fix: harden bot lifecycle, validate HTTP inputs, make YouTube truly optional

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>
2026-10-01 20:42:50 +08:00 · 2026-04-11 01:35:14 +08:00
1 parent 6e828b9c2d
commit 4643f70f4a
@@ -0,0 +1,99 @@
+"""Regression: DELETE /api/bot/:id must broadcast botRemoved so the UI drops the row.
+
+Creates an ephemeral bot, opens the dropdown, deletes the bot via API, and
+asserts the row disappears without any page reload. Does not touch any
+existing user bot.
+"""
+import time
+import requests
+from playwright.sync_api import sync_playwright
+
+BASE = "http://localhost:3000"
+EPHEMERAL_NAME = "rmbot_test"
+EPHEMERAL_NICK = "RmBotTest"
+
+
+def api(path, method="GET", **kw):
+    r = getattr(requests, method.lower())(f"{BASE}{path}", timeout=10, **kw)
+    r.raise_for_status()
+    return r.json() if r.text else None
+
+
+def cleanup():
+    for b in api("/api/bot/")["bots"]:
+        if b["name"] == EPHEMERAL_NAME:
+            try:
+                api(f"/api/bot/{b['id']}", method="DELETE")
+            except Exception:
+                pass
+
+
+def main():
+    cleanup()
+    new = api(
+        "/api/bot/",
+        method="POST",
+        json={
+            "name": EPHEMERAL_NAME,
+            "serverAddress": "127.0.0.1",
+            "serverPort": 9987,
+            "nickname": EPHEMERAL_NICK,
+            "autoStart": False,
+        },
+    )
+    bot_id = new["id"]
+    print(f"[setup] created ephemeral bot {bot_id[:8]}")
+
+    try:
+        with sync_playwright() as p:
+            browser = p.chromium.launch(headless=True)
+            try:
+                page = browser.new_page(viewport={"width": 1440, "height": 900})
+                page.goto(BASE)
+                page.wait_for_load_state("networkidle")
+                time.sleep(0.8)
+
+                # Open dropdown and confirm the new bot row is present
+                page.locator(".bot-selector-btn").click()
+                page.wait_for_selector(".bot-dropdown")
+                rows_before = page.locator(".bot-dropdown-row").count()
+                print(f"[ui] dropdown rows before remove: {rows_before}")
+
+                # Match the ephemeral row by its name text
+                present = (
+                    page.locator(".bot-dropdown-row", has_text=EPHEMERAL_NAME).count()
+                )
+                assert present == 1, f"ephemeral row not found (got {present})"
+
+                # Delete via API
+                api(f"/api/bot/{bot_id}", method="DELETE")
+                print("[api] deleted bot")
+
+                # Wait up to 4s for UI to drop the row
+                removed = False
+                for _ in range(40):
+                    if (
+                        page.locator(
+                            ".bot-dropdown-row", has_text=EPHEMERAL_NAME
+                        ).count()
+                        == 0
+                    ):
+                        removed = True
+                        break
+                    time.sleep(0.1)
+
+                rows_after = page.locator(".bot-dropdown-row").count()
+                print(f"[ui] dropdown rows after remove: {rows_after}")
+                assert removed, "ephemeral row did not disappear from UI after DELETE"
+                assert rows_after == rows_before - 1, (
+                    f"row count mismatch: before={rows_before} after={rows_after}"
+                )
+                print("[PASS] bot removal propagates to UI via WS")
+            finally:
+                browser.close()
+    finally:
+        cleanup()
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,171 @@
+"""Corner case regressions that go beyond Bugs A/B/C.
+
+A. Race — disconnect() called during connect()'s awaited handshake must
+   NOT leave the bot reporting connected=true afterwards.
+
+B. Config-only commands (vol, mode, clear) must work even when the bot is
+   disconnected (UI should stay usable while the bot is offline).
+
+C. After a disconnect mid-playback, the player must not be able to
+   auto-advance to the next queued song (trackEnd → resolveAndPlay).
+"""
+import time
+import threading
+import requests
+
+BASE = "http://localhost:3000"
+
+
+def api(path, method="GET", **kw):
+    return getattr(requests, method.lower())(f"{BASE}{path}", timeout=30, **kw)
+
+
+def get_bot(bot_id):
+    return next(b for b in api("/api/bot/").json()["bots"] if b["id"] == bot_id)
+
+
+def wait_connected(bot_id, want, timeout=15):
+    end = time.time() + timeout
+    while time.time() < end:
+        if get_bot(bot_id)["connected"] is want:
+            return True
+        time.sleep(0.1)
+    return False
+
+
+def test_config_commands_when_disconnected(bot_id):
+    """B. vol/mode/clear should succeed while bot is disconnected."""
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    wait_connected(bot_id, False)
+
+    # volume is the simplest config-only command
+    r = api(
+        f"/api/player/{bot_id}/volume",
+        method="POST",
+        json={"volume": 60},
+    )
+    assert r.status_code == 200, f"volume failed while disconnected: {r.status_code} {r.text[:100]}"
+
+    r = api(
+        f"/api/player/{bot_id}/mode",
+        method="POST",
+        json={"mode": "seq"},
+    )
+    assert r.status_code == 200, f"mode failed while disconnected: {r.status_code} {r.text[:100]}"
+
+    r = api(f"/api/player/{bot_id}/clear", method="POST")
+    assert r.status_code == 200, f"clear failed while disconnected: {r.status_code} {r.text[:100]}"
+
+    print("[PASS] config commands (vol/mode/clear) work when disconnected")
+
+
+def test_play_rejected_when_disconnected(bot_id):
+    """B (negative). play/add/next/prev should still be rejected."""
