FvRE v3.1 — Recipe → FVS App Converter (results)

Took the v3.0 recipe for Oscar, ran it through the converter, produced a FVS app spec in two modes.

What it does

Reads a v3.0 recipe, produces a FVS-compatible app:

Pages — the rendered page(s) plus discovered page list, each with content blocks
Content blocks — DOM nodes mapped to FVS block types (hero, nav, content, etc.)
Assets — image/font/video URLs ready for the FVS media library
Navigation — extracted nav structure
Design tokens — colors, fonts, spacing from CSS

Three modes

Mode	Output	Use case	Status
A — Static mirror	Rendered HTML + downloaded assets + fvs-app.json	Archival, backup, snapshot	done
B — Live edit	fvs-app.json with editable content blocks	Operator wants the content editable	done (spec)
C — Hybrid	Static shell + editable zones	Marketing sites, speed + selective edit	same code path

Test: Oscar

Source: wp-oscar.freshvibeapps.com (WordPress + Elementor)

Recipe: 7 pages, 28 assets, 10 interactions, 134 DOM root nodes

Converter output (Mode A):

1 rendered HTML file (341KB) at samples/fvs-oscar-mode-a.html
10 downloaded images at samples/fvs-oscar-mode-a/
FVS app spec at samples/fvs-oscar-mode-a.json

Converter output (Mode B):

FVS app spec with editable blocks at samples/fvs-oscar-mode-b.json

Sample of mapped blocks (Mode B, first 5 of Oscar's home page)

[
  {"type": "site-header", "source_role": "header", "fvs_confidence": 0.6},
  {"type": "nav", "source_role": "navigation", "fvs_confidence": 0.6},
  {"type": "modal", "source_role": "modal", "fvs_confidence": 0.6},
  {"type": "rich-text", "source_role": "content", "fvs_confidence": 0.6,
   "text": "Learn with Nature"},
  {"type": "rich-text", "source_role": "content", "fvs_confidence": 0.6,
   "text": "Explore the World of Trees"},
  ...
]

Block mapper (v3.1)

The classifier looks at the DOM role hint from v3.0 and maps to FVS block types:

DOM role	FVS block
hero	hero
header	site-header
footer	site-footer
navigation	nav
sidebar	sidebar
modal	modal
form-section	form
testimonial	testimonial-carousel
team	team-grid
pricing	pricing-table
features	feature-grid
gallery	image-gallery
card-section	card-grid
cta	cta-banner
content	rich-text

Known limitations (v3.1)

Hero detection — "Learn with Nature" is mapped to rich-text but should be hero. The v3.0 role inferrer needs visual signals (large text + image + button) to detect heroes.
Confidence scores are placeholder 0.6. The classifier should learn from the converter's corrections.
Asset import in Mode A only downloads the first 10 images. Need to do them all.
Forms, carousels are detected as interactions but not yet mapped to editable FVS blocks.
Multi-page crawls not implemented — only the entry page is captured.

How to use

# Scan a site
python3 fvre_v3_scan.py https://example.com --out recipe.json

# Convert to FVS app (static mirror)
python3 fvre_v3_convert.py recipe.json --mode A --out ./my-fvs-app

# Or convert to editable FVS spec
python3 fvre_v3_convert.py recipe.json --mode B --out ./my-fvs-app-edit

Next: v3.2 (CMS as the editor)

Now we have a spec format that describes the FVS app. v3.2 wires up the FreshCards editor so the operator can edit any of the content blocks inline. The spec stays the same — v3.2 just adds the editing surface.

← back to v3 design · ← back to index