List the entities (tables) you may read, plus the available metrics.

Call this first. You never write SQL -- you pass names from here to get_records / get_metric and sql-steward compiles the query for you.

Show one entity's fields, types and PII tags.

Fields tagged with a blocked PII category are marked blocked: true; asking for them in get_records is refused before any query runs.

List pre-approved metrics and the dimensions/filters each one allows.

Read rows from one entity.

filters is a list of {field, op, value}. Operators: =, !=, <, <=, >, >=, like, in, not in, is null, is not null. Blocked-PII fields are refused before anything runs; cross-entity references without a defined join are refused as unreachable.

Compute a pre-approved metric, optionally grouped/filtered by allowed dimensions. The aggregation itself is fixed by the semantic layer.

Vector similarity search over an entity's embedding column (pgvector).

query is embedded locally and matched against the entity's configured embedding; the closest rows are returned (the embedding itself never is). Requires the entity to have a search config and a local embedding model (SQL_STEWARD_EMBED_URL). PostgreSQL-only. Same PII refusal as everything else.

List the declared data-quality checks the layer can run.

Run the declared data-quality checks and return a readiness summary.

Each check compiles to a read-only violation count; zero violations passes. Returns a readiness score (percent of checks passing), an overall status, and a per-check breakdown. An 'error'-severity failure makes the status 'failing'; a 'warn'-severity failure makes it 'degraded'.

Verify the tamper-evident audit chain (agent-blackbox), if enabled.

Reports whether any previously recorded call was altered after the fact.