+    r = api(
+        f"/api/player/{bot_id}/play",
+        method="POST",
+        json={"query": "test", "platform": "netease"},
+    )
+    assert r.status_code >= 400, f"play should fail while disconnected"
+
+    r = api(
+        f"/api/player/{bot_id}/add",
+        method="POST",
+        json={"query": "test", "platform": "netease"},
+    )
+    assert r.status_code >= 400, f"add should fail while disconnected"
+
+    r = api(f"/api/player/{bot_id}/next", method="POST")
+    assert r.status_code >= 400, f"next should fail while disconnected"
+
+    print("[PASS] audio commands (play/add/next) rejected when disconnected")
+
+
+def test_disconnect_during_connect_race(bot_id):
+    """A. disconnect() called while connect() is awaiting must win the race.
+
+    Fires a stop 200ms into a start call; after things settle the bot's
+    connected state must be stable (either cleanly disconnected, or cleanly
+    connected if the stop happened after connect completed). It must NOT
+    end up in a weird state where connected=true but a subsequent query
+    shows inconsistent data.
+    """
+    # Ensure disconnected first
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    wait_connected(bot_id, False)
+    time.sleep(1)  # give TS server a moment to forget us
+
+    def delayed_stop():
+        time.sleep(0.2)
+        try:
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+        except Exception:
+            pass
+
+    threading.Thread(target=delayed_stop, daemon=True).start()
+    try:
+        r = api(f"/api/bot/{bot_id}/start", method="POST")
+    except Exception as e:
+        r = None
+        print(f"[info] start threw: {e}")
+
+    # Wait for all state transitions to settle
+    time.sleep(2)
+
+    b = get_bot(bot_id)
+    # The key invariant: if connected is false, playing must also be false;
+    # if connected is true, the transport is actually up (we can issue
+    # another command without error).
+    assert not (b["connected"] is False and b["playing"] is True), (
+        f"inconsistent state: connected={b['connected']} playing={b['playing']}"
+    )
+    print(
+        f"[PASS] disconnect-during-connect race — final state consistent "
+        f"(connected={b['connected']} playing={b['playing']})"
+    )
+
+
+def test_resolve_guard(bot_id):
+    """C. resolveAndPlay on a disconnected bot is a no-op.
+
+    We can't directly invoke resolveAndPlay from the API, but we can
+    verify by checking that after a stop, the bot stays idle even if we
+    wait for a trackEnd-like event to fire.
+    """
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    wait_connected(bot_id, False)
+    time.sleep(1.5)  # more than a frame cycle
+    b = get_bot(bot_id)
+    assert not b["playing"], (
+        f"player should stay stopped after disconnect: {b}"
+    )
+    print("[PASS] player stays idle after disconnect (no ghost autoplay)")
+
+
+def main():
+    bots = api("/api/bot/").json()["bots"]
+    if not bots:
+        print("[skip] no bots")
+        return
+    bot_id = bots[0]["id"]
+    initial = bots[0]["connected"]
+    print(f"[init] bot={bot_id[:8]} initial connected={initial}")
+
+    try:
+        test_config_commands_when_disconnected(bot_id)
+        test_play_rejected_when_disconnected(bot_id)
+        test_resolve_guard(bot_id)
+        test_disconnect_during_connect_race(bot_id)
+        print("ALL GREEN")
+    finally:
+        if initial:
+            api(f"/api/bot/{bot_id}/start", method="POST")
+            wait_connected(bot_id, True)
+        else:
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+            wait_connected(bot_id, False)
+        print(f"[restore] connected={get_bot(bot_id)['connected']}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,852 @@
+"""Comprehensive feature + corner-case test for TSMusicBot against a
+local TeamSpeak 3 server.
+
+Exercises every major HTTP endpoint, the WebSocket state-broadcast path,
+all music providers, bot lifecycle transitions, and a handful of races
+that have burned us in the past. Designed to be safe to run against a
+real installation: captures the target bot's initial connected/volume/
+mode settings and restores them in `finally`.
+
+Usage:
+    "C:/Users/saopig1/miniforge3/python.exe" scripts/test_full_feature.py
+
+Exit code: 0 if every non-skipped test passed, 1 otherwise.
+"""
+from __future__ import annotations
+
+import json
+import threading
+import time
+from dataclasses import dataclass
+from typing import Any
+
+import requests
+
+BASE = "http://localhost:3000"
+POLL_INTERVAL = 0.15
+CONNECT_TIMEOUT = 20        # tolerate occasional TS3 anti-flood grace
+ANTIFLOOD_BREATHER = 1.5    # gap between rapid cycles so TS3 stays happy
+
+
+# ----------------------------- HTTP helpers ---------------------------------
+
+
+def api(path: str, method: str = "GET", json_body: Any = None):
+    """Return (status_code, body). Never raises."""
+    try:
+        fn = getattr(requests, method.lower())
+        r = fn(f"{BASE}{path}", json=json_body, timeout=30)
+        try:
+            return r.status_code, r.json()
+        except Exception:
+            return r.status_code, r.text
+    except Exception as e:
+        return None, f"<{type(e).__name__}: {e}>"
+
+
+def get_bot(bot_id: str) -> dict | None:
+    _, data = api("/api/bot/")
+    if not isinstance(data, dict):
+        return None
+    return next((b for b in data.get("bots", []) if b["id"] == bot_id), None)
+
+
+def wait_connected(bot_id: str, want: bool, timeout: float = CONNECT_TIMEOUT) -> bool:
+    end = time.time() + timeout
+    while time.time() < end:
+        b = get_bot(bot_id)
+        if b is not None and b["connected"] is want:
+            return True
+        time.sleep(POLL_INTERVAL)
+    return False
+
+
+def start_and_wait(bot_id: str, retries: int = 2) -> bool:
+    """Start the bot, tolerating transient TS3 anti-flood by retrying with
+    exponential backoff. Returns True only when the bot reports connected."""
