browse_categories
ChatGPTUse this when the user wants to understand how products are grouped, find a category UUID (e.g. to pass to browse_products), or explore the category tree. Results are paged; pass cursor to continue. Use root=true to start from the top of the tree, then drill in with parent_category. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. - cursor (optional) — next or previous URL from a prior response; all other filters are ignored when set. - parent_category (optional) — restrict to children of this category UUID. - root (optional) — true for top-level only; false for non-root; omit for all. - include (optional) — comma-separated related entities: name, parent_category, images, text.
browse_products
ChatGPTUse this when the user wants to see what products exist on their account — to find a product ID, check what tickets are on sale, or explore the catalog before doing something with a specific product. Results are paged; pass cursor from a previous response to continue. Narrow by category or enabled status to reduce noise. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. - cursor (optional) — next or previous URL from a prior response; all other filters are ignored when set. - category_id (optional) — restrict to products in this category. - enabled (optional, default true) — false for disabled-only; omit for both. - has_external_id (optional) — filter by whether the product is linked to an external system. - current_only (optional, default false) — keep only products live today (enabled and inside their validity window). Trims mostly-stale catalogs; filters the current page only, so a page may return zero rows yet still have a next cursor. - include (optional) — comma-separated related entities: name, account, text, categories, resources, images, external_ids, barcode, taxes, scan_options.
get_analytics_schema
ChatGPTUse this to inspect the analytics schema before writing SQL. Call once per session, after reading get_analytics_skill; reuse the skill if already loaded for the current connection, account, and permissions. The SQL DDL lists every fact and aggregate table (orders, items, memberships, payments, sales-vs-budget, weather joins) with their columns, measures, and default filters. Retain the schema in context and reuse it for subsequent queries without calling this tool again. Refresh only when the connection, account, or permissions change, a schema update is reported or requested, or the cached schema is no longer available in context. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Returns {"account_slug": "...", "view": "...", "schema_ddl": "..."}.
get_analytics_skill
ChatGPTCall once per session, before the first get_analytics_schema or query_analytics call. It returns the canonical data-analyst skill: the Convious data team's playbook on which fact table owns which metric, the demo-account filter, join keys, default filters, how to read the schema, and worked SQL patterns. Read it fully, retain it in context, and apply it to subsequent data questions without calling this tool again. Refresh only when the connection, account, or permissions change, an update is reported or requested, or the cached guidance is no longer available in context. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Returns {"account_slug": "...", "skill_name": "...", "skill_markdown": "..."}.
get_search_schema
ChatGPTCall this FIRST, before `search`, whenever you intend to use structured `filters`. Returns the live function-calling schema for search on this account — the exact per-type filter fields and enum values (order/membership statuses, group-booking date modes, the account's group-booking location UUIDs, etc.), scoped to the result types you can access. The schema is generated from the same serializers search validates against, so what it advertises is exactly what search accepts. The per-account filters (notably group-booking locations and the available status enums) are dynamic, which is why you read them here rather than guessing. Call once per session per account and cache the result; you only need it when you want to pass filters (a plain free-text q search needs no schema). Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Returns {"name": "...", "description": "...", "input_schema": {...}, "types": {...}}. input_schema is the JSON Schema for search's arguments; types maps each result type to the permission it requires.
health
ChatGPTLiveness probe for the gateway. No parameters. Any authenticated caller can use it.
look_up_product
ChatGPTUse this when the user wants to inspect a specific product — its name, pricing type, category membership, barcode settings, validity window, or any other configuration detail. Takes the product UUID, or the numeric product ID that analytics results use; use browse_products first if you only have a name. Parameters: - product_id (required) — product UUID, or a numeric product ID (e.g. 27686) as returned by query_analytics. Numeric IDs are resolved by matching the analytics product name against the account's catalog, so they only work for products with sales history — prefer the UUID when you have it. An ambiguous or unresolvable numeric ID returns an error listing candidate products to pick from, and an id that is neither a UUID nor digits is rejected outright — use browse_products or search to find a product by name. - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. - include (optional) — comma-separated related entities to expand: name, account, text, categories, resources, images, external_ids, barcode, taxes, scan_options.
preview_pricing
ChatGPTUse this when the user wants to see what prices the RTP/NYOP engine would produce under a proposed configuration, without committing the change. Useful for "what would happen if I raised the floor to €18?" or showing projected daily prices before calling update_pricing. Window: future dates only, max 60 days. Parameters: - article_uuid (required) — product UUID. - date_from (required) — ISO8601 date (must not be in the past). - date_to (required) — ISO8601 date; window ≤ 60 days. - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - products_list_uuid (optional but effectively REQUIRED) — category (products-list) context UUID. Without it the engine returns the no-op identity model with an empty prices list, so always pass the UUID of a category the article belongs to (see browse_categories). - settings_payload (optional) — proposed settings to evaluate instead of the persisted ones, shaped {"productData": {"priceSettings": {originalPrice, avgTargetPrice, minAcceptedPrice, gatePrice, stepSize, ...}, "override": {...}}, "accountSettings": {...}}. Omit to preview the current live settings. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Returns {"modelId": ..., "defaultPrice": ..., "pricingStrategy": ..., "prices": [...], "validationMessages": [...]}.
