Repository navigation
Praxity Trace v0.1.0 (experimental)
Pre-release
Pre-release
Praxity Trace v0.1.0 is the first public release. Trace is a course review tool for learning designers: it reads a course and writes one report of how the course is built, so you can see its shape before you decide what to change.
- Reads Praxity Studio courses (
praxity inspect <course-dir> --schema 1; schema 0 still works), folders of static HTML pages, and unzipped SCORM packages. - Measures structure and pace, sentence length and a text profile by lesson, word frequency and acronyms, and emphasis, directly from the course.
- Draws views that need judgment from an LLM you already use, such as Claude or ChatGPT: objectives and the instruction each knowledge check needs, learner task types, concepts and prerequisites, confusable concepts, unexplained terms, diagram opportunities and a short set of recommendations. Trace prepares a bundle, you run it through your LLM, which writes the answer, and Trace checks it against the course revision and labels the result as an interpretation. Trace never calls an LLM itself.
- Gives every view a Chart mode and a Table mode over the same data, with each mark pointing to its lesson file and line. You can comment on the report and export the comments for Trace or an LLM.
- Writes
report.jsonwith the same data for agents.skill/SKILL.mddocuments the commands.
This is an experimental, source-only release. Install from source with Node 24.18 or newer and pnpm; see the README. Views and the report format may change between versions. Rare-word, sentence-structure and cohesion measures use English word lists and run on English courses only. Trace gives no scores, grades or pass marks.