One line: research decides the categories, the library fills them, and the app walks each teacher through one small step at a time.
Two readers
- The UnitEd web app reads this repo over the GitHub API. It queries
frontmatter — it never slurps bodies. A node's title is the file's first
#heading; graph edges come from wikilinks (double square brackets around an exact title). - Agents and humans add and organize content. Read top-down: this file →
the folder's
CONTEXT.md→ the file itself. Load only what the task needs.
The query protocol
- Every content file carries single-line YAML frontmatter:
type,category,tags,status(plusurlfor external content). The schema lives in20-resources/CONTEXT.md. - Frontmatter is the filter surface; bodies are read only when a resource is
served — and provider-card bodies (
20-resources/30_providers/) are read at step-time because they name which small part to fetch live. _templates/is excluded from scans — its files carry frontmatter but are stamps, not content.- Valid categories are exactly the seven defined in
10-logics/20_content-framework/—1.1through3.2. Nothing else is a category. - Matching a teacher to their next step follows
10-logics/30_matching/.
The flows
| Flow | Path |
|---|---|
| New teacher → first step | app asks the profile questions → Jev evaluates the answers → one category → one resource from 20-resources/ |
| Teacher finishes a step → next step | 10-logics/30_matching/03_progression-rules.md |
| New material arrives | dropped in 00-inbox/ → filed per 00-inbox/CONTEXT.md |
| Big provider content | never mirrored — fetched live in small parts per 20-resources/30_providers/CONTEXT.md |
Factory (stable): _templates/ and 10-logics/. Product (grows):
20-resources/ and 00-inbox/.