Netease's m701/m801.music.126.net CDN RST connections from FFmpeg's
default Lavf UA. The bot's frame loop treats an early ffmpeg exit as
trackEnd and advances the queue, which surfaced as songs auto-skipping
mid-playlist. Add a browser UA + music.163.com Referer for these URLs,
extract args into a tested buildFfmpegArgs(), and harden reconnect
flags (delay_max 5 -> 30, plus reconnect_on_network_error /
reconnect_on_http_error). Verified by A/B running ffmpeg against the
same fresh CDN URL: legacy args got 0 bytes + "End of file" reading
HTTP response; fixed args streamed the full track to completion.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two corner-case fixes for the prev-history + play-next feature:
1. cmdPrev only tried queue.prev() once. instance.playNext's auto-
advance retry-skip pushes failed songs into the same history stack,
so a single prev frequently lands on an unplayable song — returning
"Cannot play previous song" while leaving queue.currentIndex stuck
mid-failure (causing next() to skip past the actually-playing song).
Retry up to 4 times so prev finds a playable history entry, matching
the retry budget already used by playNext for auto-advance.
2. SongCard.song-actions has opacity:0 by default and is revealed via
parent :hover. Touch devices have no hover, so all three action
buttons (Play / Play Next / Add) were invisible to phone/tablet
users. Add @media (pointer: coarse) → opacity:1 to always show on
touch.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Final review of caeef65 caught a stale-currentIndex bug in
/play-next-song and cmdPlayNext: when the player is idle but
queue.currentIndex >= 0 (natural end-of-track, or playNext gave
up after retries without queue.clear), addNext splices mid-queue
and playAt(size-1) jumps PAST the inserted song to whatever
was last. Capture the insertion slot before calling addNext and
promote that exact index instead.
Add a regression test covering the [a,b,c,d] queue with
currentIndex=1 case -- splice at 2 yields x at index 2, but
size-1 would point at d (index 4).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirror of !play / /play-song but uses queue.addNext to splice in at
currentIndex+1 instead of clearing the queue. Idle bot still starts
the song immediately. Adds 'playnext' / 'pn' to AUDIO_COMMANDS, the
command switch, and the help text.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Splices into currentIndex+1 and shifts both playedIndices and the
history back-stack to keep references valid. Falls through to plain
push when nothing is playing so the idle-bot "add → start playing"
flow is unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Code review of 390d3fa flagged that dropping playedIndices.clear() from
playAt is an unnecessary behavior change. The two operations (push
history, reset random pool) are not in tension — restoring the clear
preserves the original 'explicit pick restarts shuffle' semantic that
Random mode users rely on, while still tracking the prev-history stack.
Also add a regression test that the 50-entry HISTORY_LIMIT cap works.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
In random modes, prev was just doing currentIndex-1 in the array, which
has no relationship to what the user actually played before. Add a
50-entry back-stack that's pushed by next/playAt, popped by prev, reset
on play/clear/setMode, and shifted by remove. Sequential and Loop modes
keep their old fallback for the case where history is empty.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Inert in this commit — no callers yet. Sets up the back-stack used
by the upcoming history-aware prev rewrite.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Surface failures + tighten edges:
1. Toast for /play-song & /play-playlist failures. Backend now returns
{ok, message} (localized in Chinese to match the rest of the UI).
Store stashes a notification on ok=false; new Toast.vue mounted in
App.vue shows it for 3-5s then fades. Clicking the X dismisses
immediately. Sits above the player on desktop and above the mobile
tabbar on phones.
2. QQ collected playlists pagination. fcg_get_profile_order_asset.fcg
returns max 30 per call; we now loop using has_more / short-page
detection up to a 300-playlist hard cap. Single-call users (typical)
exit the loop on the first iteration so no extra requests.
3. getPlayableSongIds chunking. 100 mids per request keeps URL well
under 8KB; chunk-level errors are isolated so a transient blip on
one chunk doesn't poison the whole batch. Returns null only when
every chunk failed (caller falls back to sequential retry).
4. SourceTabs single-source mode now renders a small subdued "网易云"
or "QQ" label instead of vanishing entirely, so the user always
knows which platform's data they're looking at.
5. Hide the "我的歌单 N" count badge when N=0 — Home and Library no
longer show "我的歌单 0" with an empty grid.