+    for attempt in range(retries + 1):
+        s, _ = api(f"/api/bot/{bot_id}/start", method="POST")
+        if s == 200 and wait_connected(bot_id, True):
+            return True
+        # If /start returned an error (e.g. connect timeout from our 15s
+        # deadline), back off and retry — TS3 server-side anti-flood
+        # usually clears in a few seconds.
+        if attempt < retries:
+            time.sleep(3.0 * (attempt + 1))
+            # Make sure we're fully stopped before the next attempt so
+            # oldBot.disconnect() doesn't double-fire
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+            wait_connected(bot_id, False, timeout=5)
+    return False
+
+
+def stop_and_wait(bot_id: str) -> bool:
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    return wait_connected(bot_id, False)
+
+
+def assert_started(bot_id: str):
+    """Helper that raises with a clear message when start fails so test
+    output points at 'could not connect' rather than an empty assertion."""
+    if not start_and_wait(bot_id):
+        raise AssertionError(
+            "could not bring bot online (TS3 server may be anti-flooding "
+            "or unreachable)"
+        )
+
+
+# ----------------------------- test runner ----------------------------------
+
+
+@dataclass
+class TestResult:
+    name: str
+    status: str  # PASS / FAIL / ERROR / SKIP
+    detail: str = ""
+
+
+passed = 0
+failed = 0
+skipped = 0
+results: list[TestResult] = []
+
+
+def run(name: str, fn):
+    global passed, failed
+    try:
+        fn()
+        print(f"  [PASS] {name}")
+        passed += 1
+        results.append(TestResult(name, "PASS"))
+    except AssertionError as e:
+        print(f"  [FAIL] {name}: {e}")
+        failed += 1
+        results.append(TestResult(name, "FAIL", str(e)))
+    except Exception as e:
+        print(f"  [ERROR] {name}: {type(e).__name__}: {e}")
+        failed += 1
+        results.append(TestResult(name, "ERROR", f"{type(e).__name__}: {e}"))
+
+
+def skip(name: str, reason: str):
+    global skipped
+    print(f"  [SKIP] {name}  ({reason})")
+    skipped += 1
+    results.append(TestResult(name, "SKIP", reason))
+
+
+# ----------------------------- test groups ----------------------------------
+
+
+def group_infrastructure(bot_id: str):
+    print("\n== infrastructure ==")
+
+    def t_health():
+        s, d = api("/api/health")
+        assert s == 200, f"health returned {s}"
+        assert d.get("status") == "ok"
+    run("GET /api/health", t_health)
+
+    def t_list_bots():
+        s, d = api("/api/bot/")
+        assert s == 200
+        assert isinstance(d.get("bots"), list)
+        assert any(b["id"] == bot_id for b in d["bots"])
+    run("GET /api/bot/ lists target bot", t_list_bots)
+
+    def t_bot_config():
+        s, d = api(f"/api/bot/{bot_id}/config")
+        assert s == 200
+        assert d["id"] == bot_id
+        assert "identity" in d
+        assert "serverAddress" in d and "nickname" in d
+        assert "serverPassword" in d, "serverPassword field missing"
+    run("GET /api/bot/:id/config returns all fields", t_bot_config)
+
+    def t_404_on_unknown_bot():
+        s, _ = api("/api/bot/does-not-exist/config")
+        assert s == 404
+    run("404 on unknown bot id", t_404_on_unknown_bot)
+
+    def t_quality_shape():
+        s, d = api("/api/music/quality")
+        assert s == 200
+        for p in ("netease", "qq", "bilibili"):
+            assert p in d, f"{p} missing from quality response"
+    run("GET /api/music/quality shape", t_quality_shape)
+
+
+def group_auth_status():
+    print("\n== auth status per platform ==")
+
+    def t_netease_ok():
+        s, d = api("/api/auth/status?platform=netease")
+        assert s == 200
+        assert d.get("platform") == "netease"
+        assert "loggedIn" in d
+    run("auth status netease", t_netease_ok)
+
+    def t_qq_ok():
+        s, d = api("/api/auth/status?platform=qq")
+        assert s == 200
+        assert d.get("platform") == "qq"
+    run("auth status qq", t_qq_ok)
+
+    def t_bilibili_ok():
+        s, d = api("/api/auth/status?platform=bilibili")
+        assert s == 200
+        assert d.get("platform") == "bilibili"
+    run("auth status bilibili", t_bilibili_ok)
+
+    def t_youtube_routed():
+        # Regression: /auth/status?platform=youtube used to fall through
+        # to NetEase and leak the NetEase user's nickname/avatar.
