Results And Inspection

Use result rows, plan metadata, and provenance to understand an answer. When your deployment includes the semantic console, the panels below organize that information. API clients can inspect the corresponding response fields and use EXPLAIN or PROFILE directly.

Data

The Data tab shows the returned rows. For semantic answers, fields come from the selected semantic plan and output projection. For direct Cypher/SPARQL, fields come from the query’s RETURN or SELECT projection.

Provenance-like fields such as source_row_hash may be compacted visually. Hover or focus the cell to inspect the full value.

Visuals

Visuals are derived from returned rows. They do not change query semantics.

Common visual types:

  • sortable/searchable tables;

  • small graph evidence views;

  • bar charts or trend charts for aggregate answers;

  • heatmaps when a two-dimensional rollup is present.

If the visual does not match the question, check the Data and Query tabs first. The visual may be faithfully rendering rows from an unsuitable plan.

Reasoning

The Reasoning tab explains:

  • selected semantic family;

  • route or template;

  • constraints ShrikeDB recognized;

  • unsupported boundaries or clarification needs;

  • assumptions that need confirmation before interpreting the result.

Knowledge

The Knowledge tab shows the profile and evidence basis. Useful fields include:

  • profile ID and hash;

  • ontology expansion or grounding evidence;

  • active lakehouse mapping IDs;

  • row-group and source hashes;

  • validation, fallback, or skip diagnostics.

Query

For natural-language answers, the Query tab shows typed semantic plans and Cypher-equivalent inspection output. That does not mean literal Cypher was executed.

For direct Cypher or SPARQL, the tab may show the submitted direct query and the parsed/lowered plan.

Inspect

The Inspect tab is the deeper engineering view. Look here for:

  • selected substrate: graph, lakehouse graph view, ontology, analytics, transfer, vector, or exact fallback;

  • row counts and result hashes;

  • timing and cache reports;

  • row groups, bytes read, and edge relaxations;

  • fallback reasons and unsupported-shape diagnostics.

Cypher EXPLAIN And PROFILE

Prefix a supported query with EXPLAIN to inspect plan metadata without executing it:

EXPLAIN MATCH (p:Person) RETURN count(*) AS people

PROFILE executes the child query and returns inspection fields including operator_count, row_count, elapsed_ms, plan, and result_hash:

PROFILE MATCH (p:Person) RETURN count(*) AS people

row_count counts the child query’s output rows; an aggregate returning one row is not evidence that only one graph node was scanned. Use a read-only child query when you want read-only profiling: PROFILE is an execution request, not a dry run.