BoOngThuoi

Developers debugging unfamiliar code get a screenshot, a vague report, and scattered logs — then hours of manual tracing. ARI turns that into a structured investigation: it reads the screenshot with Muse Spark, searches the repository with deterministic tools (AST, dependency graph, BM25), ranks competing hypotheses with supporting and contradicting evidence, and only calls a cause "verified" when a real test it ran confirms it. When the evidence is not enough, it abstains and tells you exactly what is missing.