Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

Pyxis: product analytics without cookies, IP addresses or personal data

Описание и применение

Pyxis – это открытая система аналитики, которая собирает данные о пользователях без хранения cookie, IP‑адресов и персональных данных, обеспечивая подробный анализ поведения пользователей и соблюдение конфиденциальности.

Задача

The author needed product analytics for BolsoVerde without storing cookies, IP addresses, or personal data, and without relying on third‑party tools that require user consent.

Результат

Pyxis enabled the author to collect detailed product analytics without storing cookies, IP addresses, or personal data, while keeping all data within the existing infrastructure and respecting user privacy signals.

Рабочий процесс

  • 1. Generate a random session ID in sessionStorage for each visit.
  • 2. Template the page URL by replacing identifiers with ":id" and removing sensitive segments.
  • 3. Drop query string parameters except UTM fields and flag ad click sources.
  • 4. Read IP address only to determine country/state, then discard it.
  • 5. Classify and discard user agent information; block automation browsers.
  • 6. Filter event properties to remove emails, phone numbers, CPF/CNPJ, and other personal data.
  • 7. Respect Do Not Track and Global Privacy Control signals before queuing events.
  • 8. Send events to the Pyxis API, which re‑templates paths and enforces the same privacy rules.
  • 9. Run a daily job to delete data older than 13 months.
  • 10. Use the analytics dashboard to view funnels, feature usage, and failed requests per visit.

Доказательства

I wrote Pyxis, open‑source product analytics that writes no cookie, keeps no IP address and keeps no personal data.

Источник

The function that builds the template applies it to each segment and replaces anything that is an identifier with ":id".

Источник

The API templates every path again with the same rules, drops properties that look like personal data and refuses the whole event when the user id looks like an email address or a document number.

Источник

  • Дата записи: 2026-10-24T15:11:18.974Z