As i am pretty new to n8n I would like to asks, can someone explain in simple way, what is the difference between these 2 processes. I looked at the templates and youtube videos and sometimes I see the left version and sometime the right version. What is actually the role of additional nodes on the right side. I can’t find much information by google it.
The left version of this workflow is pretty recent (introduced in 1.74.0). This allows you to pass in the results from the vector store as-is back to the agent. The version on the right is the “old” way of doing it, where the results of the vector store are processed by the language model to answer the question, and that result is then passed back to the agent.
The second method can still be useful if you want to reduce the number of tokens processed by the agent, and if you want to use different types of models for the agent and the answering of questions. The downsides are that the question-answering model doesn’t have all the context that the agent does, and that it doesn’t pass back meta-data that the agent might need.
