Passage (Working Memory Jev)
Localhost teaching aid that uses Jev to estimate where instructional text asks readers to hold too many ideas at once.
Overview
Passage (Working Memory Jev) helps educators inspect where instructional text may ask a learner to hold too many ideas or relationships at once. Running on localhost with the real Jev API, it reviews a whole passage, then sentences and smaller reading steps, proposing active idea groups, source connections and changes in demand, and compares estimated demand with a selected budget of 1-5 slots given reading grade, prior knowledge, teaching goal and presentation mode.
It highlights source text, draws connections between exact occurrences, reassesses what matters at each step, and shows average and peak estimates with categorical confidence, keeping estimates provisional when checks remain unresolved. The README stresses that slot estimates and confidence labels are not validated measurements or calibrated probabilities, and a list summary notes its evaluation opens by naming two tests its own model fails. It is free for individuals and non-commercial use under a custom licence, not open source.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How many ideas must the reader hold at this step?(inferred) | — |
| Yes / no | Are these two source occurrences connected?(inferred) | yes · no |
How it works
- Enter a passage and context (grade, knowledge, goal).
- Jev proposes active groups and connections per step.
- Compare estimated demand to the slot budget.
- Review groupings manually and export.
Strengths
- Fine-grained view of reading demand.
- Honest about validation limits.
- Runs locally with your own key.
Caveats
- Not a validated measure of working memory.
- Custom non-commercial licence; organisations need permission.
Getting started
Clone the repo, install Python 3.12 requirements and build the web app with npm ci && npm run build, then add your TypeSafe key to .env.At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- localhost web app (Node frontend)
- Has code
- yes
- Has tests
- yes
- License
- Working Memory Jev License (custom, non-commercial)
Use cases and integrations
How curated lists describe it
Edit \`.env\` locally and replace the placeholder with your own key. Keep this file private; Git ignores it. Do not paste the key into frontend code, a browser field, an issue, or a command that will be saved in shell history.
logicrw/awesome-jev-projectsLocalhost tool for educators that flags where instructional text may ask a reader to hold too many ideas at once. Its evaluation opens by naming the two tests its own model fails. Custom licence, not open source.
MrJev/awesome-jev
Frequently asked questions
What does it estimate?
Where a passage may overload a reader's working memory.
Are the numbers validated?
No, the README says they are not validated measurements or calibrated probabilities.
Can my school use it?
Organisations with more than one employee need written permission.
Related projects
Semantic mapping site that places phrases in 2D using Jev's scores for how strongly each relates to two chosen axis concepts.
Article- jev-samples★ 1
Runnable Jev samples: README checks, AGENTS.md readiness, PR triage, MovieLens next-watch ranking and issue triage, with measured runs.
ProjectbetaPython Analytics engine that rates published posts on pre-registered rubrics with Jev and tests ratings against real engagement.
ProjectbetaTypeScriptDocument evaluation CLI and CI gate using Jev or OpenAI Decisions, with pre-registered validation and documented reliability limits.
CLIbetaPythonLocal SvelteKit workbench and CLI for organizing, editing, running and comparing Jev experiments defined as JSON.
CLIbetaTypeScriptLearning lab for DeepEval with Jev as the judge: replayed evaluations, adjustable decision rules, release gates and drift alerts.
ProjectprototypePython