SparkLang: CPU model training in one reviewable file

Field notes

SparkLang uses a small declarative program to keep training inputs, job polling, and output assertions together. Its dry-run path relies on checked fixtures, including accepted and succeeded status files, so a repository can exercise the workflow before connecting to a trainer. The page also points to a reference HTTP service with four CPU methods and shows both passing and failing example programs. The displayed release is v0.6.3, with installers, a changelog, and MIT-licensed source linked from the site.