Skip to main content
Starting data is the baseline a version uses to prepare or reset synthetic Tools. It is different from the current records your agent has already changed in a running environment.

​
Edit or import records

  1. Open the source draft for your selected Tool version.
  2. If that exact source revision has no compiled version, publish it first so Firedrill can read its Tool contracts.
  3. In Starting data, choose the Tool copy and collection you intend to change.
  4. Edit records or import a local JSON array. Use synthetic or redacted data, not private production records.
  5. Preview the change. Schema errors are shown before anything is saved.
  6. Review the before-and-after source comparison and save the candidate revision.
Records use stable row IDs:
Use properties accepted by your selected Tool’s actual schema. This is not a universal record format for every Tool. Replacing a collection’s records removes omitted baseline rows from the new source; review the complete candidate before saving. The editor preserves the other collections, named Tool copies and unrelated source fields. It supports JSON and YAML world source without asking you to convert the project format. Local JSON import does not change the source until you explicitly preview and save.

​
What saving changes

Saving creates a new draft revision. It does not mutate a running Tool, reset an environment, approve executable code or activate the new version. Publish, review and activate the new revision when you want to use it. For a reviewed library Tool, a data-only revision can reuse the Tool’s reviewed behavior. Firedrill checks the original source, the new compiled version and the exact library package bytes. Tool behavior, browser apps, permissions and all non-data source must remain unchanged. The version shows Starting data reviewed, not a new test pass. Code or permission changes require their own executed validation and approval. Activating a version changes which starting data new sessions use. Existing sessions keep their state and version; their history is not rewritten. If another edit changed the draft, reload before previewing again. A new, unpublished revision does not silently reuse a previous revision’s schemas. Collections with ordered state actions or more than 500 records remain accessible through the source editor instead of being flattened by this editor. For live data changes, use the selected Tool’s supported operation or browser app. For restoring live state, see reset and time controls. Results and evidence remain separate from mutable Tool state. To keep several named sets of records and reuse them across scenarios, see saved datasets. Saving a dataset retains data; applying one creates a source revision; resetting a connection is a separate explicit action.