A followup question is , I am unclear about the below flow customer asks questions to the Agent , now agent will have to somehow decide , when it has enough data to actually convert the asks from the customer to an intent parse the intent and map it to supported operation by the actual platform . I would see that I can expose well documented API's for the platform which can be referred by the agent as a tool via MCP or something to decide if the intent maps to the supported platform operations (API ) and then invoke the parser and if not ask more question based on the API documentation we have. would I need a RAG which the agent can refer to decide if the asks form the user can be mapped to a valid operation or not . Am I thinking rationally or am I in totally wrong direction. User --> agent --> LangGraph --> (tool) RAG API documentation --> (tool: if valid intent ) Parse Intent (using LLM) --> Map intent to operation (NOT sure how to do this right now any help would be great) --> Generate DAG node and add it to the DAG definition --> Repeat above until the conversation is complete , not sure again how can we complete a conversation with the user , do we use some specific phrase which we again pass to another LLM tool to decide if conversation is complete or not ... Sorry for so many questions ... Any help will be great ...😀 The Link for the Flow in my thoughts https://www.mermaidchart.com/raw/5c08f414-75a5-4fa1-bf8c-a8d852c83ccd?theme=light&version=v0.1&format=svg

I am confused or unaware of how to manoeuvre the user to navigate towards the right operations supported by our platform . The conversations can be vague and will not directly map to our supported operations ...