- blindtest is only available with >=3 players AND >=2 humans (a bot can't be DJ):
lobby gate + tooltips updated, and server game:start rejects otherwise
- new botDifficulty setting (easy/normal/hard) drives how often bots answer
correctly (quiz + blindtest who-added / title-artist); host picks it in the
lobby when at least one bot is present
Verified e2e: blindtest start rejected with 1 human + 2 bots (NEED_THREE).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- server-side bot players (isBot): host adds/removes them in the lobby
(lobby:addBot / lobby:removeBot), capped at 8 players
- bots auto-vote each round: GameRound.botAnswer generates a plausible answer
(quiz ~50% right, blindtest guesses a random player / title); engine schedules
the votes within the answer window and triggers early end when all have voted
- bots are never DJ (pickDj excludes them) and are excluded from the blindtest
track-submission gate (they don't submit)
- snapshot players carry `bot`; lobby shows a 🤖 marker + add/remove controls
Verified e2e: 2 bots added, they vote, game ends with all 3 players scored
and rounds end early.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- shared: RoundRecap gains answers[] (RoundPlayerAnswer: text + correct)
- server: quiz/blindtest recap() build per-player answers (choice text / free
text / blindtest guess formatted by mode), with correctness
- client: recap lists each player's answer per round, green/red by correctness,
with their points when they scored
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- shared: RoundRecap (+ scorers); game:end carries recap[]
- server: GameRound.recap(ctx) per mode (quiz: prompt + correct answer +
image; blindtest: "title — artist" + addedBy + links); engine accumulates a
history per round (with that round's scorers from the score deltas) and sends
it at game:end
- client: GameEndView shows a collapsible "Récapitulatif des manches" — each
round's label, answer, media thumbnail (image/youtube) and who scored
- engine test dummy round implements recap
Verified e2e: game:end recap lists every round with answer + scorers.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Server:
- quiz-round handles image_reveal like a text-answer format (tolerant match
on acceptedAnswers), exposing imageUrl in the round payload; degressive
scoring comes for free from the existing speed bonus (earlier = blurrier =
more points)
- repo includes image_reveal in the pool (requires imageUrl + acceptedAnswers)
- back-office: image upload (@fastify/multipart) saved to UPLOADS_DIR and
served statically on /uploads (@fastify/static); create supports image_reveal
- env: UPLOADS_DIR; gitignore uploads/
Client:
- admin form: "Image à deviner" — upload + preview + accepted answers
- in-game: ImageReveal shows the image with a time-cadenced decreasing blur
(synced via server timestamps) + text input; reveal shows the clear image
and the answer
shared: QuizQuestionPayload.imageUrl
Verified against real server+DB: upload (url), static serve (200), create
(201), validation (400).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mixed mode (default):
- gameType is now mixed | quiz | blindtest, default "mixed"; the lobby
interleaves quiz + blindtest rounds so modes alternate
- lobby reworked: 3-way game-type switch, quiz count and/or blindtest config
+ track submission shown per selected type
Free-text quiz questions:
- quiz supports the `free` format (e.g. maths) alongside mcq/truefalse:
payload without choices, {text} vote, tolerant matching on acceptedAnswers
- shared QuizQuestionPayload.choices optional; reveal truth carries answer
- match util moved to game/match.ts (shared by quiz + blindtest)
- repo/seed include free + acceptedAnswers; in-code bank gains free questions
- client quiz view renders a text input for free questions
Mode-aware transitions:
- RoundTransition themed per mode (colors/label/emoji); announces the mode
on change (quiz <-> blindtest), lighter teaser within the same mode
- custom media buckets per mode: assets/transitions/{quiz,blindtest}/ + shared
- store tracks roundModeChanged
Verified e2e: mixed game alternates quiz/blindtest to completion; free-text
scoring covered by tests (22 pass).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Drizzle (postgres-js) setup: schema (quiz_category, quiz_question,
quiz_played, game_history), client (nullable — no DATABASE_URL → null),
drizzle.config, initial migration, migrate script
- Open Trivia DB seed: curated geek categories, mcq + truefalse, base64
decode (incl. type/difficulty), 1 req/5s rate limit, idempotent
(prompt unique); also seeds the FR in-code bank as source 'manual'
- quiz wired to the DB: per-room pool preloaded at game start (loadQuizPool
excludes already-played), records quiz_played for cross-session
anti-repeat; falls back to the in-code bank when no DB
- docker-compose for local Postgres; db:* scripts; DATABASE_URL in env
Roadmap V1 step 5. Verified against a real Postgres: migrate + seed (270
OpenTDB + 10 manual), DB-backed game serves OpenTDB questions and fills
quiz_played; no-DB fallback still serves the in-code bank.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