+        s, d = api("/api/auth/status?platform=youtube")
+        assert s == 200
+        assert d.get("platform") == "youtube", (
+            f"youtube auth status leaked to {d.get('platform')}"
+        )
+    run("auth status youtube routes correctly", t_youtube_routed)
+
+    def t_youtube_cookie_rejected():
+        s, d = api(
+            "/api/auth/cookie",
+            method="POST",
+            json_body={"platform": "youtube", "cookie": "fake"},
+        )
+        assert s == 400, f"youtube cookie should be rejected, got {s}: {d}"
+    run("POST /auth/cookie rejects youtube", t_youtube_cookie_rejected)
+
+
+def group_search():
+    print("\n== multi-platform search ==")
+
+    def search(platform: str, query: str = "test"):
+        return api(f"/api/music/search?q={query}&platform={platform}&limit=1")
+
+    def t_netease():
+        s, d = search("netease")
+        assert s == 200
+        assert isinstance(d.get("songs"), list)
+    run("netease search", t_netease)
+
+    def t_qq():
+        s, d = search("qq")
+        assert s == 200
+        # QQ may return 0 results if no cookie, but shouldn't error
+        assert isinstance(d.get("songs"), list)
+    run("qq search (empty ok)", t_qq)
+
+    def t_bilibili():
+        s, d = search("bilibili")
+        assert s == 200
+        assert isinstance(d.get("songs"), list)
+    run("bilibili search", t_bilibili)
+
+    def t_missing_query():
+        s, _ = api("/api/music/search?platform=netease")
+        assert s == 400, "missing q should 400"
+    run("400 on missing query", t_missing_query)
+
+    _, auth = api("/api/auth/status?platform=youtube")
+    youtube_available = isinstance(auth, dict) and auth.get("loggedIn") is True
+
+    if youtube_available:
+        def t_youtube():
+            s, d = search("youtube", "lofi")
+            assert s == 200
+            songs = d.get("songs", [])
+            assert len(songs) >= 1, "expected at least 1 YouTube result"
+            assert songs[0]["platform"] == "youtube"
+        run("youtube search (yt-dlp installed)", t_youtube)
+    else:
+        skip("youtube search", "yt-dlp not installed")
+
+
+def group_lifecycle(bot_id: str):
+    print("\n== connection lifecycle ==")
+
+    def t_stop_from_any_state():
+        api(f"/api/bot/{bot_id}/stop", method="POST")
+        assert wait_connected(bot_id, False), "bot did not stop"
+        b = get_bot(bot_id)
+        assert not b["playing"], "playing should be false after stop"
+    run("stop from any state \u2192 disconnected+idle", t_stop_from_any_state)
+
+    def t_start_completes_quickly():
+        t0 = time.time()
+        s, d = api(f"/api/bot/{bot_id}/start", method="POST")
+        elapsed = time.time() - t0
+        assert s == 200, f"start failed: {d}"
+        assert elapsed < 10, f"start took {elapsed:.1f}s (expected <10s)"
+        assert wait_connected(bot_id, True)
+    run("start completes well under 15s deadline", t_start_completes_quickly)
+
+    def t_identity_persists():
+        assert_started(bot_id)
+        _, cfg1 = api(f"/api/bot/{bot_id}/config")
+        id1 = cfg1["identity"]
+        assert id1, "identity empty after first start"
+        assert stop_and_wait(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+        assert_started(bot_id)
+        _, cfg2 = api(f"/api/bot/{bot_id}/config")
+        assert cfg2["identity"] == id1, (
+            f"identity changed across restart: {id1} \u2192 {cfg2['identity']}"
+        )
+    run("identity preserved across stop/start", t_identity_persists)
+
+
+def group_playback(bot_id: str):
+    print("\n== playback ==")
+
+    # Bring the bot online ONCE for the whole playback group, then only
+    # toggle player state (play/pause/stop) between tests. This keeps the
+    # TS3 reconnect count for this group at exactly 1.
+    assert_started(bot_id)
+
+    def t_play_song():
+        s, d = api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        assert s == 200, f"play failed: {d}"
+        time.sleep(1.2)
+        b = get_bot(bot_id)
+        assert b["playing"] is True, f"not playing after /play: {b}"
+        assert b["currentSong"] is not None
+    run("play netease song \u2192 playing=true", t_play_song)
+
+    def t_pause_resume():
+        # Previous test left a song playing
+        api(f"/api/player/{bot_id}/pause", method="POST")
+        time.sleep(0.4)
+        b = get_bot(bot_id)
+        assert b["paused"] is True, f"pause failed: {b}"
+        api(f"/api/player/{bot_id}/resume", method="POST")
+        time.sleep(0.4)
+        b = get_bot(bot_id)
+        assert b["paused"] is False and b["playing"] is True, f"resume failed: {b}"
+    run("pause \u2192 paused, resume \u2192 playing", t_pause_resume)
+
+    def t_volume_change():
+        s, _ = api(
+            f"/api/player/{bot_id}/volume",
+            method="POST",
+            json_body={"volume": 42},
+        )
+        assert s == 200
+        time.sleep(0.2)
+        b = get_bot(bot_id)
+        assert b["volume"] == 42, f"volume not applied: {b['volume']}"
+    run("volume change", t_volume_change)
+
+    def t_mode_cycle():
+        for m in ("seq", "loop", "random", "rloop"):
+            s, _ = api(
+                f"/api/player/{bot_id}/mode", method="POST", json_body={"mode": m}
+            )
+            assert s == 200
+            b = get_bot(bot_id)
+            assert b["playMode"] == m, f"mode {m} not applied: {b['playMode']}"
+    run("all four play modes apply", t_mode_cycle)
+
+    def t_queue_endpoint():
+        s, d = api(f"/api/player/{bot_id}/queue")
+        assert s == 200
+        assert isinstance(d.get("queue"), list)
+        assert "status" in d
+    run("GET /player/:id/queue returns queue+status", t_queue_endpoint)
+
+    def t_elapsed_endpoint():
+        s, d = api(f"/api/player/{bot_id}/elapsed")
+        assert s == 200
+        elapsed = d.get("elapsed")
+        assert isinstance(elapsed, (int, float)) and elapsed >= 0, (
+            f"elapsed should be non-negative number: {elapsed}"
+        )
+    run("GET /player/:id/elapsed returns finite number", t_elapsed_endpoint)
+
+    def t_add_autoplay_on_idle():
+        # This specific test needs an IDLE bot — stop first (but keep
+        # connected), then add and confirm auto-play.