6. Auth state change invalidates fetchHomeData cache. Previously, a
user who logged out as account A and into account B within 5
minutes would see A's playlists. Now we always re-check auth at
the top of fetchHomeData and bypass the TTL cache when authStatus
has changed.
Out of scope:
- NetEase analogous batch precheck (per request).
- 60s TS3 UDP idle disconnect — that's the bundled @honeybbq/teamspeak-
client UDP layer kicking when no server packet arrives in 60s. It's
baked in (constant `v=6e4`) and not exposed as an option, and root
cause is server-side or network-layer behavior we can't reach from
here.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three small but real correctness fixes from auditing recent commits:
1. getPlayableSongIds returned an empty Set for both "endpoint failed"
and "succeeded but all unplayable" — the caller couldn't tell which.
Return Set | null now: null = error (fall through to sequential
retry), empty Set = authoritative "all unplayable" (short-circuit
to a clear message instead of wasting 20+ retries).
2. Web store: fetchHomeData unconditionally wrote lastFetchTime even
when every fetch rejected (network blip, server down). That cached
the failure for 5 minutes — user had to hard-reload to recover.
Now only commit lastFetchTime if at least one auth-status call
succeeded.
3. Settings profile section: if GET /profile failed, profileConfigs
stayed undefined and the row showed "加载中..." forever. Track a
per-bot error state and render an inline "加载失败 / 重试" link
so the user can recover without page reload.
Out of scope but documented:
- ein=29 hardcoded in fetchCollectedPlaylists (no pagination yet —
users with 30+ collected QQ playlists get truncated).
- /play-song single-failure UX (returns "Cannot play" message but
frontend ignores; needs a global toast/notification primitive).
- NetEase has no analogous batch precheck (could surface same
"click and wait silent" issue if user has region-restricted NetEase
playlists).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Many users' QQ playlists (especially collected/subscribed ones) contain
a large fraction of songs that return result=104003 (region/copyright
restricted) for the current account. Sequential retry-skip wasted time
guessing — for the user's ACG古风 collected playlist, only 3 of 115
songs are actually streamable, so a 20-retry budget had < 50% chance
of finding a hit before giving up.
Add a getPlayableSongIds(ids) method to the QQ provider that calls the
upstream /getMusicPlay with a comma-separated batch of mids — the
wrapper resolves all of them in a single upstream call (~2-3s for 100+
songs). /play-playlist now duck-types this method and, when present,
filters the playlist down to playable songs before queueing.
Response message reports "Loaded N of M songs (rest are copyright/
region restricted)" so the user sees exactly what's happening instead
of guessing why the queue is short or playback didn't start.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous fix for "播放全部 没有反应" hooked into playNext but had
two problems:
1. Used !!queue.current() to detect success, which is always true
after queue.next() advanced — so the response message would lie
about playback starting even when all retries failed.
2. The default 3-retry budget is sized for "next song couldn't
resolve, skip it" cases. User-initiated playlist plays commonly
hit long contiguous runs of QQ 104003 songs in collected playlists
(entire ACG/古风 packs can be uniformly unstreamable in some
regions), so 3-4 attempts wasn't enough.
Make playNext return whether a song actually started, and accept a
maxRetries parameter. /play-playlist now uses 20 retries — high
enough to clear typical unplayable runs, bounded enough that fully
unstreamable playlists still surface a clear "none were playable"
message rather than hanging.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
QQ Music sometimes refuses to issue a play URL for individual songs
(upstream returns result=104003, "no copyright/payment"). The
/play-playlist endpoint only called resolveAndPlay once on the first
song, so when that one was 104003 the bot would sit silently with no
feedback — exactly what looked like "播放全部 没有反应".
Make BotInstance.playNext() public and call it as a fallback when the
first resolveAndPlay fails. playNext already has the 3-attempt
retry-skip used by trackEnd auto-advance, so we get the same
graceful skip behavior for explicit playlist plays.
Also tighten the response message so the user can tell when the bot
loaded songs but none were playable.
Single-song /play-song still returns "Cannot play" on failure (no
queue to advance through); UX surfacing of that goes in a follow-up.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The /playlist/:id/detail route was hardcoded to call the NetEase API
path /playlist/detail on whatever provider was selected. For QQ this
hit a non-existent path, so clicking into any QQ playlist showed
"歌单不存在或加载失败".
