TutorLoom
The learning platform for companies, colleges and training institutes. Free and open source, with an Enterprise edition.
Vaulter watches what students ask Gemini and tells a counselor when something needs a human. Google Apps Script on a Google Sheet, built by Orono Public Schools.
EducationHTML · JavaScriptGitHub @Orono-Public-SchoolsOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
No updates detected yet. GitHub releases will appear here.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Watches student Gemini prompts from Google Vault, flags concerns with keyword rules and AI triage, and gives counselors a review dashboard. Google Apps Script on a Sheet.
Taken from the repository README
Loading the full README.
Read the full README ↗Vaulter watches what students ask Gemini and tells a counselor when something needs a human. It pulls Gemini activity from Google Vault, scans each prompt against keyword rules, asks Gemini whether a match is a real concern, emails the flagged ones, and gives reviewers a dashboard to work the queue. It is a Google Apps Script bound to a Google Sheet: nothing to host, nothing to buy.
Built by the technology department at Orono Public Schools, Minnesota.
Site: https://orono-public-schools.github.io/Vaulter · Live demo: the dashboard on sample data · Setup walkthrough: five stages, about an afternoon
Vault export → parse prompts → keyword rules → Gemini triage → email digest → dashboard review
A scheduled run asks Vault for the Gemini activity since the last run, waits for the export, reads
each prompt in memory and keeps only the ones that match a keyword rule. Each match goes to Gemini
for triage: likely false positives land in a log, everything else is written to the Flagged tab,
emailed to the counselors and waits in the dashboard queue. Students are masked in the dashboard
until a reviewer clicks Reveal.
Follow the setup walkthrough. In short:
Using an AI coding agent such as Claude Code or Gemini CLI? The walkthrough's first stage has a prompt to paste that lets the agent clone, create the Sheet with clasp, push the code and hand you the steps only a person can do.
apps-script/src with the same name (or push with clasp).setupSheets from the editor and approve the permissions. Eight tabs appear with defaults.Setup (OUs to import, schedule), App_Config (alert recipients), Alerts_Config
(keywords) and Access_List (who can open the dashboard). Add GEMINI_API_KEY as a script property.DASHBOARD_URL script property.installTrigger once, then press Run import now in the dashboard.Later code changes: Deploy → Manage deployments → edit → New version keeps the same URL.
With clasp: clasp push, then clasp deploy -i <deployment id> -d "<note>".
| Property | Required | What it does |
|---|---|---|
GEMINI_API_KEY | no | Enables AI triage at import and in the dashboard |
DASHBOARD_URL | no | The deployed web app URL, for the Sheet's Vaulter menu |
SPREADSHEET_ID | no | Use a Sheet other than the one the script is bound to |
VAULT_MATTER_ID, PENDING_EXPORTS | set by the script | The Vault matter it created and exports in flight |
| Path | What it is |
|---|---|
apps-script/src/ | The script. Config.js constants, VaultImport.js export and scan, Scanner.js rules, Alerts.js email, Scheduler.js triggers and the Sheet menu, SheetSetup.js tab scaffolding, WebApp.js dashboard backend, index.html dashboard UI, Demo.html sample data for demo mode |
apps-script/.clasp.json.example | Copy to .clasp.json with your script ID to use clasp |
docs/ | The site: landing page, setup walkthrough, and the dashboard running on sample data. node docs/build-demo.js regenerates the demo from the dashboard source |
Nothing in this repo is specific to one district. Per-install values live in Script Properties and
the Sheet; .clasp.json is ignored so your script ID stays on your machine.
cd apps-script
cp .clasp.json.example .clasp.json # paste your script ID
npm install && clasp login
clasp pull # fetch what's live; the script can also be edited in the browser
clasp push # send local changes to the script's HEAD
The dashboard is one HTML page with no framework. Colours and spacing are CSS variables at the top
of index.html. Demo mode (the Demo toggle, or ?demo=1) answers every server call from
Demo.html without touching the sheets, Vault, Gemini or email; the public demo is the same page
built statically. After a dashboard change, run node docs/build-demo.js and commit the result.
Vault_Data staging tab is cleared after every scan.appsscript.json.Keyword rules are blunt and need tending. The AI verdict is triage, not a decision. Runs are scheduled, not real time, because Vault exports take minutes to hours. It sees Gemini prompts only. Apps Script's six-minute execution cap is why exports are collected by a checker trigger.
MIT. See LICENSE.
Joel Mellor, Technology, Orono Public Schools · [email protected]