Skip to content
Zevqi

Changelog

What shipped

Small releases, shipped often. Planned work lives on the enterprise roadmap.

  1. November 2026

    Clearer website analytics

    • Public pages now record time on page, which sections were seen and when a form is started, so we can see what visitors actually read. Never form values or copied text
    • The privacy page lists exactly what is measured, and signed-in and private pages are left out of the new tracking
  2. November 2026

    No scores for people

    • The People page no longer shows pass rates, levels, XP or streaks. It lists, in alphabetical order, the files each person has reviewed, the files where they are the only reviewer, and how many reviews they have completed
    • The activity feed and file history no longer show a person’s quiz score. Streaks and levels stay private on each developer’s own machine
    • The dashboard is for finding files nobody understands, not for grading developers
  3. November 2026

    The product is now called Zevqi

    • The product name is Zevqi, the same as the company. The old name, zlearn, was shared by several unrelated products and was hard to find
    • The command-line tool is now the npm package zevqi, with the command zevqi: npm install -g zevqi; zevqi setup --key <your-team-key>
    • The old zlearn command still works. The package installs both, so nothing you set up before stops working. The old package zlearn-cli points to the new one
    • Settings can use ZEVQI_ names (ZEVQI_TEAM_KEY and so on). The older ZLEARN_ names are still read
    • Author ids are described accurately as pseudonymous: a one-way hash of the email, not an anonymous id
    • Owners can now delete their team and its data from Settings, and a refund and cancellation policy page is live
  4. November 2026

    One command to start, and a map from git

    • One paste installs Zevqi, joins your team, turns it on and scans your repository: npm install -g zevqi; zevqi setup --key <your-team-key>
    • Knowledge map from git (zevqi scan): who holds the knowledge of each file, files with one author, owners who have left, and AI-assisted commits with no review. No AI and no API key. zevqi scan --dry-run shows exactly what would be shared
    • Attested reviews with no model: read a change and describe it in your own words. Not graded, and your words stay on your machine
    • A model key is now optional. It is only needed if you want AI-written, diff-specific questions
    • Lessons can be written by your own coding agent through AGENTS.md (early access, not yet verified in live Codex or Cursor)
    • A live setup page on your dashboard: four steps that update by themselves as your first scan and review arrive, each with a command to copy
    • Copy buttons on every command, across the docs and the dashboard
    • The weekly email and Slack summary now list the decisions for the week, each linking to the right dashboard page, and never include full email addresses
    • A table of every way Zevqi can learn about your code, what each shares and what exists today, in the docs and on the security page
  5. November 2026

    Evidence you can trust

    • Review evidence now expires: a review counts for one version of a file, and goes out of date when the file changes or after 90 days
    • Developers say how they feel after the questions: understood, need clarification, or need further investigation. Only "understood" counts as evidence
    • Every flag on the dashboard states its reason in plain words. There is no composite score; risk levels follow fixed, documented rules
    • Knowledge coverage shows its formula and what it does not measure
    • New AI changes view with a detail drawer, and decisions for the week with the next step for each
    • zevqi review: review recent committed changes to a file, so a second person can cover it
    • Content fingerprints (git blob hashes) sync with each lesson; source code still never leaves the machine
    • Questions now cover why, failure cases and assumptions, and lessons list decisions, dependencies and what could break
    • Pilot checklist and an in-dashboard feedback form
    • The pull-request gate reports out-of-date files and the reason for each
  6. November 2026

    Email sign-in, team plans and checkout

    • Slack weekly summary through an incoming webhook (connect it in Settings)
    • README badge showing your team’s understanding score (signed link, one percentage, nothing else)
    • Pull-request gate: `zevqi check` fails CI when a changed file has AI changes nobody has understood, with GitHub annotations
    • Weekly team summary and a trial-ending reminder by email, with one-click unsubscribe
    • Passwordless sign-in with a one-time email code: reach your dashboard and billing from your account, no link needed
    • 7-day card-free trial for up to 5 developers
    • Prepaid Team plans through Razorpay (UPI, cards, netbanking): the team is created the moment payment succeeds
    • Seats and paid-until date enforced on the server, with renew and add-seats from the dashboard
    • Hosted team dashboard: members, blind spots, single points of knowledge and coverage by area
    • Scores-only sync with an offline outbox, so lessons are never lost when you are offline
    • Team keys and dashboard tokens stored as SHA-256 hashes
    • Support for Supabase secret keys (sb_secret_)
  7. November 2026

    Zevqi CLI 0.1

    • Claude Code hooks: capture, explain in the background, commit gate
    • Terminal quiz and local web dashboard on 127.0.0.1
    • Bring-your-own key for Anthropic or any OpenAI-compatible endpoint
    • MIT licence, open source