Explore Data

The Explore Data screen (Discover menu — previously called "Explore Graph") shows, on a visual map, what exists in each context of the platform: the graph entities and relationships, the fields of the search indices and the relational tables — all driven by the sources declared in the context's registration, without writing Cypher or SQL.

Navigating — a 5-level drill-down

1. Root — the DATTA node in the center. Click to discover the contexts.
2. Contexts — click a context to open its data sources.
3. Data sources — each source declared on the context becomes a node, colored by engine: Neo4j (databases, one per origin — e.g. febraban, cvm, base-negativa, bcb, cnpj), OpenSearch (indices) and Trino (schemas). The panel lists the sources with counts (entities, documents, tables); click a source — on the map or in the panel — to open it.
4. Source objects:
   • Neo4j database → the entities and relationships of THAT database, with local counts; an empty database says so.
   • OpenSearch index → the fields on the map and, in the panel, the exact document count and sample data (first documents of the index).
   • Trino schema → the schema's tables; querying the data goes through a JDBC agent (panels) or DATTAX (FROM TRINO).
5. Records — in the panel of a Neo4j entity, the View records button opens the instances of that label as a graph: the first 100 records and the relationships among them. Click a record to see every property with its value and the list of relationships (type and the name of the other end, clickable); double-click or the Expand Node button fetches its neighbours from the database. Click a relationship to see its properties.

The breadcrumb (DATTA › Contexts › context › source › label) and the Back button go up one level at a time.

Filters at the Records level

The Filter bar above the map builds filters of the form target › name › property › operator › value — for instance Node › Empresa › uf › equals › SP. Each filter becomes a removable chip; several filters combine with AND. The operators are: contains, does not contain, equals, differs from, starts with, ends with, greater than, greater or equal, less than, less or equal, empty and not empty. Text is compared case-insensitively; greater/less compare as numbers when the typed value is a number (decimal comma accepted) and as text otherwise — handy for ISO dates.

Where a filter runs depends on what it filters:

• A filter on the open label (the one from View records) goes to the database: the query is re-run with the filter in Cypher, and the count shows how many records match across the whole store (the count stops at 10,000). That is how you find a record that is not among the first 100.
• A filter on a relationship or on another label (nodes that arrived via expansion) runs in the browser, over what is already loaded.

The same bar, with the same operators, exists in the chat's Graph tab — there every filter runs in the browser, because the answer's graph is the subgraph already loaded.

Where the sources come from

The screen shows exactly what the context declares in Configure › Contexts › Databases (the per-origin dataSources — see Contexts and the public negative list admin guide). A context with a single database keeps working as before (a single source at level 3); a per-origin context shows one source per database/index/schema.

Permissions

• GRAPH_EXPLORE — opens the screen and loads the per-context sources.
• GRAPH_QUERY_EXECUTE — Records level (listing and filtering instances) and node expansion.
• SEARCH_EXECUTE — OpenSearch index fields.
• PANEL_VIEW — Trino schema tables.

Without one of these permissions the corresponding section simply does not appear — the screen never breaks for lack of permission.