Journal Picker
picker-ui calls picker-api through a service binding; Workers AI embeds the query and compact static shards supply evidence.
Verified architecture diagrams
The live path and the build path.
One view follows a user request through public experiences, Workers and managed services. The other shows how the shared content export becomes bounded runtime artefacts.
Combined runtime
Solid arrows follow the normal read path. The graph makes public experiences, Worker boundaries and data/AI services explicit.
The Journal Picker is represented here at system level; its existing architecture page contains the deeper request and data-lifecycle diagrams.
Offline materialisation
The shared snapshot fans out into four deliberately bounded build products, keeping corpus-wide analysis away from interactive requests.
The outputs differ in shape: compact picker shards, an AI Search index plus canonical records, topic snapshots and static graph bundles.
System context
These anchors are the architecture-diagram destinations used by the POC catalogue.
Portfolio Search Lab
search-test retrieves and reranks through AI Search, hydrates exact records from private R2 and optionally asks Workers AI for a grounded answer.
Author Graph POC
All graph analysis is offline. The production Worker publishes static HTML, JavaScript and precomputed JSON for browser-side rendering.
Topic Explorer
topics-ui calls a private portfolio-intelligence Worker, which reads exact materialised topic keys from R2.
Content foundation
The root Worker publishes documentation and three read-only R2-backed dataset hosts for metadata, open full text and BGE vectors.
Verification
The Mermaid sources were inspected as diagrams, verified structurally, rendered to SVG and text, and checked against their layout output.