REVIEWABLE VENDOR IMPORTS
Map a vendor CSV to your application's contract
Start with the field meanings, then validate every record before buying an artifact.
Give the import an explicit destination
A vendor's account_code may correspond to your customer_id. The names suggest a candidate; a reviewer must establish whether they mean the same thing.
Use this repository-owned synthetic source. The identifier is text, even though it contains only digits.
account_code,givenName,enabled 000401,Ada,true 000402,Grace,false
The target is a supported, flat JSON Schema subset:
{"type":"object","additionalProperties":false,"properties":{"customer_id":{"type":"string"},"first_name":{"type":"string"},"active":{"type":"boolean"}},"required":["customer_id","first_name","active"]}
Review, then validate
- Upload a UTF-8 CSV, JSON or JSONL file and paste your target contract.
- Review every correspondence and cast. For this synthetic fixture choose account_code → customer_id, givenName → first_name, enabled → active.
- Approve those choices for the named vendor and application domain.
- Validate all actual records. A failed record prevents checkout.
Preview and validation are free. A $9 purchase downloads your approved mapping, a Python runtime wrapper and a full-record validation receipt. The ZIP contains schemas and decisions, not transformed source records. Python 3.12+ and the free matching runtime are required to run the wrapper.
Run the same workflow locally
The free Apache-2.0 CLI exposes infer, map, review, resolve, approve, transform, validate, generate and drift. Begin with python -m mcp_products.schema_mapper --help. Approval must bind the exact reviewed fingerprint and the intended user/vendor/domain scope.
The hosted source limit is 1 MiB. Nested values, date/locale conversions and unsupported JSON Schema keywords are refused. This example is synthetic; it is not customer validation or a claim of accuracy.