Replace the platform-specific hack in the route handler with a proper
getPlaylistDetail method on MusicProvider. Implement it for NetEase
(porting the existing /playlist/detail logic) and QQ (calling
/getSongListDetail and reading from response.cdlist[0]).
Pre-existing bug exposed by the QQ source tab.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three fixes to the QQ Music provider, all surfaced by enabling the QQ
tab on Home/Library:
1. getUserPlaylists field names were wrong. The upstream
fcg_get_profile_homepage actually returns title/picurl/subtitle,
not dissname/imgurl/song_count, so all playlist cards rendered as
empty placeholders. Map title->name, picurl->coverUrl, and parse
'(\d+)首' from subtitle for songCount.
2. getUserPlaylists only returned playlists the user CREATED. Now also
fetches COLLECTED playlists from c.y.qq.com fcg_get_profile_order_asset
(reqtype=3, requires g_tk derived from p_skey cookie) and concatenates
them after the created list — same order as the QQ desktop app.
3. Add getDailyRecommendSongs implementation backed by /getNewSongs
(新歌速递, ~20 newest songs). The QQ provider previously didn't
implement this method, so the daily-recommend QQ tab was empty.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Library.vue: use Song type from store so SongCard's strict platform
union accepts the data (was platform: string, broke after merge tightened
SongCard prop type).
database.test.ts: switch toEqual -> toMatchObject so getBotInstances()
returning extra profile_* schema columns no longer fails the assertion.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
RandomLoop randomly selects from queue, causing the same song to play
multiple times when the queue is small (~3 songs from personal_fm API).
Random mode ensures each song plays once before repeating, relying on
refillFm() to fetch fresh songs before exhaustion.
Also fix the proactive refill condition to use unplayedCount() instead
of size() - getCurrentIndex(), which was meaningless for non-sequential
play modes.
Silent failure: logs showed "Client properties updated" / "Description
updated" / "Avatar updated" even though the bot's nickname never
changed and the avatar stayed as "loading image" on clients.
Root causes:
- TS6HttpQuery.clientUpdate ignored non-2xx responses, so 400 (bad
parameter) and 403 (insufficient permission) were reported as success.
- updateClientProperties built TS3-escaped strings (\\s for space) then
split them back into JSON props, so TS6 received literal backslashes
and rejected the nickname silently.
- handleFeatureError only matched textual "permission" errors; HTTP
4xx statuses weren't recognised and the feature retried every song.
- doAvatarUpload had no per-step logging, making it impossible to tell
whether a broken avatar came from init, the TCP 30033 transfer, or
the client_flag_avatar command.
Fixes:
- Add HttpQueryError (status/body/path); clientUpdate throws on non-2xx.
- Build a raw property map in updateClientProperties; escape only on
the TS3 wire path.
- Log HTTP status and updated prop names on success.
- handleFeatureError now treats HTTP 400/401/403 as unrecoverable.
- Debug-log each step of doAvatarUpload plus bytes/elapsedMs on success.
https://claude.ai/code/session_018NrpGWbQQTrahUVXyea5Jy
The "复制专属链接" button silently failed on public-IP HTTP deployments
because navigator.clipboard requires a secure context. Now the link is
always revealed in a modal with a read-only input (select-all on focus),
so users can copy manually even when clipboard APIs and execCommand both
fail. The dialog still tries to auto-copy when possible.
Also adds a publicUrl config option that overrides window.location.origin
for link generation (useful behind reverse proxies / with custom domains),
exposed via GET /api/config/public-url, and a trustProxy flag so Express
honors X-Forwarded-* when fronted by nginx/Caddy/Cloudflare.
https://claude.ai/code/session_019FSX3S3UUcKEYWanYmoqUv
The @sansenjian/qq-music-api module's export structure varies between
versions (2.2.10 vs 2.2.11+). Add fallback chain to find the Koa app:
try candidate.listen first, then candidate.default.listen.
