get_table_schema

readSellerSheet·

Cheaper than loading the full report-data skill when you already know which table you need and just want column names to build a query. For business-question → table mapping (e.g. "what table answers 'restock needs'?"), per-column gotchas, calibration notes, or the cross-table workflow (manual create-poll-download, disabled_reason decoding) — load the 'report-data' skill instead. This tool returns the schema only; it does NOT teach which table to use.

Call itCopy Anchor Link

{
  "name": "get_table_schema",
  "arguments": {
    "table_name": "<table_name>"
  }
}

curl -X POST https://api.sellersheetai.com/api/v1/tools/get_table_schema \
  -H "Authorization: Bearer $SELLERSHEET_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"table_name":"<table_name>"}'

ArgumentsCopy Anchor Link

ResultCopy Anchor Link

Every tool answers {notification, data, human_action}; data carries:

• data.table_name — the resolved physical table.
• data.resolved_from — the ref you passed, when it differed (else null).
• data.columns — [{name, type, nullable, default, note}] — note is a per-column trap (e.g. the days_of_supply NULL-sort trap) or null.
• data.unique_constraint — the dedup key; on a snapshot/roster table its marketplace column (country_code, country, marketplace_id, profile_id) is what report_date='latest' resolves one date per.
• data.indexes / data.fed_by_report_types / data.notes — indexes ({name, columns, unique}), the Amazon report types that feed the table, and table-level warnings (PII-blocked, schema-variant, retired/aliased; empty when healthy).