Backfield¶
Backfield helps news organizations turn articles and reporting materials into durable, structured data. Instead of treating every story as a block of text, Backfield organizes it into entities, metadata, and relationships that can support new products, services, and understanding.
The Backfield platform consists of three interconnected applications:
- Agate lets users create customizable workflows that turn articles into structured data and gives editors a place to verify article-level results.
- Stylebook cleans, standardizes and enriches the entities found across your reporting.
- Backfield API makes the resulting data available for queries by location, entity, keyword, or meaning — and for products, services, tools, and story forms your organization builds.
Agate and Stylebook are primarily editorial interfaces. Backfield API is the application-facing part of the same platform: it turns the structured and curated work into something other tools and services can use.
The three apps at a glance¶
| Agate | Stylebook | Backfield API | |
|---|---|---|---|
| Think of it as | The processing workspace | The reference desk | The delivery layer |
| You use it to | Build flows, process articles, and review results | Maintain canonical people, organizations, locations, and their relationships | Query structured data and power products, services, tools, and story forms |
| Core idea | A flow that runs on articles | A canonical record that many mentions can point to | A project-scoped interface to Backfield data |
| Who uses it | Flow builders and editors | Editors maintaining shared knowledge | Developers, data teams, and applications |
How a story moves through Backfield¶
- You feed text into an Agate flow.
- The flow enriches the article, leaving behind a processed item. Editors can compare enrichments and extracted entities with the source text and correct article-specific mistakes.
- Confirmed entity data flows into the project's assigned Stylebook. Repeated references to the same real-world person, organization, or place can point to one shared canonical record.
- Backfield API exposes the saved data so your own applications can retrieve, search, and combine it.
A useful mental model¶
Backfield separates what a story said from what your newsroom knows:
- Processed items belong to a project. They include the story itself, its extracted and enriched data, and the editor's additions and corrections.
- Canonical knowledge belongs to a Stylebook. It includes the durable identity, metadata, geography, and connections your newsroom maintains across stories.
- Programmatic access comes through Backfield API. It lets other systems use project data without turning the API into a separate source of truth.
This distinction determines where to make a correction. Fix a mistaken extraction in the Agate processed item. Edit a person's newsroom-wide title, aliases, or relationships on the canonical record in Stylebook.
How the platform is organized¶
An organization is the top-level account. It contains workspaces, which group related projects. Each project is assigned one Stylebook, and a Stylebook may be shared by several projects in the same organization.
See Organizations & workspaces, Projects, and Users & access for the boundaries that control navigation, settings, and editing rights.
Where to start¶
| If you want to… | Go to |
|---|---|
| Get oriented on your first visit | Getting Started |
| See one story go end to end | Simple Example |
| Understand how accounts and projects are organized | Organizations & workspaces |
| Learn what a project contains | Projects |
| Understand how Backfield models your data | Data model |
| Process and review articles | Agate |
| Curate shared reference knowledge | Stylebook |
| Search data or power another product | Backfield API |
| Configure models, integrations, and keys | Settings |