Skip to content

Reading & discovering

Read datasets directly, audit their provenance, list what you have, and discover shared reference data — all without writing SQL.

Start with list_memory

Before querying or re-ingesting, list what you already have — it avoids duplicating data and tells you the exact names to reference in SQL.

recall_dataset

Read a named dataset directly, without writing SQL.

Use this for a quick look at one dataset. To filter/join/aggregate, use query_memory; for provenance/lineage use inspect_dataset.

Parameters

Name Type Default Description
name str Name of the dataset to read.
mode "preview" | "schema" | "stats" "preview" preview = first rows, schema = columns/types, stats = row count + schema.
limit int | null null Max rows when mode=preview.

Returns — depends on mode: capped rows (preview), column names/types (schema), or row count + schema (stats).


inspect_dataset

Get full metadata for a dataset: schema, where it came from (source), tags, row count, and its lineage — which datasets and SQL produced it.

Use this to understand or audit a dataset before trusting or building on it. For the actual rows, use recall_dataset or query_memory.

Parameters

Name Type Default Description
name str Name of the dataset to inspect.

Returns — a MemoryObject: tenant, label, table identifier, source, schema, row/size counts, tags, timestamps, and lineage (parents + producing_sql).


list_memory

List the datasets currently in your memory (name, source, tags).

Start here to discover what's already stored before querying or re-ingesting — it avoids duplicating data you already have. Optionally filter by tags, then use inspect_dataset / recall_dataset to dig into one.

Parameters

Name Type Default Description
tags list[str] | null null Only return datasets carrying any of these tags.

Returns{datasets: [{label, source, tags, updated_at}, ...]}.


discover_reference_data

List the shared reference datasets available to every tenant (read-only).

Beyond your own private datasets, Memcove may expose shared reference data (e.g. market/reference tables) that anyone can query but no one can modify. Use this to see which shared schemas and tables exist and their columns, then read them in SQL by their qualified name, e.g. SELECT * FROM ref_market.prices.

Parameters — none.

Returns{schemas: [{schema, tables: [{name, columns: [{name, type}]}]}]}.

The shared schemas are configured by the operator (MEMCOVE_SHARED_SCHEMAS, default ["ref_market"]). See Tenant isolation.