ‘openai-credentials’ holding the OpenAI API token are not present.

And I have no idea what is it, please help.

[roberto_cadili]

Hi @anguslou,

I believe the problem is rooted in the passing of the credential flow variable to the Vector Store Retriever node in the RAG ChatApp component.

Go inside the RAG ChatApp component and make sure to include the credential variable (see screenshot). This is needed because the downstream Vector Store Retriever node in the component requires OpenAI embeddings, and therefore the connection to OpenAI service (via credential flow variable) needs to persist.

Additionally, be aware that if the RAG ChatApp contains the Chat Model Prompter node, you need to use the OpenAI Chat Model Connector node (and not the OpenAI LLM Connector node). This is because you need a conversational LLM for chatting.

Hope it helps .

Happy KNIMEing,
Roberto

[anguslou]

Hi Roberto,

Thanks for your help. I have modified the flow with the OpenAI Chat Model Connector.

However, the RAG ChatApp is not configurable as the image attached.

[roberto_cadili]

Hi @anguslou,

You need to go inside the RAG ChatApp component . That’s essential to replicate the step that I’ve illustrated in my previous answer/screenshot.

How do you go inside a component? You have 2 options mainly:

1. Right-clik on the component and pick the option “Component > Open component”
   

   

2. CTRL + double-click (on Windows) // Cmd + double-click (on MacOS)

What you’re currently seeing in your second screenshot is the component dialog, as such this is still external (not inside the component ).

Hope it helps,

Roberto

[anguslou]

That works, this is life-saving. Thank you.