About
About gigcast
What a Booking Brief is, how cultural data grounds it, and where it stops being reliable.
Overview
What it is
Concert promoters often decide on gut feel: which city, which support act, which sponsor, which price. The data that would help is spread across ticket history, social numbers and local knowledge, and none of it says who the audience is or where they live.
gigcast turns an artist, a city and dates into a Booking Brief with eight parts: audience map, demand and fit, lineup, sponsors, venues and partners, pricing, risk panel, agent trace. Each claim is labeled grounded, inferred or no coverage. It is for promoters, talent buyers and venue bookers who need a first read on a market before they commit money, and want to see what each claim rests on.
- Audience map
- Demand and fit
- Lineup
- Sponsors
- Venues and partners
- Pricing
- Risk panel
- Agent trace
Data
Where the data comes from
Behind every brief is one affinity graph across artists, brands, places, localities and demographics: Qloo's Taste AI, reached through its API. One graph means a brief can answer "who likes this artist, where do they live, what do they buy, where do they go" with results that connect to each other. These are the endpoints the agent uses.
- /search
- Resolves artists, brands and localities to entities in the graph. A city picked on /new is pinned by its entity id, so "Panama City" does not drift to Panama City, Florida.
- /v2/insights, filter.type=urn:heatmap
- Geohash cells with affinity, popularity, quality and price-level dimensions. This is the audience map.
- urn:entity:artist with signal.location
- Lineup candidates that over-index in the city.
- /v2/insights/compare
- The shared tags behind a pairing of headliner and support act.
- urn:entity:brand
- Sponsor candidates, also filtered by the dominant age bucket.
- urn:entity:place with filter.location and place tags
- Venues, bars, hotels and restaurants near the heat.
- urn:demographics
- Age and gender affinities of the audience.
- urn:tag
- Tags for the artist and for the city, used for the taste match.
- /v2/trending
- The twelve-month trend for the artist.
Evidence
The three labels
- Grounded
Backed by a result in the taste graph; the entity id was returned by the cited tool call.
- Inferred
The model reasoned on top of grounded data; not itself a data result.
- No coverage
The taste graph returned nothing usable for this item or section.
The server checks grounded picks. An entity id must appear in a tool result of the same run, and a risk item must cite a url that the risk feed returned. A pick that fails the check is not shown as grounded.
System
Architecture
Provenance
Data sources
| Source | What we take | Terms note |
|---|---|---|
| Qloo | Taste, heatmap, demographics, trending and entity data. | Hackathon API access. |
| Google News RSS | Headlines and links. | Never article bodies. |
| Local outlet RSS feeds | Headlines and links from local outlets per market. | Never article bodies. |
| Nager.Date | Public holidays. | Public API. |
| Open-Meteo ERA5 archive | Historical weather and climate. | CC BY 4.0. |
| US State Department, UK FCDO | Travel advisories. | Public. |
| OpenFreeMap tiles | Base map for the heatmap. | OpenMapTiles and OpenStreetMap attribution shown on the map. |
We store headlines and links, never article bodies. Nothing is sent to sponsors.
Limits
Honesty notes
- Affinity is an audience-overlap signal. It is not a ticket forecast.
- Demographics are affinities relative to the general population, not headcounts.
- Pricing is a heuristic, not a quote or a revenue forecast.
- Risk classification is a model reading headlines. It is not a safety assessment.
- Briefs are public read-only links. Anyone with the link can read the brief.
Elsewhere