AI
Multi-Agent Event Coordinator
LangGraph workflow for vendor sourcing, constraints, and text-to-SQL
Overview
A multi-agent coordinator that automates event-planning workflows with LangChain, LangGraph, LangSmith tracing, MCP tools, Tavily Search, and a self-correcting SQLite text-to-SQL agent.
Problem
Event planning requires matching many constraints across vendors, budgets, locations, and schedules while extracting structured information from databases and web sources.
Dataset
A SQLite event database, user constraint prompts, and live vendor/search results retrieved through MCP-connected web tooling.
Approach
Split the task into specialized agents, used LangGraph schemas to coordinate state, traced executions with LangSmith, integrated Tavily for vendor search, and added text-to-SQL correction loops.
Results
Automated three event workflows, queried two external web APIs, and resolved more than 80% of user constraints through structured extraction and tool orchestration.
Lessons Learned
Multi-agent systems work best when responsibilities, state transitions, and failure recovery paths are explicit rather than hidden inside one broad prompt.
Model / Pipeline
The implementation combines Python, LangChain, LangGraph, LangSmith, MCP, Text-to-SQL, Tavily, OpenAI API with a repeatable workflow for data preparation, evaluation, and communication.