Also resolve leftover merge conflict marker in instance.ts.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Wrap clientedit (description) with 5s timeout to prevent blocking
channel description and now-playing updates if the command hangs
- Check generation counter in clearAvatar to avoid clearing a newer
song's avatar when stop→play happens in quick succession
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add generation counter to prevent stale avatar uploads from
overwriting newer song's profile when rapidly skipping tracks
- Fix nickname truncation to use UTF-8 byte length instead of JS
string length (TS3 counts bytes, Chinese chars are 3 bytes)
- Add timeout to clearAvatar file transfer (was missing)
- Extract withTimeout helper to deduplicate timeout logic
- Add BiliBili CDN thumbnail resize support (@200w_200h)
- Bump generation on reconnect to discard in-flight updates from
the old connection
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Description via clientupdate is rejected (error 1538) on TS3 full-client
protocol. Switch to clientedit on the bot's own clid, which is how
TS3AudioBot handles it. Requires b_client_modify_description permission.
Also add sendCommandNoWait to TS3Client for fire-and-forget commands
(clientupdate, channeledit) that don't return timely responses.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add BotProfileManager that updates the bot's TeamSpeak presence when
songs change: album cover as avatar, song info in nickname, away status
toggled on stop/play. Each feature is independently configurable via
REST API and persisted to the database. Permission-safe — features that
fail due to insufficient server permissions are silently disabled until
reconnect. Description falls back to nickname display on TS3 (only
supported via TS6 HTTP Query).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Three corner cases fixed:
1. Removing the currently-playing song caused the next song in the array
to be silently marked as "played" and skipped. Root cause: playedIndices
was updated in next() by marking currentIndex, but after remove() shifts
currentIndex, it pointed to the wrong song. Fix: mark songs as played
at play-time (in play/playAt/next/prev) instead of at next-request-time.
2. Using prev() in Random mode could cause a song to play twice — the
song navigated to via prev() was not recorded in playedIndices, so
next() could randomly select it again. Fix: prev() now marks the
returned song as played.
3. Switching to Random mode mid-playback could cause the current song to
repeat because setMode() cleared playedIndices without preserving the
currently-playing song. Fix: setMode() now re-adds currentIndex after
clearing.
https://claude.ai/code/session_01W3ZncxL5VfdZeB4qqYWDmY
In Random (shuffle) mode, the queue's next() method would return the same
song indefinitely when only one song was in the playlist, and never terminate
even with multiple songs. This happened because played songs were not tracked.
Added a playedIndices Set to track which songs have already been played in
Random mode. Once all songs have been played once, next() returns null to
stop playback — matching the expected behavior where Random plays each song
once in random order, while RandomLoop is the mode for infinite shuffling.
https://claude.ai/code/session_01W3ZncxL5VfdZeB4qqYWDmY
The ffmpeg-static npm package bundles a pre-compiled binary that passes
`ffmpeg -version` but crashes with SIGSEGV during actual audio processing
inside Docker containers (incompatible glibc/missing shared libraries).
Changes:
- Install system FFmpeg via apt-get in Docker production stage, which is
always compatible with the container runtime
- Reverse FFmpeg resolution priority in player.ts: prefer system FFmpeg,
fall back to ffmpeg-static (for non-Docker environments like Windows)
- Optimize Dockerfile multi-stage build: compile native modules (opus,
better-sqlite3) in builder stage and copy to production, eliminating
the need for build tools (python3, make, g++) in the production image
- Fix qq-music-api dependency from file:../qq-music-api (path outside
Docker build context, breaks npm ci) to npm registry ^2.2.10
https://claude.ai/code/session_01JAV8sBokoifKh4Hc8XbJws
Repoints the `@sansenjian/qq-music-api` dependency from the public npm
release to a local fork at ../qq-music-api, which ships a corrected
getMusicPlay that:
- drops the hardcoded-sign GET path (no longer honored by QQ's vkey
server for VIP entitlement lookups)
- POSTs JSON directly to u.y.qq.com/cgi-bin/musicu.fcg (mirroring
the library's own getLyric.ts pattern)
- extracts qqmusic_key from the forwarded cookie and passes it as
`comm.authst` — the inline auth field the jsososo/QQMusicApi
reference implementation sets
- uses `ct: 19` (was 24) to match the community reference
For accounts that actually have entitlement to a given track, this
now returns the real VIP URL. For accounts that don't, QQ's vkey
server still returns result=104003 with empty purl — this is correct
server-side behavior and not a bug. Verified by observing the real
QQ Music web player on y.qq.com fall back to the same 30-second
preview on a logged-in account that lacks the specific track tier.
