Intelligent Meeting Agenda Generator with Natural Language Processing

Create a JavaScript application that uses natural language processing (via compromise.js or natural npm package) to automatically generate meeting agendas from previous meeting notes, email threads, and Slack conversations. The system should extract action items, prioritize discussion topics, estimate time allocations, and generate a structured markdown output compatible with Google Workspace and Microsoft Teams integrations. Implement a machine learning model that learns from historical meeting patterns to improve agenda relevance over time.