VegeFin/Evolving-Werewolf

README 摘要

English 简体中文 Evolving-Werewolf AI players get stronger with every game. Cross-game knowledge accumulation and evolution : After each game, the engine automatically distills post-game review experience, writing it into a knowledge base partitioned by role (wolf faction / good faction / seer / witch / hunter strategy, etc.). At the start of the next game, these experiences are injected into the new batch of player agents' personas as skills. AI players learn from every win and loss — developing stealthier kill strategies, more accurate seer checks, smarter voting patterns. The more games played, the stronger the AI becomes. How It Evolves Human players don't face fixed-strategy AI, but an opponent collective that continuously evolves with accumulated game experience . Other Features Beyond the evolution mechanism, this is a complete 9-player Werewolf engine (3 Wolves / 1 Seer / 1 Witch / 1 Hunter / 3 Villagers), running on DeepSeek Harness (DSH) : - Engine-hosted : No LLM gamemaster — all flows driven by a state machine. Saves tokens, zero information distortion. - Full role abilities : Wolf two-round discussion + majority vote, Seer verification, Witch save/poison (first-night self-…

分类