jev-trip
Explainable day-trip planner where an LLM drafts the itinerary and Jev screens places, compares transport and reviews each step.
Overview
Jev Trip, subtitled Two Minds, One Trip, is a day-trip planner for mainland Chinese cities that splits work among three parts. An LLM planner chooses the next activity and repairs flagged plans. Jev screens candidate places, compares transport options, answers typed planning questions and reviews the itinerary, with listings describing pass, flag or unknown Choice answers checked against the user's stated preferences. A deterministic controller owns facts and rules: timing, routes, IDs, versions, validation and retry limits.
Live mode uses Amap places and walking and public-transit routes, the configured LLM and Jev; a replay demo runs a scripted plan locally with no credentials so you can scrub through every planning and repair event. Explanations follow the user's language while place names stay in their originals. Current scope is a single day; it does not support multi-day trips, bookings, accounts, persistent history or verified opening hours, and results should be treated as drafts. Tests, typecheck and build scripts are included, and a live demo is hosted on Vercel.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Does this candidate place or planning step pass, get flagged, or is it unknown given the user's preferences? | pass · flag · unknown |
| Choice | Which transport option is better for this leg?(inferred) | — |
How it works
- Controller gathers places and route facts
- LLM proposes the next activity
- Jev screens candidates and reviews each step
- Flagged steps are repaired by the LLM within retry limits
Strengths
- Explainable, replayable planning events
- Facts and timing stay in deterministic code
- Keyless replay demo
Caveats
- Mainland China only, via Amap
- Single-day drafts; no verified opening hours
- Live mode needs Amap, LLM and TypeSafe keys
Getting started
npm install, cp .env.example .env.local, `npm run dev -- --hostname 127.0.0.1 --port 3100`.At a glance
- Type
- App
- Status
- prototype
- Language
- TypeScript
- Runtime
- Next.js
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Jev Trip is an explainable day-trip planner. The LLM plans ahead; Jev chooses and checks. Deterministic code handles route facts, time calculations, validation, and versioning.
logicrw/awesome-jev-projectsTrip planner where an LLM drafts the itinerary and Jev answers pass, flag, or unknown Choice questions to filter candidate places, check each planning step against the user's stated preferences, and catch unreasonable plans before they propagate.
cobanov/awesome-jevTwo Minds, One Trip. https://jev-trip.vercel.app/.
hellogumbo/awesome-jev
Frequently asked questions
What is Jev's role?
Screening places, comparing transport and reviewing each planning step, while the LLM writes the plan.
Can I try it without keys?
Yes, the replay demo runs locally with no credentials.
Where does it work?
Mainland Chinese cities, using Amap data.
Related projects
Prompt reviewer for Claude Code and Codex where Jev decides whether each prompt needs review before a light model lists blind spots.
CLIbetaTypeScriptPi extension that audits whether the rpiv-todo board matches visible work and nudges the agent when it drifts off-task.
PluginbetaTypeScriptPython kernel of Jev Choice/Score/Noul gates, caches and loops used to admit or reject Lean proof candidates in research pipelines.
LibraryresearchPythonPublished Jev client and claim check from the Fake/Real evidence-review site, where Jev filters sources and judges each claim.
ProjectproductionTypeScriptFail-open Codex UserPromptSubmit hook that asks Jev for task type, complexity, risk and execution mode and injects it as advisory context.
HookbetaPythonFork of codex-with-chatgpt where Jev plus fixed rules decide when to bring in web ChatGPT to plan or review and when Codex just acts.
Projectbeta