get_premises

Add insights, charts as data, each with a one-sentence answer, its count, scope, source and date: its asking price per m² against comparable premises nearby (same type, operation and district, half to twice its size), footfall by part of the day and where its location figures rank in the city (location access only), straight-line distances to businesses of each kind within 1 km (Overture Maps), other premises on the market within 500 m, and the monthly average asking price per m² of its type in its town over three years with the change in a year (Casafari Area Insights); prices_here, what active properties of its type ask per m² for rent and for sale in its district (then town, then province: the first with 10 priced units), quartiles from 10 and medians from 5, the town's rows beside it and medians by size band; supply_flow, from Claro's nightly count (market_daily, since 8 Oct 2026): units of its type and operation on the market in its town (or province) on the latest day, properties that came on and left over 30 days, and the median asking price per m² at both ends, with a note when fewer than 7 days are recorded; and time_on_market, the average days on the market by asking price band of properties of its type within 1 km (Casafari Area Insights; the ~500 m cell's centre when the address is hidden); and similar: up to six other active premises of the same type and operation, half to twice its size, within 1.5 km (from the ~500 m cell's centre for a hidden address), else its district, else its town, nearest first then closest in asking price per m², each with its id, one Casafari thumbnail and the agency's own listing page only (never a portal or a private seller). Asking prices, not agreed ones; each unit counted once. Also adds nearby, the map's layers: businesses of the tracked kinds within 500 m (Overture Maps; exact addresses only) and other premises on the market within 500 m as points, nearest first (a hidden address as the centre of its ~500 m cell), and the area a 5-minute walk reaches as GeoJSON MultiPolygon coordinates (location access and an exact address only). Asking prices, not agreed ones; each unit counted once. Also adds nearby, the map's layers: businesses of the tracked kinds within 500 m (Overture Maps; exact addresses only) and other premises on the market within 500 m as points, nearest first (a hidden address as the centre of its ~500 m cell), and the area a 5-minute walk reaches as GeoJSON MultiPolygon coordinates (location access and an exact address only). With location access and an exact address, location_detail: figures in groups (reach, residents, visitors, health_education, transport_parking, offices_hotels, retail_food), each with a stable key, label, value, unit, scope, source, date and a percentile in the city for figures stored for every property; the 10 nearest named places per kind with walking minutes (OpenStreetMap through Targomo); and the 10-minute walk. A figure Targomo does not have for the point is left out, never 0. With location access, scorecard: every location figure held for the premises (the stored location row, location_detail, footfall_detail and access), grouped as location_detail is, each with its scope, source and date and, for figures stored for every property, its percentile among premises of its type in the city (city) and in its district when 50 or more are measured there (district): the share with a lower figure, each unit once. Live figures carry no percentile and a note saying so; a 0 footfall or population is unknown and never ranked. strongest and weakest name the five row keys with the highest and lowest city percentiles.

Business kinds to list first in the distances chart (the brief's competitors).

With insights: add site_assessment first, a site assessment written by Gemini from this premises' own figures only (its data, its facts with their quotes, the market and competition charts, and with location access the location figures and ranks): 3 to 5 summary sentences, strengths and watch-outs, and a verdict (strong, fair, weak, not_assessable) for quick-service restaurants, coffee and bakery, clinic and pharmacy, fitness, fashion and retail, and supermarket and convenience, each reason with the keys of the figures it rests on. Every number is checked against those figures and a sentence that fails is dropped (counted in notes); a missing figure is not stated, never a weakness. scope is with_location or listing_only. Stored per premises and inputs; the first request for new inputs waits for the model.