# Glossary

The vocabulary these docs use, one definition each: graph, entity, measure, lineage, confidence score, proposal, evidence, approval, ledger, governed context.

## The model

**Graph.** The one model Datapace keeps per estate: what each database means and what runs it. See [the graph](/docs/concepts/the-graph).

**Estate.** Every database a team is responsible for, across engines and environments, including the inherited and acquired ones.

**Entity.** A thing the business talks about, such as a customer or an order, resolved to one or more tables in one or more databases.

**Measure.** A number people ask for, with its definition: which column, which filter, which grain.

**Grain.** What one row of an entity or a measure stands for: one row per billing account, one row per order line.

**Relationship.** How two entities connect, whether or not a foreign key declares it.

**Semantics.** What a column and its values mean, including whether the column holds personal data.

**Lineage.** Where a value comes from and where it goes: the tables, views, jobs, exports, and dashboards that produce and consume it.

**Infrastructure facts.** What runs a database: instance class, storage, cost, performance, usage, freshness. Kept on the same graph as the meaning.

## Inference

**Signal.** What Datapace reads: schema metadata, workload signals, sample structures.

**Confidence score.** A number between 0 and 1 on every inferred fact, saying how much of the evidence points the same way.

**Inferred.** Proposed by Datapace with its score. Visible and reviewable, not yet trusted.

**Validated.** Accepted or corrected by one of your experts. Part of the confirmed graph.

**Confirmed graph.** The validated facts. What the agents draft on and what AI systems read.

**Working session.** A sitting in which experts accept, correct, or reassign inferred facts. A few sessions make a first pass. See [infer and confirm](/docs/concepts/infer-and-confirm).

**Living documentation.** The confirmed graph, kept current by the Documentation agent's proposals and the team's approvals.

## Work

**Agent.** One of the five Datapace agents: FinOps, Performance, Quality, Migration, Documentation. Each drafts proposals on the confirmed graph; none executes. See [agents](/docs/concepts/agents).

**Proposal.** A change with its case attached: what changes, what it affects, the evidence, the risk, and where it stands in approval.

**Evidence.** The signals and graph facts a proposal cites. A proposal without evidence is not ready.

**Blast radius.** What a change touches downstream, made visible by lineage.

**Mapping.** In a migration, a row that ties a source field to a target field, with a confidence score and a note. See [migration mapping](/docs/guides/migration-mapping).

## Governance

**Scope.** Which databases and environments Datapace may read, agreed with your team before anything connects.

**Policy.** The rules every action is checked against: which databases, which environments, which kinds of change, and who approves.

**Approval.** A person's decision on a proposal: approve, reject, or ask for more. Recorded with who and when.

**Audit ledger.** The one record of everything that happens through Datapace: who, what, on which database, the evidence, the approver, the outcome. See [policy, approval, and audit](/docs/concepts/policy-approval-audit).

**Governed context.** The way copilots, BI, and MCP clients reach production data: the confirmed graph, through the gate, inside policy and the ledger. See [governed context](/docs/concepts/governed-context).

**Gate.** The point every AI system reads through. Serves confirmed facts, applies policy, records the access.

**Raw access.** A connection string handed to an AI system. What governed context replaces.