EdgeAdvisor is a decision-support system that combines deterministic domain evaluation with LLM-based reasoning to produce interpretable, context-aware recommendations. The project uses Texas Hold’em poker as a constrained testbed for exploring how agents can reason over structured state and quantitative signals rather than relying on free-form text generation.
The system maintains a fully structured representation of the game state, computes objective hand strength and outcome distributions using deterministic tools, and then passes this enriched context to an LLM acting as a strategy advisor. Rather than outputting raw actions, the agent focuses on explanation—articulating why a given decision is recommended under the current conditions.
This project demonstrates a general pattern for building human-in-the-loop decision agents that combine reliable computation with transparent reasoning, applicable far beyond the poker domain.
The full write-up details the structured game-state model, deterministic evaluation tools, LLM reasoning strategy, and interactive web interface used to deliver explainable recommendations.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.