Supporting changes:
src/music/api-server.ts
The fork (v2.2.11) stopped auto-starting a Koa server on import —
it only listens when run as `require.main`. Explicitly import the
default Koa app and call .listen() with a server handle we can
clean up on shutdown. Without this fix, port 3200 silently fails
to bind and every QQ endpoint 502s.
src/music/qq.ts (getSongDetail)
The library's /getSongInfo endpoint returns upstream code 500001
because its param format no longer matches QQ's current API.
resolveAndPlay only needs `id` + `platform` to fetch a play URL,
so fall through to a minimal stub on /getSongInfo failure. This
unblocks /play-by-id and /add-by-id for QQ — they had been
returning "Song not found" for every QQ track regardless of
entitlement.
scripts/qq_browser_login.py
Visible-browser diagnostic tool that opens Chromium at y.qq.com,
auto-detects login via uin cookie poll, captures the full
post-login cookie set, tests it against /getMusicPlay for 稻香,
and writes the cookie to data/cookies/qq.json only if VIP
actually unlocks. On failure, dumps the full cookie to
data/cookies/qq.browser-capture.json for OAuth-vs-browser diff.
scripts/qq_verify_entitlement.py
Companion diagnostic: opens the real QQ Music web player at a
specific song's detail page so the user can manually click play
and verify whether their account has entitlement — independent
of any code path in this project. If the browser plays the full
song, HTTP 104003 is a request-signing issue; if the browser
also falls back to a 30-second preview, the account lacks the
tier/album purchase and no code fix can change that.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
QR login against QQ Music has been silently broken: every call to
checkQrCodeStatus returned "expired", so the scan-and-confirm cycle
never completed even when the user successfully scanned the code. The
root cause was four independent bugs in our wrapper talking past the
library's actual HTTP shape.
1. getQrCode lost ptqrtoken.
/getQQLoginQr returns { img, qrsig, ptqrtoken }, but we stored only
one of them in the single `key` field (picking qrsig, falling back
to ptqrtoken). The polling endpoint needs BOTH — passing only one
fails with 400 "参数错误". Fix: pack both into the opaque `key` as
"qrsig|ptqrtoken" and split on the receive side.
2. checkQrCodeStatus used GET.
@sansenjian/qq-music-api 2.x registers /checkQQLoginQr as POST only
(router.js: `router.post('/checkQQLoginQr', ...)`). GET returns 405
Method Not Allowed, axios throws, the catch returns "expired".
Fix: api.post(url, null, { params }).
3. checkQrCodeStatus parsed the wrong response shape.
The endpoint uses customResponse, not successResponse, so axios sees
the body directly (no { response: ... } wrapper). The actual shape
for each state is:
waiting: { isOk: false, refresh: false, message: '未扫描二维码' }
expired: { isOk: false, refresh: true, message: '二维码已失效' }
success: { isOk: true, message: '登录成功', session: { cookie } }
We were looking for a numeric `code === 0/1/2` field that does not
exist, so every state fell through to "expired". Fix: switch on
isOk / refresh / message.
4. Cookie read from the wrong path on success.
On isOk=true the cookie lives at res.data.session.cookie, not
res.data.cookie — so even if everything else had worked, the cookie
would never have been saved. Fix: read session.cookie.
Also rewrites getAuthStatus to actually validate the cookie:
5. getAuthStatus hit a non-validating endpoint.
/getUserAvatar is not registered on the library's main router; the
real route is /user/getUserAvatar, and even that just builds a
static avatar URL from a uin without round-tripping through QQ
Music with the cookie. The result: the bot happily persisted any
user-supplied cookie to disk and sent it on every request while
every downstream login check returned "not logged in". Fix: parse
uin from the cookie, call /user/getUserPlaylists (which actually
hits QQ Music with the cookie), and derive the avatar URL from the
uin via q.qlogo.cn/headimg_dl.
This is the same getAuthStatus fix that was sitting on the
claude/bot-shutdown-disconnect-cwFvF branch, now combined with the
QR login repairs.
Verification:
- tsc --noEmit clean
- vitest: 93/93 pass
- Live: POST /api/auth/qrcode platform=qq returns both tokens packed
into `key`; polling a freshly-issued QR returns {"status":"waiting"}
instead of the old {"status":"expired"}; raw library response is
{"isOk":false,"refresh":false,"message":"未扫描二维码"} as expected.