+        api(f"/api/player/{bot_id}/stop", method="POST")
+        time.sleep(0.4)
+        b = get_bot(bot_id)
+        assert not b["playing"] and b["queueSize"] == 0, f"setup failed: {b}"
+        s, d = api(
+            f"/api/player/{bot_id}/add",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        assert s == 200
+        msg = d.get("message", "") if isinstance(d, dict) else ""
+        assert "Now playing" in msg, (
+            f"add on idle bot should auto-play, got: {msg!r}"
+        )
+        time.sleep(0.8)
+        b = get_bot(bot_id)
+        assert b["playing"] is True, f"not playing after add: {b}"
+    run("add on idle bot auto-plays", t_add_autoplay_on_idle)
+
+    # Leave the bot in a clean state for the next group
+    api(f"/api/player/{bot_id}/stop", method="POST")
+
+
+def group_queue_ops(bot_id: str):
+    print("\n== queue operations ==")
+    assert_started(bot_id)
+
+    def t_clear():
+        api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        time.sleep(0.8)
+        api(
+            f"/api/player/{bot_id}/add",
+            method="POST",
+            json_body={"query": "lemon tree", "platform": "netease"},
+        )
+        time.sleep(0.6)
+        b_before = get_bot(bot_id)
+        assert b_before["queueSize"] >= 2, f"expected \u22652 songs: {b_before}"
+        api(f"/api/player/{bot_id}/clear", method="POST")
+        time.sleep(0.4)
+        b_after = get_bot(bot_id)
+        assert b_after["queueSize"] == 0
+    run("clear queue empties it", t_clear)
+
+    def t_play_at_invalid_preserves_playback():
+        api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        time.sleep(1.2)
+        assert get_bot(bot_id)["playing"]
+        s, _ = api(
+            f"/api/player/{bot_id}/play-at",
+            method="POST",
+            json_body={"index": 9999},
+        )
+        assert s == 400, f"invalid index should 400, got {s}"
+        time.sleep(0.4)
+        b = get_bot(bot_id)
+        assert b["playing"], "invalid play-at killed the current song"
+    run("invalid play-at preserves current playback", t_play_at_invalid_preserves_playback)
+
+    def t_play_at_negative_rejected():
+        s, _ = api(
+            f"/api/player/{bot_id}/play-at",
+            method="POST",
+            json_body={"index": -1},
+        )
+        assert s == 400
+    run("play-at with negative index rejected", t_play_at_negative_rejected)
+
+    api(f"/api/player/{bot_id}/stop", method="POST")
+
+
+def group_input_validation(bot_id: str):
+    print("\n== HTTP input validation ==")
+
+    def t_volume_out_of_range():
+        for bad in (150, -10, 1000, -1):
+            s, _ = api(
+                f"/api/player/{bot_id}/volume",
+                method="POST",
+                json_body={"volume": bad},
+            )
+            assert s == 400, f"volume={bad} should 400, got {s}"
+    run("volume out-of-range rejected (400)", t_volume_out_of_range)
+
+    def t_volume_wrong_type():
+        for bad in ("50", None, [50], {"v": 50}):
+            s, _ = api(
+                f"/api/player/{bot_id}/volume",
+                method="POST",
+                json_body={"volume": bad},
+            )
+            assert s == 400, f"volume={bad!r} should 400, got {s}"
+    run("volume wrong-type rejected (400)", t_volume_wrong_type)
+
+    def t_volume_missing():
+        s, _ = api(
+            f"/api/player/{bot_id}/volume", method="POST", json_body={}
+        )
+        assert s == 400
+    run("volume missing rejected (400)", t_volume_missing)
+
+    def t_volume_valid():
+        for good in (0, 1, 50, 100):
+            s, _ = api(
+                f"/api/player/{bot_id}/volume",
+                method="POST",
+                json_body={"volume": good},
+            )
+            assert s == 200, f"volume={good} should succeed, got {s}"
+            b = get_bot(bot_id)
+            assert b["volume"] == good, f"volume not applied: {b['volume']}"
+    run("valid volumes apply", t_volume_valid)
+
+    def t_mode_invalid():
+        for bad in ("bogus", "", None, 1, "SEQ"):
+            s, _ = api(
+                f"/api/player/{bot_id}/mode",
+                method="POST",
+                json_body={"mode": bad},
+            )
+            assert s == 400, f"mode={bad!r} should 400, got {s}"
+    run("mode invalid rejected (400)", t_mode_invalid)
+
+    def t_mode_missing():
+        s, _ = api(f"/api/player/{bot_id}/mode", method="POST", json_body={})
+        assert s == 400
+    run("mode missing rejected (400)", t_mode_missing)
+
+
+def group_disconnect_corners(bot_id: str):
+    print("\n== disconnected-bot corners ==")
+
+    def t_play_rejected():
+        assert stop_and_wait(bot_id)
+        s, d = api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "x", "platform": "netease"},
+        )
+        assert s >= 400
+        err = (d.get("error") or "") if isinstance(d, dict) else ""
+        assert "not connected" in err.lower(), f"expected 'not connected' error: {d}"
+    run("play rejected while disconnected", t_play_rejected)
+
+    def t_add_rejected():
+        s, _ = api(
+            f"/api/player/{bot_id}/add",
+            method="POST",
+            json_body={"query": "x", "platform": "netease"},
+        )
+        assert s >= 400
+    run("add rejected while disconnected", t_add_rejected)
+
+    def t_next_rejected():
+        s, _ = api(f"/api/player/{bot_id}/next", method="POST")
+        assert s >= 400
+    run("next rejected while disconnected", t_next_rejected)
+
+    def t_volume_allowed():
+        s, _ = api(
+            f"/api/player/{bot_id}/volume",
+            method="POST",
+            json_body={"volume": 60},
+        )
+        assert s == 200, "volume should work while disconnected"
+    run("volume allowed while disconnected", t_volume_allowed)
+
+    def t_mode_allowed():
+        s, _ = api(
+            f"/api/player/{bot_id}/mode", method="POST", json_body={"mode": "random"}
+        )
+        assert s == 200, "mode should work while disconnected"
+    run("mode allowed while disconnected", t_mode_allowed)
+
+    def t_clear_allowed():
+        s, _ = api(f"/api/player/{bot_id}/clear", method="POST")
+        assert s == 200, "clear should work while disconnected"
+    run("clear allowed while disconnected", t_clear_allowed)
+
+    def t_player_state_clean():
+        b = get_bot(bot_id)
+        assert not b["playing"] and not b["paused"], f"state leak: {b}"
+    run("no player state leak while disconnected", t_player_state_clean)
+
+
+def group_seek(bot_id: str):
+    print("\n== seek validation ==")
+
+    def t_negative():
+        s, _ = api(
+            f"/api/player/{bot_id}/seek", method="POST", json_body={"position": -5}
+        )
+        assert s == 400
+    run("negative seek rejected", t_negative)
+
+    def t_string():
+        s, _ = api(
+            f"/api/player/{bot_id}/seek",
+            method="POST",
+            json_body={"position": "abc"},
+        )
+        assert s == 400
+    run("string seek rejected", t_string)
+
+    def t_nan_literal():
+        r = requests.post(
+            f"{BASE}/api/player/{bot_id}/seek",
+            data='{"position": NaN}',
+            headers={"Content-Type": "application/json"},
+            timeout=10,
+        )
+        assert r.status_code >= 400, f"NaN literal accepted: {r.status_code}"
+    run("NaN literal seek rejected", t_nan_literal)
+
+    def t_valid_seek():
+        # Seek needs a live connection + playing song. The disconnect-
+        # corners group right before this one left the bot disconnected.
