brief_get_guide
ChatGPTThe intake guide for commissioning a Tastewise report: what to ask the requester, the allowed values for category, audience, market, brief_requested and data_origin, and which fields are mandatory. Call it before brief_submit, every time — the allowed values are live, not a fixed list you can remember. Not an analysis tool: it returns no market data. Takes no arguments. Brief type is deliberately absent — a Tastewise analyst assigns it after intake.
brief_submit
ChatGPTValidates a brief a requester has dictated, and files it once they confirm. Use when someone asks Tastewise for a custom report, analysis or deep dive. Call brief_get_guide first — it carries the allowed values and the required policy. With confirm false nothing is persisted, so call it that way while gathering: the answer names the fields still missing, and you ask the requester for each one. Never invent a value to fill a gap, and never submit on a guess.
foodservice_get_dashboard
ChatGPTThe whole Operator Insights landscape for a term in one market: source volume, the 24-month trend, four operator breakdowns, menu items and menu ingredients, as eight panels with their figures. Use for the restaurant-side picture of a topic, not one figure. Not for a single total (foodservice_get_overview), nor for one angle already known, which foodservice_rank_by_locations or foodservice_trend_item_count answers faster. Fixed panels, no query parsing: one question always draws one dashboard. digest is your only copy of the numbers.
foodservice_get_overview
ChatGPTTop-line totals for the menus matching a search in one market: menu items, restaurants and chains, plus school and c-store totals in the US. Use when the user asks how many, how big, or for a single total — the short path, one call. Not for the whole operator landscape at once (foodservice_get_dashboard), for trends (foodservice_trend_item_count), per-operator breakdowns (foodservice_rank_by_locations) or the items themselves (foodservice_list_menu_items). Draws the dashboard's source-volume strip: counts only, no ranking, no share, no growth.
foodservice_get_trend_performance
ChatGPTHow ONE term has moved over 24 months in one market, or inside one slice of it: menu presence, social presence, item count, items per operator and average price — five monthly series, each with its year-over-year growth and a written summary. ONE slice per call. Not for comparing two or more segments, chains, states or cuisines against each other: when the question asks which of them moved more, grew faster or priced higher, use foodservice_trend_item_count for counts, foodservice_trend_avg_price for price and foodservice_trend_location_share for share, each returning one series per slice. Use it when a single term is asked about with no breakdown: the market-wide price of a dish and how it moved, how a term grew inside one named segment, or whether it is still rising. Nor for the social side on its own (social_get_trend_performance). Draws the dashboard's Trend performance card, one tab per series. Each tab states its own unit — they are not all percentages — and a business_type slice leaves the menus and social tabs market-wide.
foodservice_list_ingredients
ChatGPTIngredient analytics for the menus matching a search in one market: per-ingredient menu, recipe and social counts, monthly and yearly menu change, channel split and lifecycle stage. Use for on-menu ingredient questions — the ingredient mix of matcha menus, what sits on taco menus by chain. Not for the consumer/social side (social_list_ingredients_ranked), nor for whole menu items (foodservice_list_menu_items). Draws the dashboard's Menu ingredients table — menu share, yearly change and index per ingredient, highest share first.
foodservice_list_regions
ChatGPTWithin-region share of menus mentioning the search term, one row per region, as region code and percent. Use for map or where-is-it-served questions at state or region level. Not for cities — there is no city granularity here; use foodservice_rank_by_locations with dimension=cities. Not for over-indexing either: this is raw within-region share, not share-with versus share-without. Regions with no data are dropped. Entries the market could not place on a region are NOT dropped silently: they are counted in unplaced, whose share_percent belongs to no region and must be stated rather than ignored. That figure is small everywhere except FR and ES, where most of the answer lands there and the regional picture is not usable. USA, UK, CA, AU, DE, FR, IN, BR, MX, ES only.
