Tour Rescue

• 

  Cancelled Leipzig: compare Nuremberg, Frankfurt and a day off. Saved sample with synthetic Qloo evidence.

• 

  Synthetic experiment: extra-distance cap 350 → 354.1 km changes Nuremberg → Frankfurt. Daily 6-hour limit and four shows stay fixed.

• 

  Promoter brief for the inspected Nuremberg option: constraints, evidence and unknowns. A reviewable draft; nothing is sent or booked.

Inspiration

The show’s off. The tour isn’t.

The band is already on the road when tonight’s cancellation arrives. The crew needs a plan. Four other shows still have to happen. For the person everyone calls, a blank date becomes a decision they have to stand behind.

Tour Rescue turns that moment into three checked directions and a concrete brief for the next promoter conversation.

Try the guided demo · Watch the two-minute film. Mobile first, with a tablet and desktop workspace.

What it does

Start with a cancelled date and compare a replacement, a detour and a day off. Qloo supplies relative cultural-fit evidence across a fixed set of cities. Route checks combine that evidence with directed road distances and declared constraints while keeping four confirmed shows locked.

The manager can inspect the recommendation, rejected options, missing facts and the assumption that could change the answer. The guided journey ends with a promoter brief and an audit containing the current inputs, checks and sources.

The labelled synthetic example makes sensitivity tangible. Raising the additional-distance cap from 350 to 354.1 km changes the recommendation from Nuremberg to Frankfurt am Main. The daily-driving limit and confirmed concerts stay fixed. The interface distinguishes the current recommendation from the option being inspected.

Optional AI Rescue researches leads using the current Qloo result, route evidence, relevant excerpts from private band notes and public sources. A person explicitly accepts a lead into comparison. The same checker then evaluates it, preserves the other comparison card and carries the accepted sources into the audit.

When the guitar breaks. Or the delivery never comes.

Tour SOS demonstrates a separate incident → local supplier options → budget review journey. The Ho Chi Minh City example uses illustrative suppliers, quotes and arrival estimates. It does not contact anyone, confirm availability, place an order or make a payment.

How we built it

A pure TypeScript core separates evidence, constraints and recommendation policy. Each artist is ranked within the same frozen city universe; missing values remain explicit. Without Qloo taste evidence, the live workflow requires evidence instead of letting the model invent cultural scores. Comparison-artist omissions, a distinct control artist and separate fit and distance axes help users examine the result.

Cloudflare Pages and Workers serve the mobile-first PWA and protect provider credentials. Claude Sonnet 5.5 uses tools for research and revalidation. An OSRM road snapshot makes the saved route experiment repeatable. The workspace exports an A4 brief, a PNG and a JSON audit so the reasoning can travel into the next conversation.

Challenges

Cultural fit is relative evidence, not a ticket-demand forecast. Venue availability, capacity, fees and terms remain UNKNOWN until confirmed. This prototype prepares comparisons and drafts; no promoter is contacted and no booking is made. SOS and agency extensions are labelled samples. Rescued revenue, faster booking and customer adoption remain unmeasured.

Accomplishments

The final release passes 398 offline tests, strict type checks and builds. A measured live AI session rechecked a human-accepted Kassel lead and retained its sources in the exported audit.

We also recorded real Qloo profiles for Radiohead, Arctic Monkeys and Dua Lipa across 24 cities, with complete coverage for each. All three retained Nuremberg at the 350 km cap; control and comparison omissions also kept the winning city unchanged. That study is shown separately from the synthetic breakpoint experiment.

What we learned

A strong recommendation includes the condition under which it would change. Equally, an unchanged result is evidence worth preserving. Clear sources, explicit unknowns and a reproducible comparison make the next promoter discussion more useful.

What's next

Work through one real cancelled-date brief with a tour manager. Compare the shortlist with their map, spreadsheet and contact workflow, and identify which evidence changes the next call.

Our business hypothesis is per-tour or agency licensing, with a later SOS marketplace earning fees or commissions on completed local services. That requires onboarding a vetted network for backline, repairs and urgent delivery, and building fulfilment and payments. Willingness to pay, service economics and revenue remain unmeasured.

Built With

• claude-sonnet-5-5
• cloudflare-pages
• cloudflare-workers
• durable-objects
• html
• osrm
• pwa
• qloo-api
• sqlite
• typescript