+        assert_started(bot_id)
+        api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        time.sleep(1.5)
+        s, _ = api(
+            f"/api/player/{bot_id}/seek",
+            method="POST",
+            json_body={"position": 25},
+        )
+        assert s == 200
+        time.sleep(0.5)
+        _, d = api(f"/api/player/{bot_id}/elapsed")
+        elapsed = d.get("elapsed")
+        assert isinstance(elapsed, (int, float)) and 24 <= elapsed < 40, (
+            f"elapsed after seek(25) wrong: {elapsed}"
+        )
+        api(f"/api/player/{bot_id}/stop", method="POST")
+    run("valid seek produces finite elapsed", t_valid_seek)
+
+
+def group_races(bot_id: str):
+    print("\n== race conditions ==")
+
+    def t_disconnect_during_connect():
+        assert stop_and_wait(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+
+        def delayed_stop():
+            time.sleep(0.2)
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+
+        threading.Thread(target=delayed_stop, daemon=True).start()
+        api(f"/api/bot/{bot_id}/start", method="POST")
+        time.sleep(2)
+        b = get_bot(bot_id)
+        assert not (b["connected"] is False and b["playing"] is True), (
+            f"inconsistent state: {b}"
+        )
+    run("disconnect during connect", t_disconnect_during_connect)
+
+    def t_stop_during_url_resolve():
+        assert stop_and_wait(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+        assert_started(bot_id)
+
+        def delayed_stop():
+            time.sleep(0.15)
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+
+        threading.Thread(target=delayed_stop, daemon=True).start()
+        api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json_body={"query": "the mass", "platform": "netease"},
+        )
+        time.sleep(3)
+        b = get_bot(bot_id)
+        assert not (b["connected"] is False and b["playing"] is True), (
+            f"inconsistent state: {b}"
+        )
+    run("stop during URL resolve", t_stop_during_url_resolve)
+
+    def t_rapid_volume_change():
+        # Volume is a config-only command and works while disconnected,
+        # so this test deliberately doesn't call assert_started — we're
+        # validating the API's last-write-wins behavior, not the TS
+        # transport. That also spares the TS3 anti-flood budget.
+        for v in (10, 25, 50, 75, 100, 1):
+            s, _ = api(
+                f"/api/player/{bot_id}/volume",
+                method="POST",
+                json_body={"volume": v},
+            )
+            assert s == 200, f"volume POST failed: {s}"
+        time.sleep(0.3)
+        b = get_bot(bot_id)
+        assert b["volume"] == 1, f"final volume wrong: {b['volume']}"
+    run("rapid volume changes converge", t_rapid_volume_change)
+
+
+def group_websocket(bot_id: str):
+    print("\n== websocket broadcasts ==")
+    try:
+        from websocket import create_connection
+    except Exception as e:
+        skip("websocket state broadcasts", f"websocket lib unavailable: {e}")
+        return
+
+    try:
+        ws = create_connection("ws://localhost:3000/ws", timeout=5)
+    except Exception as e:
+        skip("websocket state broadcasts", f"connect failed: {e}")
+        return
+
+    ws.settimeout(0.3)
+    messages: list[dict] = []
+    stop_reader = threading.Event()
+
+    def reader():
+        while not stop_reader.is_set():
+            try:
+                raw = ws.recv()
+                if not raw:
+                    break
+                try:
+                    messages.append(json.loads(raw))
+                except Exception:
+                    pass
+            except Exception:
+                # recv() timeout or closed — keep trying until stop_reader
+                if stop_reader.is_set():
+                    break
+                continue
+
+    reader_thread = threading.Thread(target=reader, daemon=True)
+    reader_thread.start()
+
+    try:
+        def t_init():
+            time.sleep(0.6)
+            types = [m.get("type") for m in messages]
+            assert "init" in types, f"no init message; got: {types}"
+        run("init message on connect", t_init)
+
+        def t_state_change_on_play():
+            assert_started(bot_id)
+            messages.clear()
+            api(
+                f"/api/player/{bot_id}/play",
+                method="POST",
+                json_body={"query": "the mass", "platform": "netease"},
+            )
+            time.sleep(1.5)
+            types = [m.get("type") for m in messages]
+            assert "stateChange" in types, (
+                f"no stateChange after play; got types: {types}"
+            )
+            api(f"/api/player/{bot_id}/stop", method="POST")
+        run("stateChange broadcast on play", t_state_change_on_play)
+
+        def t_bot_disconnected_event():
+            assert_started(bot_id)
+            messages.clear()
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+            time.sleep(1.5)
+            types = [m.get("type") for m in messages]
+            assert "botDisconnected" in types or "stateChange" in types, (
+                f"no disconnect event; got: {types}"
+            )
+        run("botDisconnected event on stop", t_bot_disconnected_event)
+    finally:
+        stop_reader.set()
+        try:
+            ws.close()
+        except Exception:
+            pass
+
+
+# ----------------------------- main ----------------------------------------
+
+
+def main() -> int:
+    _, data = api("/api/bot/")
+    if not isinstance(data, dict) or not data.get("bots"):
+        print("[fatal] no bots registered — create one via the WebUI first")
+        return 2
+
+    target = data["bots"][0]
+    bot_id = target["id"]
+    initial_connected = target["connected"]
+    initial_volume = target["volume"]
+    initial_mode = target["playMode"]
+
+    print(f"[init] target bot  = {bot_id[:8]}  ({target['name']})")
+    print(
+        f"[init] initial state: connected={initial_connected} "
+        f"volume={initial_volume} mode={initial_mode}"
+    )
+
+    try:
+        # Read-only / no-lifecycle groups first — they don't consume TS3
+        # anti-flood budget.
