Skip to main content
Mappings is where the context layer is actually shaped. Sources name things their own way. In Mappings, you decide that COCA COLA CLASSIC 330ML and Coke Classic 330 are both your one canonical entity for Coca-Cola Classic. Once mapped, every AI tool that queries your data via MCP rolls those values up the same way. You curate one concept at a time and can select the concept from the switcher at the top. The module has two tabs.

​
Canonical entities tab

Your master records for the selected concept. Here, you can see every canonical entity with its attributes and how many sources map to it. You can create a new entity, edit an entity’s name and attributes, and deprecate one you no longer use.

​
Mappings tab

The curation queue has two parts:
  • Proposed — raw source values waiting for a decision.
  • Approved — the ledger of decisions already made.
Each proposed row shows the source it came from, the raw value, a suggested entity, the volume (how many rows carry that value), and who proposed it: Each row also carries a status:
  • Pending — the AI hasn’t evaluated this value yet. It has no suggested entity or confidence, and it’s left out of bulk-approve and one-click approve until it does.
  • New — evaluated, but no confident match exists; needs a canonical entity chosen or created.
  • Review — one confident suggestion to confirm.
  • Conflict — more than one candidate; pick the right one.

​
See why the AI suggested it

Every AI suggestion is transparent. Open a proposed row to see the AI’s reasoning in its own words, the exact source data it looked at, and the runner-up candidates it considered. The AI only ever sees descriptive attributes — product names, pack sizes, categories — never your sales or other measures.
Stores and depots map themselves. Retailer data already carries their proper names, so Corvera approves them automatically and they never enter the review queue. The queue is where your products get curated.

​
What you can do

  • Approve a proposal onto its suggested entity or pick a different one.
  • Create a new canonical entity inline and approve the value onto it in one step.
  • Reject a proposal that shouldn’t map.
  • Remap an already-approved value to a different entity.
  • Bulk-approve many proposals at once.
Curating mappings — approve, reject, remap, create, and bulk-approve — requires a role with the Edit mappings permission (admins always have it). Everyone else can view the queue and the ledger but not change them. See Governance for roles.
Curation is available to your AI tools so the same actions via the UI are also available to agents over MCP. The UI and your agents draw on one governed mapping path. As fresh sources arrive, additional “New” proposals will surface here.