Discord Channel Selector / LangSmith Multi-Agent System

Multi-agent recommendation system that helps users navigate large Discord communities by collecting preferences, ranking channels, and tracing the workflow with LangSmith.

Highlights

• Built a two-agent workflow for preference gathering and channel recommendation.
• Added LangSmith tracing to improve observability, debugging, and prompt iteration.
• Ran prompt compression and efficiency experiments to reduce unnecessary token usage.
• Returned structured channel rankings rather than free-form suggestions.

Tech Stack

Python, LangChain, OpenAI API, LangSmith, and PyYAML.

Why It Matters

This project demonstrates practical LLM orchestration with structured outputs, prompt optimization, and the kind of observability needed for real agent workflows.