[FEATURE] Adding a choice whether to add node output to state.messages in sequential agents

[kpzen]

In sequential agent flows, by default the State contains an immutable state.messages array, which acts as the transcript or history of the conversation between the user and the agents. This includes the output from "in between" nodes, and also tool nodes, meaning the state can grow quite large from longer flows with several document retrievals. Some of these outputs might not be needed in state, and depending on the size, make the state.messages too large to be of any use. It would therefore be a nice feature to be able to choose whether to add the output of a node to the state.messages array, allowing us to manage it in a new way. This can be done with a simple yes/no choice under "additional parameters". Right now I can think of the tool, llm and agent node, but possibly it makes sense for the condition agent too.

[HenryHengZJ]

this feature already exists:

[kpzen]

@HenryHengZJ I am talking about something different. The feature you linked concerns the construction of each llm call and whether to include the state messages as input, but I'm talking about an option for each node whether its output should be added to the state in the first place. Maybe I was a bit unclear.

[HenryHengZJ]

@HenryHengZJ I am talking about something different. The feature you linked concerns the construction of each llm call and whether to include the state messages as input, but I'm talking about an option for each node whether its output should be added to the state in the first place. Maybe I was a bit unclear.

ah okay my mistake, misunderstood the question. will reopen it!