Product terminology
pChronicle uses a small, consistent vocabulary in the CLI, Web UI, and task-oriented documentation:
| Term | Meaning |
|---|---|
| Dataset | One Agent trajectory store identified by a path (local path or object-store URI). |
| Source file | One input file or storage source inside a Dataset. |
| Run | One stored Agent execution record. A Session ID identifies the record; an optional Run ID may group it with runtime activity. |
| Step | One ordered user, agent, system, or tool-related entry in a Run. |
| Event | A low-level recorded or reconstructed item linked to a Step. |
| Result | Rows returned by a query or analysis. |
| Snapshot | The write/read sync protocol after a path is opened: which Sources exist and which version each is pinned to for one operation. |
The UI uses Recorded events for events captured directly from a runtime and Reconstructed events for events derived from imported run data. It never labels reconstructed events as raw data.
The following terms are reserved for technical documentation and APIs:
- Storyline is the normalized storage and exchange model behind Runs and Steps.
- Canonical Event is the append-only recorded event model.
- Directory is the optional platform locator that resolves a name to a path
and issues a ticket. CLI flags and HTTP paths may still say
catalog. - Snapshot (in APIs, sometimes still
DatasetCatalogSnapshot) pins Source membership and versions. It is not the Directory listing. - projection, revision, fragment, and column page describe storage and consistency mechanisms, not primary user workflows.
Dataset pins (@name), Warehouse mount names, and Directory library names are
locators. After resolution the engine only sees a path.
Older API paths and schema fields may retain these technical names for compatibility. User-facing labels follow the simpler vocabulary above.
See the pChronicle concepts and Discover and query for how these terms appear in workflows.