foodservice_rank_by_locations
ChatGPTChains, states, business types, cuisines or cities ranked by how many LOCATIONS carry the search term. Each row also carries that slice's share of matching locations, averaged over the period, and the year-on-year change in THAT SHARE — not in the count. The widget labels them Menu Share and Menu Growth, as the dashboard does. Use for top chains for a term, or the top states for a dish — questions about reach and how many outlets serve it, never about how many items each carries. Not for menu depth per operator (foodservice_rank_by_menu_depth), nor for pricing (foodservice_trend_avg_price). Point-in-time ranking only. dimension decides what a row is.
foodservice_trend_avg_price
ChatGPTMonthly AVERAGE PRICE of matching menu items per slice in one market, in that market's currency. Each series also carries its average share of matching locations over the period and the year-on-year change in PRICE — this is the one foodservice tool whose growth is a price move rather than a share move. The widget labels the two Menu Share and Menu Growth, as the dashboard does. Use to compare price levels across segments — which chains price a dish highest, how QSR and fast casual differ. Not for share of voice (foodservice_trend_location_share), counts of locations or items (foodservice_trend_item_count), nor for how one term's price moved with no breakdown asked for — that is foodservice_get_trend_performance, which returns a single price series instead of one per slice. Sorted most expensive first; the most recent month's price is each line's last point, so it is not repeated as a field of its own.
foodservice_trend_item_count
ChatGPTMonthly menu-item COUNT per slice for a search in one market. Counts are averaged across delivery platforms. Each series also carries, beside the line the chart draws, its average share of matching locations over the period and the year-on-year change in THAT SHARE — not in the count. Read those two when asked which slice is gaining or losing ground, rather than differencing the line. Use to compare two or more slices over time — chains against chains, states against states. Not for share trajectory (foodservice_trend_location_share) nor price (foodservice_trend_avg_price) — those are separate measures, not options here — and not for how one term evolved with no breakdown asked for, which is foodservice_get_trend_performance. dimension decides what a series is.
foodservice_trend_location_share
ChatGPTMonthly LOCATION-share per slice for a search in one market — the percent of matching locations that fall in each slice, which sums to about 100 across the slices of a month. Each series also carries the year-on-year change in that share. The widget labels them Share and Share Growth, as the dashboard does. Use for how operator composition shifted over time — gained or lost share, share trajectory, which states are gaining. Not for item counts (foodservice_trend_item_count), price (foodservice_trend_avg_price), nor for one term's own trajectory with no slices to compare (foodservice_get_trend_performance). Slices with under 24 months of history are dropped, so fewer than five series is normal and the slices drawn here need not match those of any other card.
retail_get_dashboard
ChatGPTA rendered multi-panel shelf dashboard for a broad e-commerce question, drawn in one shot with the figures behind it. Use when the user wants a product's whole shelf landscape — the panels show it, your answer reads it. Not for one number or one angle: retail_get_overview answers how big the shelf is, and retail_rank_brands or retail_list_products answers one ranking. Returns the panel layout and digest, the same figures as text — your only copy.
retail_get_overview
ChatGPTTop-line shelf totals for a search in one market: how many sales categories, online products, brands and retailers carry it, recent launches, and the average price. Use when the user asks how big the shelf is, how many products or brands sit on it, or the average price. Not for the products themselves (retail_list_products), for a ranking (retail_rank_brands, retail_rank_retailers, retail_rank_categories), nor for the whole shelf drawn at once — that is retail_get_dashboard. Six numbers, no rows, no ranking, no growth.
retail_list_products
ChatGPTThe actual products on shelf matching a search in one market — names, brands, prices and review data. Use when the user wants to see the products themselves rather than a count or a ranking: what is being sold, at what price, under which brand. Not for how many there are (retail_get_overview), nor for a ranking of brands, retailers or categories (retail_rank_brands, retail_rank_retailers, retail_rank_categories). One page of products, the same rows the shelf widget shows.