query_analytics
ChatGPTUse this when the user asks a data question about their venue — revenue, GMV, refunds, top products, memberships, payment failures, budget vs. actuals, or anything else in the analytics dataset. Load get_analytics_skill (the playbook), then get_analytics_schema (the column list), once per session. If both are already available in context for the current connection, account, and permissions, query directly without calling them again; apply the cached skill and schema to every query. Refresh the affected guidance only when the connection, account, or permissions change, an update is reported or requested, or cached context is lost. Follow the skill's demo filter, fact-table-per-metric, and default payment-flow filters. SQL rules: SELECT or WITH only (no DDL/DML); must filter by account_slug somewhere. Parameters: - sql — the SQL to run. Required for a new query; omit it when paging with cursor. - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. - cursor (optional) — pass the next_cursor from a previous response to fetch the next page of the same result set. When set, sql is ignored. Returns {"account_slug": "...", "sql": "...", "columns": [...], "rows": [...], "row_count": N, "total_row_count": M, "next_cursor": "..."|null}. Up to 1000 rows per call; if next_cursor is non-null, call again with that cursor to get the next page. total_row_count is the whole result set; row_count is this page. Cursors stay valid ~24h.
read_price_settings
ChatGPTUse this when the user wants to see how a product is currently priced, or before calling update_pricing — the response includes the version token required for a safe optimistic-lock write. Returns the original price, RTP/NYOP target and floor, gate price, step size, and the engine's computed defaults. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - product_ids (optional) — list of product UUIDs, or numeric product IDs (e.g. 27686) as returned by query_analytics. Numeric IDs are resolved to UUIDs first via analytics name-matching, so they only work for products with sales history; any that can't be resolved come back under unresolved_product_ids instead of being dropped. Omit to read every currently-live product on the account (bulk mode). - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Returns {"priceSettings": [{originalPrice, avgTargetPrice, minAcceptedPrice, gatePrice, stepSize, pricingType, version, lastModifiedAt, lastModifiedBy, ...}]}, plus unresolved_product_ids when a supplied numeric ID could not be resolved. Bulk mode pages through the whole account catalog and adds resolved_product_count (and partial: true only if an internal page cap is hit).
search
ChatGPTUse this to find a specific entity across a Convious account in one shot — an order, membership, article (product), product list (category), or group booking (pay-later cart). Free-text q matches identifiers, names and emails (order number, reference id, customer/article name, email); matching is lexical (exact + fuzzy), so near-misses still hit. This is the fastest way to answer "find order #…", "which membership belongs to this email", "is there an article called …". For structured filtering (by status, date range, group-booking location, etc.), call get_search_schema FIRST to learn the exact per-type filter fields and enums, then pass them in filters. Parameters: - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - q (optional) — free text: order number, reference id, email, customer or article name. Use customer data only for an authorized venue-support workflow. Terms under 2 characters are ignored. - types (optional) — list restricting which result types to return (order, membership, article, products_list, group_booking); omit to search every type you can access. Types you lack permission for are silently skipped. - filters (optional) — per-type structured filters keyed by type, e.g. {"order": {"status": ["CONFIRMED"]}}. Get the exact shape from get_search_schema. A type whose search fails reports why in its own group's error (the rest of the search still runs). - limit (optional, default 10) — per-type result cap, 1–50. - account (optional) — inferred from your session; pass explicitly when you have access to more than one. Provide at least q or filters (or types) — an empty search returns nothing useful. Returns {"groups": {"<type>": {"total": N, "results": [...], "partial": false, "error": null}}} — one group per searched type, each hit carrying id, title, subtitle, deeplink, score and type-specific fields.
search_copies
ChatGPTUse this to go from text seen on a venue's widget, kiosk, POS or email to its CMS key, or to find copies whose key, name or category contains a word. Typical flow: the user shows or quotes a text, you call this with a short distinctive fragment of it, then export_copies with the returned key shows every translation, and preview_copies / create_copies with on_conflict: "update" change it (those three need Control Panel admin; this search does not). Results are the served copies that apply to the venue: global copysets plus the venue's own; isDefaultCopyset false means the venue overrides the global text. Only translations in the locales the venue serves are searched: a text in another language never matches, so search a venue that serves that language instead. Backoffice checks the read_copies permission on the venue and answers 403 without it. Search tips: stored text may hold placeholders (a venue name, a price), and rich text holds HTML, so a fragment of five to eight plain words matches better than a whole sentence. A highlighted block on a page can be two copies joined; search each part. The match is case-insensitive and partial. Parameters: - search (required) — the fragment to look for. - search_by (optional, default text) — text matches translations; key matches category_slug.copy_slug (a dot searches category exactly and copy slug as prefix) or, without a dot, part of the key, copy name or category. - intent (required) — use the exact fixed purpose from the input schema's enum. Never include personal data, identifiers, credentials, or user quotes. The server records a fixed purpose. - account (required for Convious employees) — venue slug. Partners with a single account have it inferred. Returns {results: [{key, text, locale, name, editorType, isDefaultCopyset}]}, one row per key and locale, venue-specific copies first.
