Output nodes¶
Output nodes decide what happens to a flow's results. Every flow ends with one.
| Node | What it does |
|---|---|
| Output | Presents the combined results as JSON for viewing — useful for inspection and testing |
| Backfield Output | Saves results into your project, including matching people and places into your Stylebook catalog |
| S3 Output | Writes the results back to Amazon S3 cloud storage as files |
JSON Output¶
JSON Output keeps the combined result available for inspection in the run and processed item's JSON views. It is useful while designing a flow or when another system will consume the result without saving Backfield articles and entities.
Backfield Output¶
Backfield Output is the bridge into the rest of the Backfield platform. It saves article content, metadata, images, custom records, and entity evidence in the project when the flow produces them. A resolvable article body alone is enough to save the article — for example, Text or JSON Input wired directly to Backfield Output in a metadata-only or passthrough flow. People, organizations, and locations are matched against the project's assigned Stylebook when those domains are present.
As it saves, canonicalization may link an existing record, create a new record when policy allows, or send an uncertain candidate to editors. See Canonicalization.
Its settings can also enable supported semantic indexing and automatic connection inference. Automatic connection creation defaults off and can use a dedicated model separate from Stylebook matching. These features depend on compatible project models and the corresponding upstream data. Connection status may appear in Agate, while the canonical Stylebook record remains the place to edit a saved connection.
The output also defines how a rerun reconciles machine-generated evidence that was already saved:
| Policy | Conceptual effect |
|---|---|
| Add Only | Add new results without treating omissions as removals |
| Smart Merge | Update compatible machine output while retaining prior data where appropriate |
| Replace | Treat the newly emitted machine domains as authoritative, including empty results |
Replace can retire prior machine-generated associations that the new run no longer emits. Editor-added and editor-modified evidence is preserved.
The Stylebook is inherited from the project and is not selected on an individual output node.
S3 Output¶
S3 Output writes the structured result to an Amazon S3 bucket and prefix. It is useful for downstream data pipelines or archives that consume files instead of the Backfield database.
The node needs S3 credentials from Integrations. After review, the processed item's JSON tools can synchronize reviewed output back to the configured object when that output path is available.
Choosing an output¶
Use JSON Output for inspection, Backfield Output for the full editorial and Stylebook workflow, and S3 Output for file-based delivery. A flow's output choice determines which review and downstream features are available; merely extracting an entity does not save it into Stylebook.