retail_rank_brands
ChatGPTThe brands selling a searched product in one market, ranked by the metric you pick, with a written narrative. Use for which brands lead a retail category, who is launching the most new SKUs, or which brands price highest. Not for the shops that carry them (retail_rank_retailers), nor for the individual products (retail_list_products). This is e-commerce shelf data, not social conversation — for who is talking about a brand, use the social_list_* tools. Ranked rows only, no time series.
retail_rank_categories
ChatGPTThe retail sales categories a searched product sits in, ranked by the metric you pick, with a written narrative. Use for which aisles or shelf categories a product belongs to, and which of them carry the most of it. Not for the categories behind a social trend, with their share of conversation — that is social_list_categories, a different dataset. Ranked rows only, no time series.
retail_rank_claims
ChatGPTThe claims printed on the PACKAGING of products sold online in one market — organic, gluten free, high protein and the like — ranked by the metric you pick, with a written narrative. Use for what packs claim, which claims appear most on shelf, and which are appearing on new launches. Not for which claims come up when people talk about a food — that is social_list_consumer_motivations, and a question about the claims around a product, rather than on its packs, belongs there. Nor for the consumer-need categories those sit in (claims_list_categories). Ranked rows only, no time series.
retail_rank_ingredients
ChatGPTThe ingredients and flavours in the packaged products matching a search in one market, ranked by the metric you pick, with a written narrative. Use for what is in the products on shelf, which flavours sell most, and which are arriving on new launches. Not for the consumer/social side either (social_list_ingredients_ranked), nor for restaurant menus (foodservice_list_ingredients). Ranked rows only, no time series.
retail_rank_retailers
ChatGPTThe retailers and e-commerce platforms carrying a searched product in one market, ranked by the metric you pick, with a written narrative. Use for where a product sells, which chains stock the most of it, or which price it highest. This is the shops, not their shoppers — for who buys, use social_list_audiences. Not for the brands on the shelf (retail_rank_brands), nor for the whole shelf drawn at once (retail_get_dashboard). Ranked rows only, no time series.
social_get_dashboard
ChatGPTA rendered multi-panel dashboard for a broad social-trend question, drawn in one shot, together with the figures behind it so you can analyse them. Use when the user wants a topic's landscape — the panels show it, and your answer reads it. Not for one specific number or one angle: call the matching social_list_ / social_get_ tool instead of drawing eight panels to answer one question. Returns the panel layout for the view, and digest — the same figures as text, which is the only copy you get.
social_get_overview
ChatGPTHow much consumer data sits behind a social trend in one market: people, social posts, recipes, dishes and restaurants mentioning it. Use when the user asks how much data a topic rests on, or how big its conversation is. Not for growth (social_get_trend_performance), for the whole landscape at once (social_get_dashboard), for operator menu totals (foodservice_get_overview) or the shelf (retail_get_overview). Five counts, no ranking, no growth; a count the source lacks shows as a dash.
social_get_trend_performance
ChatGPTHow a social trend performs over time in one market, across five tabs — social mentions, menu penetration, recipe adoption, forecast, audience — each a chronological series, with a written narrative. Use when the user asks whether a topic is growing or declining, or wants a forecast. Not for ranked breakdowns, which every other tool here returns. Always five rows, one per tab; the audience tab carries at most 3 named lines.
social_list_audiences
ChatGPTThe consumer audiences behind a social trend in one market — generations, behavioural cohorts, a retailer's shoppers, a chain's customers — with a written narrative. Use when the user asks who is driving a topic, or which cohort over-indexes on it. Not for why they choose it (social_list_consumer_motivations), nor for an audience curve over time (social_get_trend_performance). Not for the shops themselves either — this is WHO shops, not WHICH retailers stock a product, which is retail_rank_retailers. Returns up to 10 by share. Can come back empty for a sparse topic.
