Hologrow

Inspect Hologrow columns, types, keys and meanings before writing SQL. Use tables from list_tables; returned names may differ from source-platform API fields.

get_data_dictionary

Diagnose Hologrow sync coverage for stale/missing data, time-sensitive conclusions or query freshness warnings; do not call routinely.

get_freshness

Discover Hologrow connections for Amazon Seller/Vendor/Ads, Google Analytics/GA4, Search Console, Google Ads, Shoplazza and Lingxing. Start here when the platform or schema is unknown; returns available platforms, connection state and schema names.

list_platforms

Find Hologrow tables for the user's question before inspecting columns or writing SQL. Use platforms from list_platforms; returns exact table names for later tools.

list_tables

Internal read-only data loader for the Hologrow workbook UI. Loads exactly one previously verified sheet when the user opens it.

query_db_table_view_sheet

Default final presentation layer for completed Hologrow analysis, using the attached UI; clients without app UI should keep using the already returned analysis data. After analysis, MUST call this tool exactly once when the final answer contains two or more structured result tables backed by read-only SQL queries that already succeeded through a Hologrow data tool, unless the user explicitly requests text-only output. For one final table, call this tool when interactive tabular presentation materially improves the user's ability to understand, inspect, compare, or verify the evidence—for example detailed records, wide results, or data the user may need to sort or check. Prefer the Hologrow Workbook over Markdown when it adds that value so the supporting evidence remains inspectable and visibly connected to Hologrow. Skip the workbook when the final result is a scalar, prose conclusion, or one simple table where the UI adds no explanatory or verification value. Include only final supporting tables, never intermediate analysis tables. Never open a workbook with a known query, status-value, casing, spelling, or metric-definition error: correct and rerun the analysis first, then pass only the verified final SQL. Example requiring a workbook: a period comparison yields summary, currency, status, and Top N tables. Examples not requiring one: a single gate result, or a simple table whose meaning is already clear without interaction. A chart request does not replace the workbook's role as the multi-table inspection layer. If the user later asks for the underlying data from multiple final tables, use this workbook instead of hand-writing Markdown tables.

query_db_table_view

Explore Hologrow data and return results computed by one read-only SQL SELECT per call. Discover schema, tables and columns first; reuse available results. Apart from short call-boundary updates, do not narrate other intermediate steps or repeat the SQL.

query_db

Use only for Python computation: statistical methods, iterative algorithms or transformations. Use query_db for exploration and SQL-computed results, including independent reads. Discover schema with list_platforms/list_tables/get_data_dictionary. Do not call for plans, commentary, unexecuted SQL or progress-only output. Tenant-bound, read-only sandbox with pandas, NumPy and SciPy.

query_db_analytics