The Content Graph

## The Content Graph — references as first-class edges The Content Graph project rebuilt how Barkpark thinks about relationships. Instead of treating a reference field as an opaque foreign key, every reference now materialises as a typed edge in a live graph — so content carries its connections wherever it goes. ### The problem A headless CMS is only as useful as the links between its documents. An author has posts; a page has a hero asset; a project has the case study that explains it. Historically those links were resolved one query at a time, and search returned isolated rows that gave no sense of how the site fit together. We wanted search results that arrive already connected. ### The approach We modelled the dataset as a directed graph where documents are nodes and reference fields are edges, dogfooded natively in Barkpark's own production dataset. Plain-slug references — the v1 model — resolve into edges at write time, so a page pointing its hero asset at a media slug, or a post pointing at its author, immediately produces a traversable relationship. The graph view became the visual dual of the Indx retriever seam: search finds the node, the graph shows the neighbourhood. ### The outcome Search results now read like a real, polished site. Click a project and see the pages that link to it; open an author and find their posts; land on a page and follow its hero asset straight into the Media collection. The Content Graph turned a pile of documents into a navigable web — and proved the model by shipping it to barkpark.cloud first.

Slug

the-content-graph

Client

Barkpark (core engineering)

Featured

Yes

Start date

2025-11-03T10:00:00Z

Status

completed

Updated

June 16, 2026

Created

June 16, 2026