Where this comes from

Where our data comes from

We actively ingest from 19 sources today, and carry 12 live video feeds. We also keep a catalogue of 491 local sources we have identified and assessed — most of those are not being collected, and are listed there so we can work through them honestly rather than claim them.

Live video we carry

Every feed below is either a public body's own published embed or a public channel whose owner has embedding enabled. We checked each one individually and recorded why we believe we may show it.

• Wellington TV

  Wellington TV, courtesy of the Village of Wellington. Granicus is the Village's own streaming vendor and publishes this player route as an embed ("Copy this html to embed into your page"). Verified 2026-08-26: HTTP 200, no X-Frame-Options, CSP is Report-Only with no frame-ancestors, page declares isLive = true. Public meetings of a public body.

• Royal Palm Beach Village Council

  Village of Royal Palm Beach. YouTube's own embed player, which exists to be embedded; channel is public with embedding enabled (oEmbed 200) — verified 2026-08-26. Public meetings of a public body.

• Palm Beach County TV — Channel 20

  PBCTV Channel 20, Palm Beach County. YouTube embed player; channel public with embedding enabled — verified 2026-08-26. The county's own discover.pbcgov.org player page is NOT usable (X-FRAME-OPTIONS: SAMEORIGIN). Public meetings of a public body.

• Palm Beach County School Board

  School District of Palm Beach County. YouTube embed player; channel public with embedding enabled — verified 2026-08-26. The district's published /c/sdpbc/live URL is SAMEORIGIN and cannot be framed; the channel row resolves the live broadcast instead. Public meetings of a public body.

• WPB-TV — City of West Palm Beach

  WPB-TV, City of West Palm Beach. YouTube embed player; channel public with embedding enabled, resolved to a live broadcast running continuously since 2024-02-02 — verified 2026-08-26. Municipal government channel.

• Wellington International — show ring

  Wellington International. YouTube embed player; channel public with embedding enabled — verified 2026-08-26. Free public classes only; the paywalled ShowGroundsLive / ClipMyHorse feeds are deliberately NOT used (CMH §6.2 restricts to "private and non-commercial purposes").

• Delray Beach — Atlantic Avenue

  Live beach cam courtesy of the City of Delray Beach (hosted by Brownrice Internet). Purpose-built embed endpoint (/embed/, loader invoked with framed=true), HTTP 200, Access-Control-Allow-Origin: *, no X-Frame-Options, no CSP, live HLS ladder confirmed — verified 2026-08-26. No published written embed terms found; shipped on intent-of-design, courtesy confirmation requested from the City.

• Jupiter Reef Club — oceanfront

  Live cam courtesy of Jupiter Reef Club. YouTube embed player; public with embedding enabled (oEmbed 200) and broadcasting (isLiveNow: true) — verified 2026-08-26.

• Sawfish Bay Park — Loxahatchee River

  Live cam courtesy of the Town of Jupiter. YouTube embed player; the Town of Jupiter publishes this cam as an iframe on its own .gov site. Public with embedding enabled (oEmbed 200) and broadcasting (isLiveNow: true) — verified 2026-08-26. Municipal government work.

• Jupiter Veterans Memorial

  Live cam courtesy of the Town of Jupiter. YouTube embed player, published by the Town of Jupiter on its own .gov site. oEmbed 200, isLiveNow: true — verified 2026-08-26. Municipal government work.

• Cinquez Park Dog Park

  Live cam courtesy of the Town of Jupiter. YouTube embed player, published by the Town of Jupiter. oEmbed 200, isLiveNow: true — verified 2026-08-26. Municipal government work.

• Catopia — Furry Friends cat room

  Live cam courtesy of Furry Friends Humane Society, Jupiter. YouTube embed player; public with embedding enabled (oEmbed 200) and broadcasting continuously since 2023-05-17 (isLiveNow: true) — verified 2026-08-26.

What we deliberately don't use

This is the part of a transparency page that is actually worth reading. Where permission hasn't been granted, we say pending — not granted.

• WeatherBugNot used

  Their terms explicitly forbid framing and mirroring, restrict use to personal use, and bar bulk imagery. That is a clear no, and we treat it as settled: it is not revisited without written permission from them. Our weather comes from the National Weather Service instead.

