Question Around Lang-chain and action mapping

I am working on a project where I want to use agentic AI to help users define there requirement more precisely and once that is done map it to a dynamic DAG , which will eventually be executing a large orchestrated task , where each node in the dag is one action or intent of the user . Now I am new to this realm of AI and and I am unsure how I can achieve this , mapping from user inputs via agents to DAG Nodes . My plan is as below I am unsure if that makes sense and will it scale for large user base ?

User Message

│

▼

Agent

(LangChain Chain)

│

▼

[ Branch: isIntentConfidentEnough? ]

├── Yes ──▶ RasaHandler ➝ Intent ➝ Action Executor

└── No ──▶ LLM Reasoner ➝ LangChain Tools / Planner ➝ Action DAG

LLM Reasoner to DAG I am not sure how to achieve this , should I leverage MCP server or its an overkill for the use case , now the DAG generation from the user agent interaction is the most important part of the idea . Any help around this can be greatly appreciated....

7 comments