feat: add Rotten Tomatoes mirror site

[derenlei]

Add Rotten Tomatoes mirror site

Mirror URL: https://www.rottentomatoes.com/
Port: 40015

Overview

A faithful Rotten Tomatoes clone with real movie data scraped from RT. Supports browsing, searching, movie detail pages, celebrity pages, user authentication, watchlists, ratings, and reviews.

Data Summary

Table	Rows
Movies	147
Persons	895
Movie Cast	909
Genres	18
Movie-Genre links	335
Critic Reviews	258
Audience Reviews	28
Users	4
Watchlist Items	16
User Ratings	12

Assets: 147 real movie posters + 777 actor headshots (906/909 cast members have photos)

Features

• Browse — filter by genre, certified fresh, and streaming platform (At Home)
• Search — title-based movie search
• Movie detail — Tomatometer, Audience Score, cast, synopsis, critics consensus, Movie Info (director, producer, screenwriter, studio, release date, box office, runtime, rating)
• Celebrity pages — bio, filmography
• Auth — register, login/logout
• Watchlist — add/remove movies
• Ratings — rate movies 0.5–5.0 stars
• Reviews — write and delete reviews
• Account — profile page, edit name/email

Benchmark Users

Username	Email	Password
alice_j	alice.j@test.com	TestPass123!
bob_c	bob.c@test.com	TestPass123!
carol_d	carol.d@test.com	TestPass123!
david_k	david.k@test.com	TestPass123!

Tasks

20 benchmark tasks covering: browse + filter, search, movie info lookups, authentication, watchlist CRUD, ratings, and review operations.

Screenshots

Verification

Health check: {"movies": 147, "ok": true, "persons": 895, "site": "rotten_tomatoes"}
Reset: {"pid": ..., "ready": true, "site": "rotten_tomatoes"}

HuggingFace Assets

rotten_tomatoes.tar.gz uploaded to ChilleD/WebHarbor dataset (commit 7583bbd)

[MufanQiu]

Review — rotten_tomatoes (PR #26)

Verdict: REQUEST CHANGES.

We deployed the environment locally and ran the full review checklist; every finding below was independently reproduced, and fidelity was checked against the real upstream site.

Issues (summary)

• BLOCKER: .assets-revision pinned to 'main' — new-site tarball not on main, CI fail-closed
• BLOCKER: Port-slot collision with merriam_webster (port 40015) — branch is stale vs main
• MAJOR: T18 infeasible via UI — producer not searchable, no producer route
• MAJOR: T3 ill-posed/non-deterministic — two shared producers, two different 'third movies'
• MAJOR: seed_data.py cannot reproduce shipped DB (producer/screenwriter never seeded)
• MINOR: Answer-leak via browse count label for count tasks (T4, T6)

Mechanical checks: FAIL

• Byte-identity: PASS. After boot + serving routes, md5(instance/rotten_tomatoes.db) == md5(instance_seed/rotten_tomatoes.db) (5a5af6e6…). seed_all is gated by a single whole-function early-return, so /reset is byte-stable.

Visual fidelity: PASS

Convincing RT look: dark header, "Movies & TV Shows" hero, New & Upcoming / Popular Streaming / Certified Fresh / Top Box Office sections, dual Tomatometer/Audience badges with real posters.

Functional depth: PASS

All flows driven via Playwright and persisted to DB: login (all 4 benchmark bcrypt hashes validate TestPass123!), register, watchlist add/remove, rating, review, account edit. Form validation enforced server-side (Email, review Length min=10, NumberRange 0.5–5.0); CSRF on all POSTs.

Task quality: FAIL

20 tasks with broad coverage and good distractor density; ~13 are solid. But multiple have answer-leaks, ill-posedness, or UI-infeasibility that disqualify them for autonomous grading.

Required fixes before approval

1. (BLOCKER) Rebase onto origin/main and resolve the merge conflict / port collision: move rotten_tomatoes to index 16 / port 40016 in BOTH websyn_start.sh SITES and control_server.py SITES, bump Dockerfile EXPOSE to 8101 40000-40016, and update tasks.jsonl web to http://localhost:40016/ on all 20 tasks.
2. (MAJOR) Pin .assets-revision to a merged HF dataset SHA on ChilleD/WebHarbor that contains rotten_tomatoes.tar.gz (neither main nor 54882a6a was updated by this PR to include the new site).
3. (MAJOR) Kill the browse count-label leak for T4/T6: remove {{ movies|length }} movies found from browse.html, or re-scope these to a non-count answer that requires a detail page.
4. (MAJOR) Disambiguate T6 platform semantics: make the filter "available on" (substring/relationship) or normalize compound platforms, and set the answer key accordingly, so 8/12 vs 16/18 is unambiguous.
5. (MAJOR) Fix T18 solvability: add a producer browse/filter route, index producer in search, or make producer names clickable; otherwise rewrite so the filmography is discoverable through the UI.

[jackjin1997]

I’m taking over the review and remediation of this Rotten Tomatoes contribution. I’ll build a separate Review: PR on top of @derenlei’s original commits, preserve the original authorship, and link the reviewer PR here.

The review will cover current-main integration and pinned assets, browser/UI fidelity, task solvability, and deterministic verifiers backed by recorded executions and positive/negative tests. The existing review findings will be rechecked against the frozen candidate. Maintainers will retain final approval and merge responsibility.

[jackjin1997]

The reviewer continuation is now available as Draft #87: #87.

It preserves @derenlei's original commits and movie identities, with source-data and functional repairs, responsive UI comparisons, and seven proposed grading contracts. The companion immutable asset candidate is HF PR #55.

This remains a draft: full-image verification is blocked by local Docker storage recovery, and revised-task runs, human visual acceptance, and independent review are still outstanding. The linked report distinguishes completed offline tests from those uncompleted checks.