尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Turned my AI wellbeing app into a B2B product for HR teams: 12 prioritised requirements, privacy-first team reporting (no teams under 10), a 6-week pilot with KPIs agreed up front, funnel and A/B test tracking, and a landing page.
翻译准备中
取自仓库 README
正在加载完整 README。
查看完整 README ↗Your standup has a start time. Give it an ending.
Viraam is the last 2½ minutes of a meeting a team already holds. A team lead pastes one line into the standup invite. As the meeting ends, everyone opens the link and starts the same 2½-minute close in the same minute: stand if you can, move, rest the eyes, and finish "When this ends, I'll first…", in English, हिंदी or मराठी. On Friday the lead gets one line: how many days it ran. HR's whole involvement is one yes and one invoice; there is nothing to run.
Live product · Create a team link · Try the 2½ minutes · Pilot & pricing · Trust & data · Product case study
Status: pre-revenue and pilot-ready. The first six teams run free. All figures on the site are sample data from a fictional company, and every page says so.
Calendars can shorten meetings; they can't end them. Outlook's "end meetings early" setting shortens the slot, but doesn't help when people run past it (Petri). Viraam works in the meeting's last minutes: it ends the meeting together, and everyone writes their next step. A short ready-to-resume plan reduced attention residue and improved decisions on the next task in four studies (Leroy & Glomb, 2018). Microsoft's small EEG study found stress spikes at the transition between back-to-back meetings (WorkLab). No claim is made that the close improves performance or reduces stress.
| Stage | What it was | Why it changed | See it |
|---|---|---|---|
| AIRA → Stillwater | A workout companion, then short desk breaks with safety rules | People choose by problem ("stiff neck"), not workout type | — |
| Lull → Sukoon | An individual desk-break app with an HR layer: type how you feel (EN, Hinglish, हिं, मरा), one guided break, team totals only | Relies on people remembering; a break reminder is already included in many Microsoft 365 plans | Live archive · branch sukoon |
| Viraam | The last 2½ minutes of a standup the team already holds | — | lull-hr.vercel.app |
| Finding | Consequence |
|---|---|
| Break reminders are already included in many Microsoft 365 plans (Viva Insights), and they are private to each person by design | A personal reminder is not a product. A team's shared moment, with a record the lead can forward, is. |
| Links dropped in a channel depend on memory and fade after week one | Attach the break to the standup that already happens daily. The meeting is the reminder. |
| Per-seat pricing put a timer at par with complete HR software | ₹1,499 per team a month: less than a team lunch, one invoice, no seats. |
| HR teams are overloaded with admin | No rollout, no accounts, no chasing, no dashboard. Unused teams stop by themselves. |
| Meetings overrun anyway | It does not add 2½ minutes to the day; it takes them out of the overrun. |
/start (team name and standup end time). Out comes one invite line, a Google Calendar link, an Outlook .ics, a WhatsApp share and a QR code./close/?team=…&at=HHMM). Everyone who opens it sees the same countdown and starts in the same minute; anyone late joins at the right step. The link hosts it, not a person.| Time | Phase | On screen | Evidence |
|---|---|---|---|
| 0:00–0:20 | Stand · breathe | "Stand up if you can", one slow sigh | Active > passive seated breaks (systematic review, 2022) |
| 0:20–1:20 | Move: a different set each weekday | Mon neck/shoulders/wrists · Tue upper back · Wed legs · Thu hands · Fri whole body; 20 s each | Micro-break meta-analysis (Albulescu et al., 2022) |
| 1:20–1:40 | Eyes | Look across the room, blink slowly | 20-20-20 study (Aston University) |
| 1:40–2:20 | Plan | "When this ends, I'll first…" (never stored) | If-then plans, 94 tests, d = .65 (Gollwitzer & Sheeran, 2006) |
| 2:20–2:30 | Done | "Back at 10:15." · "Leaving with:" card · "Tomorrow: legs" | — |
Standup ran over? If the team opens the link with under 90 seconds left before the meeting's end, the close compresses to a 30–45 second landing (one slow breath, then the plan) and still finishes on the boundary, so a late standup never means a skipped close.
The shape never changes; only the move block does. The page reads the day itself (Indian time), so everyone on a team gets the same set; weekends use Monday's.
The promise is deliberately narrow: less fatigue, rested eyes, and a clear first step. No productivity or pain-relief claims (see /safety).
src/shared/pilot.ts). What people type, how they feel, and every safety stop stay on the device.src/, kept as history) understood about 900 phrases across English, Hinglish, Hindi and Marathi without AI, with an optional AI layer; safety always stayed on fixed rules. The close itself needs no typing to understand.| Path | Audience | What it is |
|---|---|---|
/ | Team leads, HR | The pitch: the close, how it works, and HR's "nothing to run" |
/start | Team leads | Invite generator: one line, Google/Outlook/WhatsApp, QR |
/close/ | Employees | The close: shared countdown and the 2½-minute timeline, in English, Hindi and Marathi |
/workspace | HR | HR dashboard: teams, seats, hidden teams, Friday email, six-week history, invoice |
/report | HR, Finance | Week-6 report, print-ready as a one-page A4 PDF |
/pilot | HR, Procurement | The offer, the six weeks, launch kit, pricing, procurement, FAQ |
/trust | HR, IT, Legal | Data map, under-10 rule, roles, subprocessors, retention, DPA summary, status |
/safety | Everyone | Red flags, when not to use it, review status, what the evidence supports |
/pm | Product managers | Case study: problem, bets with kill criteria, metrics, what was not built, iteration log |
/changes | Everyone | Dated decisions and the reasons behind them |
npm install
npm test # typecheck, build and run the test suite
npm start # the app on 127.0.0.1, offline mode (no AI)
npm run site:app # rebuild the hosted demo into site/app
Serve site/ with any static server to preview the marketing site. In production, Vercel strips the .html extensions.
| Folder | Contents |
|---|---|
site/close/ | The close (current product): one self-contained page, no AI, no accounts |
src/server | The earlier individual-break engine (v1–v4), kept as history: five breaks, safety rules, language understanding (lull.ts), offline responder (mock.ts), optional AI providers |
src/web | The employee app, translations (i18n.ts, i18n-strings.ts), reminders, pilot counting |
src/shared | Contracts, the screen document model, pilot allow-list, team insights |
site/ | The marketing site, HR dashboard, report, trust, safety and pilot pages, plus the hosted app |
tests/ | ~100 automated tests: safety in every language, flows, translations, pilot rules, site syntax |
docs/ | BRD, case study, one-pager and pilot proposal as PDFs with their HTML sources; earlier versions in docs/archive |
site/trust.html#status is the single source of truth for what exists, including what does not exist yet: no paying customers, no lawyer-reviewed MSA, and a native-speaker review of the Hindi and Marathi safety screens still due. site/changes.html records every change and why it was made.
Built by Rupak Ghanpathi, product, research, safety design and build. The project was previously named Lull (see /changes for why it was renamed); some file and repository names still carry that history.