No user-submitted related servers found.
Scored across 32 tools
Most tools have sharply distinct roles (build vs inspect vs answer vs export), and the verbose Chinese descriptions spell out preconditions and return shapes. A few pairs could still be confused: find_instances vs object_rows (lookup by name vs enumerate all instances), and the ask_question / list_questions / fix_question / drop_filter cluster requires careful reading to pick correctly.
Everything is uniformly snake_case and most names follow verb_noun (find_instances, list_objects, save_card, set_rules, build_ontology, export_ttl). The deviations are noun-only names like health, data_check, data_layout, object_rows, run_overview, which form a recognizable secondary pattern rather than random mixing.
32 tools is heavy for one server, and there is clear bloat: the four export_* tools repeat an identical ~200-character paragraph about out_dir, manifest files and overwrite semantics, and card/question tooling is split across many narrow operations. The multi-phase workflow (ingest, model, verify, answer, export) justifies breadth, but the surface should be consolidated.
The surface covers the full lifecycle: preview upload, build ontology, health-check data, confirm types/relations/rules, answer questions with derived-metric approval, record customer answers, and export to CSV/Mermaid/Turtle/pipeline spec. Minor gaps: no way to delete or re-run a stored run, and instance-level data export beyond per-object forms is absent.