Automated tool-calling benchmark dataset, regenerated weekly by the SakThai nightly job (benchmark-v3-publish).
Generated by: .opencode/scripts/run-benchmark-v3.py
Categories: simple, parallel, multi_turn, irrelevance
Schema: each row has messages, tools, expected_tools, category
Files: data/sakthai-bench-v3.jsonl (source), data/train-00000-of-00001.parquet (served by the Datasets Server)
License: MIT