• Palm Beach County traffic camera streams (pbctrafficstreams.pbc.gov)Permission pending

  These are the best live video of Wellington roads that exists, and we do not publish them. Resharing requires explicit written permission, which we do not have. The endpoint happens to be technically open — permissive CORS, no token check — but a misconfiguration is not a licence, so we treat it as closed. No PBC camera stream appears anywhere on this site today. The traffic cameras we do show come from Florida DOT's public FL511 system.

• Wellington International competition videoNot used

  Server-side authenticated; there is no anonymous URL and we do not try to find one. They publish no terms of use at all, and the absence of terms is the absence of a grant, not the presence of one. We show their free public YouTube channel and their published schedule only.

• ShowGroundsLive / ClipMyHorse paid feedsNot used

  ClipMyHorse's terms restrict viewing to private, non-commercial purposes. We link to them; we never embed or restream them.

• Facebook and Nextdoor groupsNot used

  Scraping these violates the platforms' terms of service. We do not scrape them, and nothing posted in a private group is ingested, quoted or indexed here.

• Article body text from any publisherNot used

  We never republish a publisher's article body. This is enforced in code rather than promised in a policy: the article text is searchable in our index but is never placed in the AI's prompt, and a test fails the build if that ever changes.

How the AI works, honestly

Fiona is the assistant on this site. Here is what it does and what it will not do.

It answers from local sources, and cites them

Fiona answers from what we have indexed from the sources listed on this page, and every answer carries citations you can click through to the original. Each record it works from includes the date it was published (or when the event is) and which source it came from, so an answer can tell you how current something is — and you can check it yourself.

It refuses rather than guesses

When local sources do not confirm something, it says so and points you at the official source instead of producing a plausible answer. A refusal is the designed behaviour, not a failure.

It never speaks as the Village

This is an independent service. Fiona is not a Village of Wellington channel and does not answer as one. For official Village information the answer is a link to wellingtonfl.gov.

It cannot reproduce a publisher's article

The AI is shown headlines, excerpts, dates and links — never a publisher's article body. It cannot summarise text it is never given.

It stays neutral on local disputes

On contested local matters — the equestrian preserve and development questions above all — the rule is facts, dates and citations only, with no position taken and no sponsor with a stake in the outcome.

Emergency information is human-reviewed

Anything safety-critical is reviewed by a person before it goes out. In an emergency, official sources come first: follow the Village, Palm Beach County and the National Weather Service.

It can be switched off

The AI passes that write editorial copy run under hard per-run limits and a kill switch. If something reads wrong, it can be stopped without taking the rest of the site down.

Our crawler

If you run a site we read, you will see us in your logs as WellingtonRoundupBot/0.1 (+felican.ai).

It identifies itself

Our crawler sends the user-agent WellingtonRoundupBot/0.1 (+felican.ai) on every request, with a contact address in it, so any site owner can see us in their logs and find us.

It is slow on purpose

At most one request per second to any single host, with retries backed off and capped. We fetch conditionally where a site supports it, so an unchanged page costs the publisher a 304 and nothing more.

It runs rarely, not constantly

Every source declares how often it may be crawled, and the pipeline refuses to run any crawler that does not. A weekly paper is checked weekly, not every hour.

It will not impersonate a disallowed bot

Some publishers disallow specific AI crawlers by name. We refuse at startup to run under any user-agent on that list — the process will not start rather than fetch under a forbidden identity.

One honest exception

One site (the National Polo Center) rejects non-browser user-agents outright, so we present a browser user-agent there. We still fetch only public pages, at the same one-request-per-second rate, on the same daily schedule. We would rather tell you that than claim a blanket rule we do not keep.

It reads and obeys robots.txt

Before our crawler fetches anything from a site, it fetches that site's robots.txt, and it obeys it. A page a site disallows is not fetched — the crawler refuses and records the refusal, rather than fetching and logging a warning. Where a site publishes a Crawl-delay we slow down to it. If we cannot read a site's robots.txt at all, we stop rather than assume permission. A person still reviews every source's terms before it is added; the parser is the floor, not the ceiling.

Three documented exceptions, named

Three machine endpoints are called despite a blanket robots.txt disallow: the National Weather Service's public JSON API, the OpenStreetMap Overpass query API, and a text-extraction proxy. Each is an endpoint its operator publishes for programmatic use, each is written down in the code with the reason and the date, and each logs a warning every time it is used. We would rather list them than claim a blanket rule we do not keep.