Synthetic data for everyone — create realistic test data in the browser with a complete, searchable Faker catalog.

Synthora is a static GitHub Pages product that combines:

• build-time Python introspection (truthful coverage metadata)
• in-browser Faker generation via Pyodide (no backend)
• optional plugin packs (explicitly labeled)

Live site: https://arvind3.github.io/fakerUI/ Repository: https://github.com/arvind3/fakerUI

Current status: production-ready UI with full feature set.

Phase 2 UI/UX shipped:

• Full-page viewport layout — Explore page fills the browser window, two panels scroll independently
• Method cards replace the old table — Try button always visible on every card, no horizontal scroll
• 3-step guided flow: "Search → Choose provider → Click Try" with live step indicator
• Synthora rebrand: DM Sans + Instrument Serif typography, emerald (#0A6E5C) + Signal Gold (#F0A500) palette
• Dark Midnight hero landing page with large Synthora logo, Instrument Serif headline, and single CTA
• Consistent 1100px max-width layout shell across all pages, 40px horizontal padding
• Templates page with industry-ready schemas across 5 domains (Retail, Azure DevOps, ServiceNow, CRM, General)
• Improved TryPanel with skeleton loader, copy button, and collapsible Python code snippet

Core functionality is stable and covered by unit tests (Vitest) and E2E smoke tests (Playwright).

• QA and engineers who need realistic sample data quickly
• product teams building demos, tests, and mock datasets
• non-technical users using the Schema Builder for CSV/JSON exports

• Browse all bundled Faker providers/methods for a pinned version
• Fuzzy-search methods and provider modules
• Generate sample data per method (locale, count, seed, unique mode, kwargs)
• Build table-like datasets with multiple faker fields (Schema Builder)
• Export results as JSON/CSV
• Work offline after first load (service worker cache)

In this project, "no gaps" means:

• all built-in formatters for the pinned Faker version are cataloged
• all enabled plugin pack formatters are cataloged
• runtime only allows calling methods in that generated whitelist

Scope labels shown in UI:

• Built-in providers (Faker vX.Y.Z)
• Plugin Pack: name + version
• Not included: arbitrary third-party/custom providers not bundled

See docs/coverage-policy.md for full policy details.

1. catalog_build/build_catalog.py introspects Faker + enabled packs and writes:
   • web/public/catalog.json
   • web/public/runtime-manifest.json
2. React app (web/) renders Catalog, Try Panel, Schema Builder, Coverage page.
3. Web worker (web/src/workers/fakerWorker.ts) lazy-loads Pyodide and runs generation safely.
4. Service worker (web/public/sw.js) caches app + catalog + runtime assets for offline use.

Prerequisites:

• Python 3.12+ (python --version)
• Node.js 20+ (node --version)

From repo root:

python -m pip install -r catalog_build/requirements.txt
python catalog_build/build_catalog.py

cd web
npm install
npm run pyodide:fetch
npm run wheels:fetch
npm run dev

Then open the local URL printed by Vite (usually http://127.0.0.1:5173/).

From repo root:

python3 -m pip install -r catalog_build/requirements.txt
python3 catalog_build/build_catalog.py

cd web
npm install
npm run pyodide:fetch
npm run wheels:fetch
npm run dev

When changing Faker version, plugin packs, or runtime logic:

# repo root
python catalog_build/build_catalog.py

# web/
npm run pyodide:fetch
npm run wheels:fetch
npm run build

Build is intentionally blocked if required generated assets are missing.

From repo root:

# Python catalog tests
python -m pytest catalog_build/tests -q

From web/:

# Unit tests
npm run test

# E2E smoke test
npx playwright install chromium
npm run e2e

Live deployment smoke check:

cd web
npm run live:check

Override live URL:

PowerShell:

$env:LIVE_URL = "https://<username>.github.io/<repo>/"
npm run live:check

Bash:

LIVE_URL="https://<username>.github.io/<repo>/" npm run live:check

Deployment is automated by .github/workflows/pages.yml on push to main/master.

Pipeline:

• install Python + Node dependencies
• build catalog
• run Python tests + frontend tests
• fetch Pyodide runtime + wheels
• build web/dist
• deploy Pages artifact

VITE_BASE_PATH is set automatically for project-site routing.

• strict method whitelist derived from generated catalog
• regex validation of method names
• kwargs must be valid JSON object
• no eval of arbitrary Python in UI path
• best-effort network blocking in Python runtime
• no backend API required for generation

• catalog_build/: build-time catalog + wheel tooling + Python tests
• plugin_packs/: plugin manifest (enabled/disabled packs)
• versions.lock.json: pinned Faker/Pyodide/plugin versions
• web/src/: React UI + worker client/runtime bridge
• web/public/sw.js: offline caching service worker
• web/e2e/: Playwright smoke tests
• docs/: policy, release notes, product copy, acceptance checklist

• Disable Windows App Execution Aliases for python.exe and python3.exe, or
• use full path to Python, or
• ensure installed Python is before WindowsApps in PATH.

Run:

python catalog_build/build_catalog.py
cd web
npm run pyodide:fetch
npm run wheels:fetch

Expected. Pyodide and wheel assets are large and only fetched once, then cached by service worker.

Usually means catalog/runtime assets were not generated before build/deploy.

Perspective pages deployed alongside the app on GitHub Pages:

These are static HTML pages in web/public/ — Vite copies them into web/dist/ automatically as part of the standard build.

• Coverage policy: docs/coverage-policy.md
• Acceptance checklist: docs/acceptance-checklist.md
• Release/versioning notes: docs/release-versioning.md
• Product copy snippets: docs/product-copy.md