It tackles the challenge of building strong poker bots by providing a full Rust stack: ultra‑fast hand evaluation, hierarchical clustering, optimal‑transport utilities, and a game‑agnostic MCCFR solver. The project is split into many crates (cards, transport, mccfr, gameplay, database, server, etc.) that interoperate to train and serve agents via PostgreSQL persistence and an async game coordinator. Researchers and developers who need a high‑performance, end‑to‑end poker AI framework will appreciate the nanosecond evaluator and ready‑made training orchestration. Compared to ad‑hoc scripts, it offers a polished, production‑grade pipeline with proven algorithms like Sinkhorn and external‑sampling CFR.
View on GitHub →faangbait/rbp-train