- Regression: netease and bilibili QR flows still produce non-empty
qrUrl/key — no collateral damage to the other providers.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Version 0.2.1 ships a universal clientinit format that works natively
against both TS3 and TS6 servers:
client_version: "3.?.? [Build: 5680278000]"
client_version_sign: DX5NIYLvfJEUjuIbCidnoeozxIDRRkpq3I9vVMBmE9L2qnekOo
BzSenkzsg2lC9CMv8K5hkEzhr2TYUYSwUXCg==
The old ts6-compat.ts workaround (monkey-patching handler.sendPacket to
rewrite clientinit's client_version to "3.6.2") is now actively wrong:
it replaces the library's new correct version/signature pair with a
stale one that TS6 servers reject, which is why the first 0.2.1 TS6
handshake attempt still hung at `received initivexpand2`.
Changes:
- package.json: "@honeybbq/teamspeak-client": "^0.1.0" -> "^0.2.1"
- src/ts-protocol/client.ts: remove patchClientInitVersion import and
the sendPacket monkey-patch block. Leave an inline comment so anyone
reading the git blame understands why the shim is gone.
- Delete src/ts-protocol/ts6-compat.ts and ts6-compat.test.ts — no
callers remain.
Other 0.2.x notes worth knowing (no code change here, just documenting):
- ClientOptions gained serverPassword / defaultChannel /
defaultChannelPassword that are sent DURING clientinit. We still call
our own joinChannel() post-connect because the existing flow works
and switching is an orthogonal refactor.
- 0.1.1 contains the P-256 DER encoding fix (PR #5 by ZHANGTIANYAO1).
Identities generated by 0.1.0 are cryptographically incompatible with
0.2.x's corrected handshake path — a bot whose identity column was
populated before this upgrade will hang at `received initivexpand2`
and fall through the 15s connect deadline. Workaround: clear the
identity column so the next start generates a fresh key. Server
groups assigned to the old UID must be re-granted once against the
new one.
Verification:
- tsc --noEmit clean
- vitest: 93/93 unit tests pass (1 test file removed with ts6-compat)
- scripts/test_full_feature.py against a local TS6 server: 51/51 pass,
including handshake, voice playback, identity persistence, WebSocket
stateChange broadcasts, and all corner-case regressions.
- Live: bot connected to TS6 in ~80ms after identity regeneration,
played NetEase audio through the voice channel, clean stop.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Major bug fixes and corner-case hardening across the backend, plus a
comprehensive feature test suite. All 94 unit tests + 51 integration
tests pass against a local TS3 server.
Lifecycle & state consistency
-----------------------------
- Bug A: startBot() now wraps connect() in a 15s deadline. A hung TS
handshake no longer blocks the /start HTTP call forever; the failing
instance is torn down and the caller gets a clean 500.
- Bug B: executeCommand rejects audio-dispatching commands (play, add,
next, skip, prev, playlist, album, fm) when the bot is disconnected.
Config-only commands (vol, mode, clear, stop, queue, now, lyrics)
still work so the UI stays usable while offline.
- Bug C: the tsClient 'disconnected' handler always clears player state
now, even when connect() never completed. A separate disconnectEmitted
flag guards duplicate external event emission. Previously an orphaned
connect attempt that idle-timed-out would leave playing=true forever.
- resolveAndPlay re-checks this.connected AFTER the URL-resolve await so
a stop() during the network call can't spawn ffmpeg on a disconnected
bot.
- connect() throws if disconnect() fired during the handshake await,
preventing a concurrent stop from being overwritten by a late connected
flag flip.
- startBot always disconnects the outgoing BotInstance before creating
a replacement, covering the mid-handshake case where isConnected()
still returned false but the library client was live.
- startBot now reuses the stored identity so server groups granted to
the bot survive restarts (was regenerating a fresh UID each time).
WebSocket reliability
---------------------
- BotManager extends EventEmitter and emits 'botInstance' whenever a
new instance is created. websocket.ts listens and re-attaches its
stateChange / connected / disconnected listeners immediately, fixing
the bug where player-bar UI never updated until manual refresh.
- attachedBots map now stores the BotInstance reference and detaches
stale listeners when the instance is replaced. Safety-net interval
(5s) also reconciles to catch anything missed.
