How Najimi uses Qloo
Qloo's taste graph is the engine under every screen. Every endpoint Najimi calls, what it asks, and how the agent gets the whole surface.
Built on , the taste graph that knows how the things people love connect across culture.
Najimi would not work without Qloo. Someone who loves Insecure and Americanah tends to love Jorja Smith, a certain kind of restaurant, a certain kind of room: Qloo knows those links across music, film, TV, books, podcasts, brands, places, destinations, games and people. Every pick, word, map, match and plan stop in Najimi is a Qloo answer, and so are the hidden answers in Guess my wavelength and Blind Taste.
- Qloo endpoints, used for everything
- 6
- kinds of culture in every taste
- 10
- Qloo tools the agent composes freely
- 5
- of calls logged, with who, why and how long
- 100%
The endpoints
| Endpoint | What Najimi asks it |
|---|---|
GET /search | Turn a name into a Qloo id: the loves you pick, cities, a business near its address, imports, the agent's lookups. |
GET /entities | One thing's full record (the sheet for anything you tap); imports matched exactly by Goodreads, MusicBrainz, X or Instagram id; poll options' artwork. |
POST /v2/insights | The workhorse. Your loves go in as weighted { id, weight } signals with explainability on. See the uses below. |
GET /v2/tags | Find tags by meaning: "candlelit", "jazz bar", "rainy day", a business's kind. |
GET /v2/analysis/compare | What two tastes share and what's only each one's: your match with anyone, a business against its street, the agent's comparisons. |
GET /v2/audiences | Audiences the agent can lean a taste with. |
The many shapes of /v2/insights
| Ask | Used for |
|---|---|
filter.type = an entity kind, weighted signal | Your picks in ten kinds, every page of the feed, more like this, the graph's growth, games, tune-up accounts |
filter.type = urn:tag | Your taste's words: aesthetics, decor, ambience, style, cuisines, dishes |
filter.type = urn:heatmap | Where a taste lives in a city, square by square: Earth, trip areas, a business's people |
filter.type = urn:demographics | Who a taste leans toward by age and gender: business plans, the agent |
signal.location, no person | A spot's own scene within 2.5 km: Earth's scene sheet, a business's street |
filter.results.entities | How well a shortlist fits someone: Guess my wavelength and Blind Taste answers, each friend's fit in a group, the feed score |
audience_growth | Places rising among a business's fans |
The agent gets the whole surface
Najimi isn't handed a "recommend" button. Its five Qloo tools mirror the API itself, and it composes them for each ask:
| Tool | Endpoint | |
|---|---|---|
qlooSearch | /search | names → ids |
qlooTags | /v2/tags | meanings → tags |
qlooInsights | /v2/insights | picks, words, heatmaps, demographics, with Qloo's signals and a wide set of its filters |
qlooCompare | /v2/analysis/compare | two tastes side by side |
qlooAudiences | /v2/audiences | audiences to lean with |
See Every tool for every filter and how it composes them ("hidden-gem music" is artists with a low popularity ceiling; "is Burna Boy for me?" scores exactly his id against your taste).
Fast, cached and fair
Next: feature by feature, and explainability.