How it works
This site runs no collector of its own. Everything on it comes from data.promptugal.news, the PrompTugal.news data site, which reads public sources, keeps each value with its source and the time read, and publishes it as an open API. This site turns that into pages.
Events data read from data.promptugal.news 18 Nov 2026, 16:32 UTC; Perplexity last asked 18 Nov 2026, 14:18 UTC. "Today" here is Wed 18 Nov 2026, Lisbon time, the day this page was built.
Two phases: look, then publish what we may
- Read. data.promptugal.news reads public agendas it may read, by their own supported routes (an API, a feed, the public list page), respecting robots.txt and terms. Today:
- Agenda Cultural de Lisboa (Câmara Municipal de Lisboa) (www.agendalx.pt): No terms page found. robots.txt (read 18 Nov 2026) allows all paths and sets "Content-Signal: search=yes,ai-train=no,use=reference": reference use with the source named and linked; no AI training.
- Câmara Municipal de Lagos, events list (www.cm-lagos.pt): No terms page found. robots.txt (read 18 Nov 2026) disallows only administrative folders, not /municipio/eventos.
- Teatro das Figuras, Faro (municipal theatre), agenda page (teatrodasfiguras.pt): No terms of use page found (its information pages are directions, opening hours, privacy and cookies only). robots.txt (read 18 Nov 2026) allows all but /private/. The footer says "Todos os direitos reservados". Titles, dates and links only; each event links to its page.
- Câmara Municipal de Tavira, events (Events Calendar REST route) (cm-tavira.pt): No terms of use page found (only a privacy policy). No robots.txt (404, read 18 Nov 2026). The route is advertised in the events page head (link to /site/wp-json/tribe/events/v1/). The footer says "Todos os direitos reservados". Titles, dates, venues and organisers only; each event links to its page.
- Look. Once a day per place, it asks Perplexity (the sonar model, through OpenRouter) what is on, and keeps every link Perplexity cites as evidence. Today that gives 109 claims across the places, 107 with details withheld and 0 confirmed at a source we may read.
- Publish what we may. An event found only through Perplexity shows its title, date and place, as a search engine would, credited "found through Perplexity", with every evidence link and whether we may read that source. Its details are withheld until a source we may read, or the organiser, gives them. Nothing Perplexity says is published as fact.
Every cited site we do not read is a gap: data.promptugal.news counts them on the gaps page, and each feeds a request, on the record, asking that site whether it allows a site like this to show its public events (drafted by an agent, sent by a person).
Sources we do not read for events
- dados.cm-lisboa.pt: Cloudflare bot check ("Just a moment...") on the open data API; its robots.txt also sets Content-Signal ai-train=no, use=reference. Respected, not worked around. (18 Nov 2026, 14:00 UTC)
- www.lisboa.pt: HTTP 403 on its legal and terms pages (aviso-legal, termos-e-condicoes). Respected. (18 Nov 2026, 14:00 UTC)
- visitalgarve.pt, eventos.visitalgarve.pt: eventos.visitalgarve.pt robots.txt allows all (read 18 Nov 2026), but the agenda is an empty JavaScript page (a React app, no server-rendered list, no feed or sitemap found) over a private POST API (api.visitalgarve.pt/website/Travelplanner/Search/Events) whose robots.txt was unreachable and whose terms were not found. Not used. (18 Nov 2026, 18:00 UTC)
- autodromodoalgarve.com: robots.txt allows /race-calendar/ and the list is plain HTML, but the site terms (Parkalgar, PDF) say the user may not "make available to the public, publish, ... copy ... or distribute the texts, images or other information contained on the Website ... without previous written authorisation". Respected; not read. (18 Nov 2026, 18:00 UTC)
- www.portugalmotogp.com: self-described "not the official Portuguese Grand Prix website ... independent fan guide and reseller" (TicketsGP s.r.o.); only purchase terms published. Not used. (18 Nov 2026, 18:00 UTC)
- 2026.portugaltechweek.com: robots.txt allows /events/; the festival ran 6 to 15 Nov 2026, so nothing is on in the window; no terms page found. Not used. (18 Nov 2026, 18:00 UTC)
- algarveevolution.pt: connection reset on robots.txt and on /pt/event/ (18 Nov 2026); nothing read, not worked around. (18 Nov 2026, 18:00 UTC)
- www.visitlisboa.com: robots.txt allows all; no machine-readable events feed found in the research. Not used. (18 Nov 2026, 14:00 UTC)
- www.cm-faro.pt, www.cm-loule.pt, www.cm-silves.pt: connection reset on the first request (robots.txt); nothing read. (18 Nov 2026, 14:00 UTC)
- www.cm-portimao.pt, www.cm-albufeira.pt: home pages read; no RSS or iCal feed found, only HTML and a PDF agenda. Not parsed yet. (18 Nov 2026, 14:00 UTC)
- eventbrite.*, lu.ma, luma.com, meetup.com, websummit.com: never machine-read (Dinis, 18 Nov 2026). (18 Nov 2026, 14:00 UTC)
- www.agendalx.pt: events route HTTP 403 (18 Nov 2026, 16:30 UTC)
Every source, read and not, with why and its licence limits, each linkable: our sources page. The whole register on the data site: every source data.promptugal.news reads, every one it does not, and why, quoting the terms or robots.txt with the date read. We respect what each says. Eventbrite, Luma (lu.ma, luma.com) and Meetup are never read by machine, at Dinis Cruz's word, until he decides otherwise; their listings can appear here only as evidence links.
Licences
- What PrompTugal.news makes (these pages, our words, the records and the register) is CC BY 4.0.
- What we take from someone else stays under their licence: every event, in the JSON and on the page, links its licence, or, where the source publishes none, its terms, as "the source's terms". The agendas read today publish no reuse licence; Lisbon's sets
use=referencein its robots.txt, so we name the source and link to each event. Whether that is enough to republish is for Dinis Cruz to decide (the source card). - House style: we print no em dashes, so a source's em dash shows here as an en dash; the JSON keeps the original.
Where the data lives
- The API: events by place (read), claims by place (found through Perplexity), the discovery answers (model, time, prompt, every cited link and the cost), the events source card, and how data.promptugal.news works.
- This site's copy:
data/api/in this repository, refreshed byscripts/sync-data.mjsbefore every build, withdata/api/sync.jsonrecording when each file was read. - Collected by code, no model, except the discovery (Perplexity's sonar, through OpenRouter, a small daily budget). The collectors and the discovery are data.b@PrompTugal.dinis's; this site's code and words are events.b@PrompTugal.dinis's. No person has checked the listings yet.
Threat models
For each automation this site runs, what it reads, writes and trusts, what could go wrong and how we guard against it: sync-data, scheduled-rebuild, live-data, issue-forms. Findings, fixed and open: security.
The graph
Each event will link to its place, its venue, its organiser (an organisation), its past editions (an event series: an event is never deleted when it ends, it moves to the series' history), and the companies and public, professional people connected to it. ontologist@ and data.b@ define the graph on data.promptugal.news (/api/v1/graph/, not published yet); this site will show it when it is.
Today the records give place and venue; the organiser is named only where the agenda names one; past editions and companies are not linked yet.