Content storage layout

Everything content-shaped follows one convention. Learn it once and you can navigate, script against, and bulk-edit any Devani site.

The layout

site-root/
├── about/                      # a page: /about/
│   ├── content.html            # the page body (inside <main>)
│   └── meta.json               # SEO + status metadata
├── blog/
│   └── my-post/                # a post: /blog/my-post/
│       ├── content.html
│       ├── meta.json
│       └── featured-image.txt  # Tonta image id
├── products/
│   └── my-product/             # a product: /products/my-product/
├── header/content.html         # site-wide header region
├── footer/content.html         # site-wide footer region
├── global-styles.css           # the site stylesheet
└── devani/                     # the application itself
    └── data/                   # storage, navigation, snapshots

The three files per item

• content.html — the rendered body. Plain HTML using the site's canonical classes (.section, .container, .vs-btn…). No PHP allowed — writes containing <? are rejected.
• meta.json — status (draft/published), title, published date, author, and all SEO fields: meta_title, meta_description, meta_keywords, canonical, robots, open_graph {title, description, image}, twitter {card, title, description, image}, schema.
• featured-image.txt — just the Tonta image id. Devani derives size-appropriate URLs from it (e.g. _512 for cards).

Rules that keep tooling happy

• Status truth lives in meta.json. Index tables are derived from it — the filesystem is authoritative, and sync flows re-read meta rather than invent state.
• Slugs are lowercase letters, numbers, hyphens. A set of directory names is reserved (devani, header, footer, blog, products, assets, api, …) and tooling refuses to treat them as page slugs.
• Media ids, not URLs. Store the Tonta id; derive URLs. That keeps every size variant reachable and survives CDN-side changes.
• Don't hand-edit while the site is being written to. If you script bulk changes, go through the MCP tools — they snapshot before every write and keep index tables in sync.