- removeBot emits 'botInstanceRemoved' -> WS broadcasts a new
{type:"botRemoved", botId} message. Client drops the bot from its
local store instead of showing it as permanently offline.
HTTP input validation
---------------------
- /volume rejects non-number, NaN, Infinity, and out-of-range values
with a proper 400 instead of a 200 OK wrapping a usage-text string.
- /mode rejects anything not in {seq, loop, random, rloop} with 400.
- /seek rejects NaN / Infinity / negative (previously NaN slipped
through typeof==="number" and poisoned seekOffset).
- /play-at validates index < queue.size() BEFORE stopping current
playback (was silently killing the current song on invalid input).
- /play, /add, /playlist, /play-by-id, /add-by-id, /play-playlist
all honour platform=youtube now (previously fell through to netease
and silently played the wrong platform).
YouTube made truly optional
---------------------------
- Lazy checkYtDlpAvailable() runs `yt-dlp --version` once, caches only
positive results so users can install yt-dlp mid-run and have it
picked up without a restart.
- getAuthStatus() returns loggedIn=false with nickname
"YouTube (yt-dlp not installed)" when the binary is missing. UI can
grey out YouTube instead of silently returning empty searches.
- findYtDlp() picks .exe on win32 and bare binary elsewhere.
- /auth/status?platform=youtube now routes to the YouTube provider
instead of falling through to NetEase and leaking the NetEase
user's nickname + avatar.
- /auth/cookie rejects platform=youtube with 400 instead of clobbering
the NetEase cookie entry.
- README documents yt-dlp install paths (bin/ local vs PATH) and adds
a dedicated "Optional: YouTube source" section.
Bot Selector UI
---------------
- New power button in each row of the dropdown with play-state-aware
styling: disabled + wait-cursor during API call, green highlight when
connected, greys out when the bot is offline.
- Dropdown always visible when >=1 bot exists, bigger font + padding.
Queue correctness
-----------------
- PlayQueue.remove(current) now decrements currentIndex so next() in
sequential mode advances to the shifted song. Previously removing
the currently-playing track silently skipped the next track because
current() falsely reported it as active and next() then incremented
past it.
Vote-skip hardening
-------------------
- cmdVote: needed threshold is Math.max(1, ceil(users/2)) so a single
voter in an empty channel can't unanimously pass a vote with
needed=0.
- resolveAndPlay clears voteSkipUsers on every new track load so votes
can't leak across songs via cmdPlay/cmdPlaylist/cmdAlbum/cmdFm paths.
cmdAdd parity
-------------
- cmdAdd auto-plays the newly-added song if the player was idle,
matching /api/player/:id/add-by-id behaviour. Previously add'ing to
an empty queue on a connected+idle bot silently enqueued without
starting playback.
Test suite
----------
- scripts/test_full_feature.py — 51 tests across 10 groups exercising
every HTTP endpoint, WebSocket broadcasts, all music providers, bot
lifecycle, disconnected-state corners, seek validation, input
validation, and the main race conditions. Captures and restores the
target bot's initial state. Resilient to TS3 anti-flood via retry
with exponential backoff. Runs against a real local TS3 server.
- scripts/test_rapid_cycle.py — Bugs A/B/C regressions
- scripts/test_corner_cases.py — disconnect-during-connect race, config
commands while disconnected, etc.
- scripts/test_more_corners.py — resolveAndPlay race, seek NaN
- scripts/test_power_button.py — E2E for the new power button
- scripts/test_bot_remove.py — E2E for WS botRemoved broadcast
- scripts/test_playbar.py — player bar auto-show regression (updated
to restore bot state on exit)
- scripts/test_multibot.py — two-bot concurrent playback monitor
- src/audio/queue.test.ts — 4 new vitest cases for remove() edge cases
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
startBot creates a fresh BotInstance but the WS layer keyed its listener
map by bot.id, so ensureAllBotsAttached skipped the new object and
stateChange events were never broadcast — the player bar only appeared
after a manual refresh. BotManager now extends EventEmitter and emits
"botInstance" whenever a new bot object is created; websocket.ts stores
the bot reference, detaches on replacement, and subscribes to the event
for immediate wiring.
Also enlarges the bot selector (padding 10×20, font 16, min-height 44,
bigger dot/chevron/state icons, wider name) and adds Playwright repro
scripts for the player bar bug and navbar sizing check.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
