Processes values (phone, email, address, company name, website, job title, or person name) through the Data Quality pipeline and returns the prepared results.

The request body is bulk and polymorphic: data is a single resource whose type selects the domain, and the batch of values is the items[] array inside its attributes. Because the request targets one domain, the batch is homogeneous by construction — items is typed to the selected domain and cannot mix domains. Each domain runs a fixed set of in-house default steps automatically — the caller does not name parse/validate/normalize. Additional work is added at the batch level via enrichments (per-domain signal list) and/or custom normalization (config), both on the resource's attributes; they apply to every item.

The response data is an ARRAY of per-item result resources, in STRICT POSITIONAL 1:1 ORDER with the request's items: data[i] is the result for items[i]. Each resource's id is that positional index (id == i) — the hard link back to the originating item. Each element's type is the OUTCOME: a successfully processed item is its domain resource (type = the request domain), whose attributes echoes the request input (confirmation, and to disambiguate duplicate input values), the parsed and normalized stage products, and the final prepared value; an item that failed to process is a PrepareError resource (type = PrepareError), whose attributes carries the failure detail. Success and failure are distinct resource types (value XOR error). Items are processed independently: a failed item still occupies its ordinal slot while the rest of the batch succeeds (the data length and order always match items).

Error classes:

• WHOLE-REQUEST 400 — the request is malformed as a whole and nothing is processed: an empty items[] or one exceeding the item cap, or a body that does not parse. (A mixed-domain batch is not possible — items is typed to one domain — so it does not arise as a runtime error.) Batch-level config violations the schema does not enforce are also rejected here: supplying both inline rules and a configId, or a normalization-rule value that is not legal for its type.
• PER-ITEM PrepareError resource — the request was well-formed but an individual value could not be processed; that item's data[i] is a PrepareError resource (failure detail in its attributes) while the rest of the batch succeeds (partial success, still a 200). This is a processing error, distinct from validation: a value that processes but does not pass validation is a domain success resource reporting isValid: false (for the domains that validate — address has no validator and omits isValid), NOT a PrepareError resource. Item-level input combinations the schema permits but does not enforce fall here too: an item that supplies both its raw string and components, or neither (or a components with too little to process), becomes a PrepareError resource at its slot.
• 429 — the request was rate-limited; retry with backoff.

This operation is synchronous: values are supplied inline and prepared results are returned inline. No records are persisted.