Product
Publishing and Discovery
Versioned public records changed under revision control, with the sitemap, machine index, search and redirects derived from the same records.

Customer controlled Linux host · Browser access
On this page
What it does
Publishing and Discovery holds the platform's public records (pages, systems, services, legal documents, publications) as versioned records, lets staff change them under revision control and derives the sitemap, the machine index, the search and the redirects from the same records so no address or title is typed in a second place.
What exists now
| Capability | What it does | Where |
|---|---|---|
| Publication editor | Staff inspect a record and supersede it with a new body, title or summary. The change names the revision it was made against; if the record has moved on, it is refused as a conflict and nothing is overwritten. | /publishing, staff with the publishing grant |
| Body rules | A body may use a closed set of elements and character references and is limited in size; anything else is refused before it is stored. | Publication editor |
| Redirects | Staff create or retire a redirect. Re-pointing closes the old row and opens a new one; a redirect is retired, never deleted. | /redirects/manage, staff |
| Sitemap and index | The sitemap, llms.txt and the route list are produced from the published records and the declared routes. | Sitemap |
| Search | Searches the published records by query, optionally narrowed to one record class. | Search, GET /api/search |
Operating needs
- PostgreSQL, which holds the records and their revisions.
- A staff member with the publishing grant, who is the only person able to change public text.
Limits
- Search matches the published records; it does not read the web or private records.
- There is no draft review workflow with several approvers: one authorised staff member supersedes a record and the change is recorded.
- It does not check that what is published is true; claim review is a separate duty.
Availability
Publishing and Discovery is running on our own platform. We do not sell a licence for it as a separate product today. The commercial terms state what is on sale. It can be part of the result of an engineering engagement. The written proposal then names it and its licence terms and what you would hold at the end. To discuss it, use Work with Us.
Evidence
Revision conflicts, body validation, redirect behaviour and the derived indexes are covered by the platform's own tests. The evidence register lists the graded claims.
Works with
- Audit Chain: Requires. Each change to a public record is recorded.
- Identity and Access: Requires. Only staff with the publishing grant may change public text.
- Document Search: Works with. Public search matches the titles and summaries of published records by substring; Document Search ranks an account's own documents with full-text search.
- Evidence Register: Works with. Claims appear on the public research pages.
Where it runs
Install on a customer controlled Linux x86-64 host with PostgreSQL 18. People use it through a browser that can reach the host.
Installation, updates and recovery
| Supported target | A customer-controlled Linux host on x86-64 with PostgreSQL 18 (linux-x86_64-customer-controlled). |
|---|---|
| Not yet supported | Windows, macOS, phones and tablets, and Linux on other architectures. Each needs its own installation, update and recovery evidence before it is supported. |
| Check a device | Run rjv preflight on it. It reads the operating system, architecture, memory, disk, listen address, configuration, database, database role and schema, changes nothing, and exits non-zero if the current release cannot run there. |
| Installed as | One application and a PostgreSQL database. rjv install lays the current release (schema 162) down with a hardened service unit and an environment file, and installs every system on this site, including Publishing and Discovery. |
| Needs | Nothing beyond the release and its PostgreSQL database. |
| Data | It stays in the PostgreSQL database you run. The only outside service the release can use is an outbound mail relay that you configure. |
| Updating | A release is a numbered build. The schema is applied by a separate owner step (rjv migrate), and the application refuses to serve a schema it was not built for. |
| Backup and restore | rjv backup writes a checksummed dump. rjv restore restores it into an empty database and checks the checksum, the schema version and the audit chain of every account. |
| Leaving | The data is in your own database and can be dumped at any time. What you may do with the release is set by the terms on deployment and ownership. |
On a phone or tablet
Use the browser with a connection to the host.
See what works on your device
- Find this offering
- Browser (W)
- Carry out every authorised operation
- Client (C)
- Compute on the device itself
- Not supported (U)
- Use a host you own
- Client (C)
- Work without the host or the internet
- Not supported (U)
Nothing here needs a desktop because every operation that this site offers works in the browser on a screen down to 320 CSS px which is the narrowest width that we test. Where a row reads Not supported the browser is the alternative because we do not install the release on the device and no offline mode exists and the host that you control must be reachable and we do not use device sensors or accelerators.
- Published
- 29 October 2026
- Revision
- 2