Validation & Auto-Fix

Analytics Agent checks reports at three levels: document → blocks → queries. The idea is simple: a reader never sees a broken report — everything is caught before publishing, and every error comes with an AI fix button next to it.

Document level

  • MISSING_TITLE — the document has no title;
  • NO_BLOCKS — a document without blocks;
  • MISSING_BLOCK_ID — a block has no ID;
  • DUPLICATE_BLOCK_ID — a duplicate block ID;
  • INVALID_BLOCK_WIDTH — a block width outside the 1–24 range;
  • EMPTY_BLOCK_CONFIG — an empty block configuration;
  • INVALID_ACCESS_LEVEL — an unknown access level;
  • DUPLICATE_QUERY_ID, DUPLICATE_OVERRIDE_ID — duplicate query and override IDs;
  • INVALID_VIEWS, DUPLICATE_VIEW_NAME, UNKNOWN_VIEW_BLOCK, MISSING_FILTER_VARIABLE, INVALID_PARAMETER_TYPE — errors in report settings: tabs, filters and what-if fields.

Content level

  • the block type is known to the registry — a typo in contentType is caught immediately;
  • queryId references exist, the query’s source is found (DATA_SOURCE_NOT_FOUND);
  • the chart configuration is filled in: columns are set (EMPTY_CONTENT_DATA, INVALID_CONTENT_DATA_TYPE).

Query level

SQL queries are validated before execution: syntax, columns, sources. Details in the Queries section.

Background validation

  • all documents are checked automatically, in the background;
  • results are stored — repeated checks don’t load the server;
  • when document files change — an immediate re-check.

API and display

  • POST /api/validation/document — document validation;
  • POST /api/validation/content — block content validation;
  • GET /api/validation/data-sources — data source checks;
  • errors are shown in the interface next to the block — with the “Fix with AI” button (details in the agent in the report).

← Back to the documentation index