Evaluar beta v0.1 is a code-first evaluation framework for structured model outputs like bounding boxes, OCR, tables, and merged document records. It provides Python suite definitions, HTTP/callable/fixture connectors, ground-truth scoring, saved JSON results, report exports, and a TUI for inspecting metrics, failures, overlays, and GT edits.
This beta focuses on validating real downstream workflows: installing Evaluar, defining eval suites, running model evaluations, reviewing scorecards, and debugging detection quality with practical visual tools.