# Lore Loop

> A planner for lore trips - short trips taken for the stories behind a small town (legends, odd local history,
> haunts, roadside oddities, festivals and food with a backstory). The traveller types where they stay as a latitude
> and longitude, the stops they already know (one per line, "name | lat, lon | minutes", or a GPX file), a start date,
> one to three days and what they have heard. Free, in the browser: the stops are ordered into a loop from the base
> (nearest neighbour, then 2-opt), drive times are ESTIMATED from straight-line distance x 1.35 at a chosen average
> speed (short hops under 0.8 km are walks), sunrise and sunset are computed for every day at the base, the loop is cut
> into days that fit the daylight, dusk and night stops go last with a wait for dusk, and an SVG map draws the loop by
> day. Two metered lanes run over the same trip: SCOUT (mood eerie / quirky / historic / mixed; 4-8 stops with a story
> told as lore, era, approximate position, visit length, best time, a question to ask a local, 1-3 checkable claims and
> a respect note; what was left out and why; tips; what to confirm before going) and VET (verdict mostly-documented /
> mixed / mostly-folklore; every claim rated documented, likely, contested, folklore, doubtful or unknown with a
> confidence, reasoning, a kind of source to check and an honest one-line retelling; red flags; kinds of source).

Every reply is reconciled against the page's facts: your stops kept or dropped, pins moved, positions far from the
base, opening times or prices stated in a story, web addresses the page cannot verify, unrated claims, and a verdict
that disagrees with its own ratings. Positions placed by the model are approximate; drive times are estimates; ratings
are a model's judgement, not verification. The app never states opening hours or prices and never suggests entering
closed sites.

Derived from the SkillSafe agent skills @lijigang/ljg-travel (deep cultural travel research before you go - the scout
lane) and @elvisun/fact-check (claim extraction and calibrated verification ratings - the vet lane).

## Pages

- [App](https://lore-loop.skillsafe.ai/): plan a trip, see the map and the day strips, run either lane.
- [API](https://lore-loop.skillsafe.ai/api.html): the scout and vet tasks over HTTP - input fields, reply contracts,
  worked examples, eight languages.
- [lorekit.js](https://lore-loop.skillsafe.ai/lorekit.js): the free lane (positions, distances, sunrise/sunset, loop
  order, schedule, input building, reply checks, GPX/ICS/CSV export); runs in node too.

## Contract in brief

- Input: `task` (scout | vet), `trip_title`, `area`; scout adds `base`, `dates`, `prefs`, `stops` (JSON-encoded),
  `heard`, `notes`; vet adds `claims` (JSON-encoded `[{id, stop, stop_name, text}]`), `pasted`, `heard`.
- Output: one JSON object per task. Scout stops are `{id, name, kind, story, era, lat, lon, pos, minutes, best_time,
  ask, claims[], respect}`; vet ratings are `{id, rating, confidence, why, check, retell}`.