// the find
wormtql/genshin_artifact
莫娜占卜铺 | 原神 | 圣遗物搭配 | 圣遗物潜力。多方向圣遗物自动搭配，多方向圣遗物潜力与评分, Genshin Impact artifacts assessment, artifacts auto combination, artifacts statistics, artifacts potential, and more.
A Genshin Impact artifact optimizer that runs damage calculations and artifact combination searches in the browser via WebAssembly. The Rust core (mona_core) handles the heavy math — attribute graphs, buff stacking, per-character damage formulas — compiled to WASM and driven by a Vue frontend. Aimed at players who want to know which artifacts to equip and whether a dropped piece is worth upgrading.
The Rust/WASM architecture is the right call here: combinatorial artifact search over millions of possibilities needs native-speed loops, and shipping it client-side means zero server costs and no latency. The domain model is thorough — separate attribute graph implementations (simple vs. complicated), a trait-based system for artifact effects and buffs, and per-character skill configs that handle constellation and talent interactions. Character and artifact coverage is genuinely wide, including recent additions like Mavuika, Varesa, and Citlali. The buff system is compositional: you can stack team buffs (Bennett, Sara, Kazuha) and see their combined effect on a damage calculation, which is exactly what serious players need.
The README is almost entirely in Chinese with no English setup guide, and the English README exists but is sparse — getting it running locally requires piecing together the wasm-pack toolchain, the npm build steps, and the recursive submodule clone without much hand-holding. The codebase uses a git submodule for mona_core rather than a workspace member, which creates friction for contributors who clone without --recursive. Character update lag is a real problem: a game that ships new characters every six weeks will always have this tool running behind, and there's no clear contributor guide for adding a new character. The heuristic optimizer for artifact combinations is underdocumented — it's unclear what algorithm it uses or what its approximation guarantees are, so players have no way to know how much worse than optimal its suggestions might be.