+        group_infrastructure(bot_id)
+        group_auth_status()
+        group_search()
+
+        # Lifecycle-heavy groups — interleave with small breathers so the
+        # TS3 server's per-IP reconnect limit doesn't start throttling us.
+        group_lifecycle(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+
+        group_playback(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+
+        group_queue_ops(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+
+        group_input_validation(bot_id)
+        group_disconnect_corners(bot_id)
+        group_seek(bot_id)
+        time.sleep(ANTIFLOOD_BREATHER)
+
+        group_races(bot_id)
+        # Extra breather before websocket group — races is the heaviest
+        # consumer of TS3 reconnect budget (disconnect-during-connect and
+        # stop-during-url-resolve each burn one cycle), and the websocket
+        # group needs a clean reconnect to observe live state broadcasts.
+        time.sleep(ANTIFLOOD_BREATHER * 3)
+
+        group_websocket(bot_id)
+    finally:
+        # Restore initial state — this runs even if a test raised
+        try:
+            api(
+                f"/api/player/{bot_id}/volume",
+                method="POST",
+                json_body={"volume": initial_volume},
+            )
+            api(
+                f"/api/player/{bot_id}/mode",
+                method="POST",
+                json_body={"mode": initial_mode},
+            )
+            api(f"/api/player/{bot_id}/stop", method="POST")
+            if initial_connected:
+                start_and_wait(bot_id)
+            else:
+                stop_and_wait(bot_id)
+        except Exception as e:
+            print(f"[warn] restore failed: {e}")
+
+    print()
+    print("=" * 60)
+    print(f"  PASSED:  {passed}")
+    print(f"  FAILED:  {failed}")
+    print(f"  SKIPPED: {skipped}")
+    print("=" * 60)
+
+    if failed > 0:
+        print("\nFailed tests:")
+        for r in results:
+            if r.status in ("FAIL", "ERROR"):
+                print(f"  [{r.status}] {r.name}: {r.detail}")
+
+    return 0 if failed == 0 else 1
+
+
+if __name__ == "__main__":
+    import sys
+
+    sys.exit(main())
@@ -0,0 +1,166 @@
+"""More corner-case regressions.
+
+A. resolveAndPlay disconnect-during-URL-resolve race
+   The bot checks !this.connected at the top of resolveAndPlay, but the
+   URL-resolve await can take several seconds. If stop is called during
+   that window, playback would previously start on a disconnected bot.
+
+B. /seek NaN/Infinity rejection
+   typeof NaN === "number" and NaN < 0 is false, so a plain range check
+   leaks NaN through and corrupts seekOffset / getElapsed.
+"""
+import threading
+import time
+import requests
+
+BASE = "http://localhost:3000"
+
+
+def api(path, method="GET", **kw):
+    return getattr(requests, method.lower())(f"{BASE}{path}", timeout=30, **kw)
+
+
+def get_bot(bot_id):
+    return next(b for b in api("/api/bot/").json()["bots"] if b["id"] == bot_id)
+
+
+def wait_connected(bot_id, want, timeout=15):
+    end = time.time() + timeout
+    while time.time() < end:
+        if get_bot(bot_id)["connected"] is want:
+            return True
+        time.sleep(0.15)
+    return False
+
+
+def test_resolve_play_stop_race(bot_id):
+    """Fire stopBot during the /play call's URL resolve window."""
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    wait_connected(bot_id, False)
+    time.sleep(1)
+    api(f"/api/bot/{bot_id}/start", method="POST")
+    wait_connected(bot_id, True)
+
+    # Schedule a stop 150ms into the play call — that lands inside the
+    # provider.getSongUrl await, which is where the race lives.
+    def delayed_stop():
+        time.sleep(0.15)
+        try:
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+        except Exception:
+            pass
+
+    threading.Thread(target=delayed_stop, daemon=True).start()
+
+    try:
+        api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json={"query": "the mass", "platform": "netease"},
+        )
+    except Exception:
+        pass
+
+    # Give both calls time to settle fully
+    time.sleep(3)
+    b = get_bot(bot_id)
+    # Key invariant: we never want connected=false AND playing=true. That
+    # pair is the exact Bug C symptom and would indicate the resolveAndPlay
+    # post-await check didn't fire.
+    assert not (b["connected"] is False and b["playing"] is True), (
+        f"inconsistent state after race: {b}"
+    )
+    print(
+        f"[PASS] resolveAndPlay stop-race — final state consistent "
+        f"(connected={b['connected']} playing={b['playing']})"
+    )
+
+
+def test_seek_nan_rejected(bot_id):
+    """Verify that NaN and Infinity seek positions are rejected at the API
+    layer (instead of poisoning seekOffset)."""
+    api(f"/api/bot/{bot_id}/stop", method="POST")
+    wait_connected(bot_id, False)
+    api(f"/api/bot/{bot_id}/start", method="POST")
+    wait_connected(bot_id, True)
+
+    # Start a real song so there is an active playback to seek against
+    api(
+        f"/api/player/{bot_id}/play",
+        method="POST",
+        json={"query": "the mass", "platform": "netease"},
+    )
+    time.sleep(1.2)
+
+    # JSON spec doesn't allow NaN/Infinity literals, but Python's json
+    # encoder emits them as bare tokens when allow_nan=True (the default).
