What technical setup should a reseller prepare to start an intake pilot?

Once scope is agreed, a reseller can set up the technical side of a pilot in a few concrete steps.

- **Request a manifest.** It should list schema version, tables, keys, split assignments, generation parameters and scorer version, so each catalogue entry maps to one release identifier. - **Ask for schema files.** JSON Schema for JSONL and column definitions for Parquet, so you can validate on ingest and reject malformed files automatically. - **Define catalogue metadata fields.** Useful ones include `domain`, `components`, `schema_version`, `case_count`, `contains_free_text`, `contains_policies`, `licence_reference` and `generation_notice`. Fill `generation_notice` with a statement that the data is generated. - **Set up a staging area.** Load the pilot sample into a staging bucket separate from your live catalogue, and run your own validators there first. - **Decide on updates.** Ask how new releases will be versioned and whether column names stay stable. Plan for a change note you can forward to buyers. - **Wire in evaluation.** If the pilot includes a task pack, run the reset and scorer in your own environment and confirm that repeated runs from the same seed give an identical starting state. - **Record provenance.** Keep the release identifier, your validation results and the licence reference together in one record.

None of this requires a fixed delivery method. Agree transfer arrangements in scoping.