How to build chatbot in fitness company (cross all domain of company and easy to scale)

[QuanduongAI]

My idea is building a chatbot to help the coach (who have a lot of trainee in their workspace, website flat form). This bot can help coach faster to check information, visualize information and recommend inside based on that data (such as tracking exercises, their workspace revenue,... ex : Show me Jane Doe’s progression in bench press over the last 4 weeks? or some questions like : Show me my revenue last month and do my VIP (Long Nguyen) still doing Bench Press last month, if having, give me his record? )
I decide to build multi-agent and then will scale it bigger and bigger :
- Technical core : Langchain , langraph

• Init project to easy scale more agent : Agent manager, Tool manager, Recommend system, Language Model Manager,... that will help to easy to build new workflow if have other domain in core will added
  My architecture get idea from building position in company
  Hope have someone who have experience to review my idea and my diagram of chatbot architecture and give me some advice.
  And i have debate with my co-worker : My idea is break an agent ( search agent - will bind tool to search data by question of coach, It is the agent's own specific task to perform and have supervisor and note taker to get the useful information - because this is a specific data => should really understand what is in fitness and trend to analyze) but my co-worker said that just build agent (collection data agent - it have a both tasks collecting and answering , visualize agent) i feel it have stuck if it will be scale bigger and he say i dont understand what is function calling