Amazon's online assessment ends its coding section with a 60-minute debugging task in an existing repository, and one of the repos candidates keep reporting is a movie database web app. You choose a stack at the start (reports name Spring Boot and Python; the Java version reported was a different variant of the app), read the task description and the failing tests, and fix the code until the tests pass. The built-in AI assistant can help you find code but will not make the fix. Reports describe two features with failing tests, and this page has one graded part for each:

Part 1 — Watch lists: the six failing tests. One candidate lists the six failing tests of the watch-list version: add a movie to a watch list, add a movie the list already has, delete a movie, add many movies and then delete them one by one, add to a watch list that does not exist, and delete from one. They advise checking whether each endpoint returns the correct status code, checks that the movie and the list exist, and saves the new movie to the database.

Part 2 — Movie search. In another version the search finds nothing. One candidate gives its rules: partial match, on the title, the stars or all of them, ignoring case, sorted by popularity, and only published movies. Another traced their failure to a query term that was never lower-cased while the stored values were.

This page is a stand-in. The code is a small Python backend written for AceOffer that reproduces the six reported watch-list tests, the reported search rules and the kinds of bugs candidates describe. The README's exact status codes and messages, the tie-break, the data and the particular bugs are ours; the solution says which bugs follow a report and which do not. Amazon's repo is a full-stack app with many more files, in the stack you choose.

The README (given at the top of the code). add_movie returns 201 with the saved list, remove_movie and get_watchlist return 200. Errors, checked in this order: a missing list is 404 Watch list not found (every handler); in add_movie a missing movie is 404 Movie not found and a movie the list already has is 409 Movie already in watch list; in remove_movie a movie the list does not have is 404 Movie not in watch list. Search returns 200 with the matching titles, most popular first, equal popularity ordered by title A to Z. Every error body is {"error": <message>}.

How the tests call it. Each test builds db = seed_db() and calls a handler through request(handler, *args), which turns a crash into a 500: get_watchlist(db, list_id), add_movie(db, list_id, movie_id), remove_movie(db, list_id, movie_id) and search_movies(db, query, field), where field is "title", "stars" or "all". A Response has status and body.

Other versions of this repo (not built here): the home page's movie recommendations not showing; a custom movie list that will not take a movie; an actor site where users follow actors and should be notified when a followed actor gets a new movie, whose tests did not cover every bug the task described; and a Django report, likely the same app though it does not name it, with filters by genre, director and writer, rating and year bounds, and sort tie-breakers.