social_list_categories
ChatGPTRanked food categories behind a social trend in one market, with a written narrative summarising them. Use when the user asks which categories drive a topic, or which are growing or declining year over year. Not for named dishes (social_list_dishes_ranked), ingredients (social_list_ingredients_ranked), audiences (social_list_audiences) or claims (social_list_consumer_motivations). This is conversation share, not the retail shelf — for the sales categories a product sits in, use retail_rank_categories. Returns up to 10 by social share, highest first; shares do not sum to 100. Can come back empty for a sparse topic.
social_list_consumer_motivations
ChatGPTThe claims and dietary motivations behind a social trend in one market — vegan, high-protein, gluten-free, keto and the like — with a written narrative. Use when the user asks why people reach for a topic, which diets and claims are growing around it, or which claims come up in conversation about it — named claims like vegan, keto or high protein included. Not for who those people are (social_list_audiences), for movement over time (social_get_trend_performance), nor for the need categories those claims fall under (claims_list_categories). This measures what people say; for what is printed on the packaging, use retail_rank_claims. Returns up to 10 claims ranked by volume, or 7 when the query implies a line chart.
social_list_curated_posts
ChatGPTUp to 6 Instagram posts from real consumers about a food or drink term in one market, best first — the same posts the Tastewise platform shows under On the plate. Use when the user wants example posts, real consumer voices, or visual proof of a trend. Each post comes as a tile: the account's handle, the link to the post on Instagram, and a caption saying why it was picked. How posts are picked: the top 100 Instagram posts for the term, narrowed to the 6 that are mainly about the term, are in the market's language, and come from regular people, not brands or celebrities. curated says whether that picking worked: when false, the picking step failed and these are the plain top posts, which may include brand or off-topic posts. Not for how much is said about a term (social_get_overview), for whether it is growing (social_get_trend_performance), nor for who posts about it (social_list_audiences). Only the top 6 posts are ever given: if the user asks for more, say that only the top 6 are given. Fewer, or none, come back for a rare term or a small market.
social_list_dishes_by_lifecycle
ChatGPTNamed dishes grouped by maturity stage — early, emerging, trending, mature, declining — for a social trend in one market. Use when the user asks what is emerging versus fading, or what the next dish for a topic looks like. Not for one flat ranked list (social_list_dishes_ranked) or for ingredients (social_list_ingredients_by_lifecycle). Up to 7 dishes per stage, ranked by social share. Carries no narrative — the title is fixed, matching the platform widget.
social_list_dishes_ranked
ChatGPTThe named dishes carrying a social trend in one market, with a written narrative. Use when the user asks which dishes are trending for a topic, or wants the leading dishes in a category. Not for how dishes spread across maturity stages (social_list_dishes_by_lifecycle), for ingredients (social_list_ingredients_ranked) or for categories (social_list_categories). Returns up to 10 rows ranked by the metric the query implies. The dish name sits in the ingredient field — an upstream schema alias, not a mistake.
social_list_ingredients_by_lifecycle
ChatGPTNamed ingredients grouped by maturity stage — early, emerging, trending, mature, declining — for a social trend in one market. Use when the user asks which ingredients are emerging versus fading, or what the next ingredient for a topic looks like. Not for one flat ranked list (social_list_ingredients_ranked) or for dishes (social_list_dishes_by_lifecycle). Up to 7 ingredients per stage, ranked by social share. Carries no narrative — the title is fixed, matching the platform widget.
social_list_ingredients_ranked
ChatGPTThe named ingredients carrying a social trend in one market on the consumer side, with a written narrative. Use when the user asks which ingredients are trending for a topic, or wants the leading ingredients in a category. Not for maturity stages (social_list_ingredients_by_lifecycle), for whole dishes (social_list_dishes_ranked), for restaurant-side menu analytics, which this surface does not expose at all, nor for what is in packaged products on shelf (retail_rank_ingredients). Returns up to 10 rows ranked by the metric the query implies.
execute_endpoint
Claudefetch
Claudefind_endpoints
Claudeget_endpoint_detail
Claudeget_endpoints
Claudesearch
Claude