+    # Express's body-parser rejects them as invalid JSON, which itself is
+    # a form of rejection. We additionally verify that sending a string
+    # "NaN" or a negative value is also rejected with a clean 400.
+    r = api(
+        f"/api/player/{bot_id}/seek",
+        method="POST",
+        json={"position": -5},
+    )
+    assert r.status_code == 400, f"negative seek should be rejected, got {r.status_code}"
+
+    r = api(
+        f"/api/player/{bot_id}/seek",
+        method="POST",
+        json={"position": "fifty"},
+    )
+    assert r.status_code == 400, f"string seek should be rejected, got {r.status_code}"
+
+    # Directly send NaN in raw body (body-parser will likely 400 it)
+    r = requests.post(
+        f"{BASE}/api/player/{bot_id}/seek",
+        data='{"position": NaN}',
+        headers={"Content-Type": "application/json"},
+        timeout=10,
+    )
+    assert r.status_code >= 400, f"NaN seek should be rejected, got {r.status_code}"
+
+    # After the junk attempts, a valid seek still works and the elapsed
+    # time is a finite number (not NaN).
+    r = api(
+        f"/api/player/{bot_id}/seek",
+        method="POST",
+        json={"position": 30},
+    )
+    assert r.status_code == 200, f"valid seek failed: {r.text[:120]}"
+
+    elapsed_resp = api(f"/api/player/{bot_id}/elapsed")
+    elapsed = elapsed_resp.json().get("elapsed")
+    assert elapsed is not None and isinstance(elapsed, (int, float)), (
+        f"elapsed should be a number, got {elapsed}"
+    )
+    # Could be exactly 30 or a tiny bit more if a frame has advanced
+    assert 29 <= elapsed < 40, f"elapsed after seek(30) out of range: {elapsed}"
+    print(f"[PASS] seek NaN/Infinity rejected; valid seek produces finite elapsed={elapsed:.2f}")
+
+
+def main():
+    bots = api("/api/bot/").json()["bots"]
+    if not bots:
+        print("[skip] no bots")
+        return
+    bot_id = bots[0]["id"]
+    initial = bots[0]["connected"]
+    print(f"[init] bot={bot_id[:8]} initial connected={initial}")
+
+    try:
+        test_resolve_play_stop_race(bot_id)
+        test_seek_nan_rejected(bot_id)
+        print("ALL GREEN")
+    finally:
+        if initial:
+            api(f"/api/bot/{bot_id}/start", method="POST")
+            wait_connected(bot_id, True)
+        else:
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+            wait_connected(bot_id, False)
+        print(f"[restore] connected={get_bot(bot_id)['connected']}")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,212 @@
+"""Stress-test two bots playing music concurrently on the same TS server.
+
+Creates two temporary bots (or reuses existing named ones), starts them,
+plays music on both, and polls /api/bot/ every 2 seconds to detect when
+(if) either bot disconnects or stops playing. Cleans up on exit.
+
+Usage:
+    python scripts/test_multibot.py --minutes 3
+    python scripts/test_multibot.py --minutes 10 --host 127.0.0.1 --port 9987
+"""
+import argparse
+import sys
+import time
+from dataclasses import dataclass
+
+import requests
+
+API = "http://localhost:3000"
+POLL_INTERVAL = 2.0
+TEST_BOT_NAMES = ("mbtest1", "mbtest2")
+TEST_BOT_NICKS = ("MBTest1", "MBTest2")
+QUERIES = ("the mass", "lofi")  # one different song per bot
+
+
+@dataclass
+class BotSnapshot:
+    t: float
+    connected: bool
+    playing: bool
+    song: str | None
+
+
+def api_get(path: str):
+    r = requests.get(f"{API}{path}", timeout=5)
+    r.raise_for_status()
+    return r.json()
+
+
+def api_post(path: str, json=None):
+    r = requests.post(f"{API}{path}", json=json, timeout=15)
+    r.raise_for_status()
+    return r.json()
+
+
+def api_delete(path: str):
+    r = requests.delete(f"{API}{path}", timeout=10)
+    r.raise_for_status()
+    return r.json()
+
+
+def cleanup_existing(names: tuple[str, ...]) -> None:
+    bots = api_get("/api/bot/")["bots"]
+    for b in bots:
+        if b["name"] in names:
+            try:
+                api_post(f"/api/player/{b['id']}/stop")
+            except Exception:
+                pass
+            try:
+                api_delete(f"/api/bot/{b['id']}")
+                print(f"[cleanup] removed existing bot {b['name']} ({b['id']})")
+            except Exception as e:
+                print(f"[cleanup] failed to remove {b['name']}: {e}")
+
+
+def create_bot(name: str, nickname: str, host: str, port: int) -> str:
+    res = api_post(
+        "/api/bot/",
+        json={
+            "name": name,
+            "serverAddress": host,
+            "serverPort": port,
+            "nickname": nickname,
+            "autoStart": False,
+        },
+    )
+    bot_id = res["id"]
+    print(f"[create] {name} -> {bot_id}")
+    return bot_id
+
+
+def start_bot(bot_id: str) -> None:
+    api_post(f"/api/bot/{bot_id}/start")
+
+
+def play(bot_id: str, query: str) -> None:
+    api_post(f"/api/player/{bot_id}/play", json={"query": query, "platform": "netease"})
+
+
+def snapshot(bot_id: str, t0: float) -> BotSnapshot:
+    bots = api_get("/api/bot/")["bots"]
+    b = next((x for x in bots if x["id"] == bot_id), None)
+    if not b:
+        return BotSnapshot(time.time() - t0, False, False, None)
+    song = b["currentSong"]["name"] if b.get("currentSong") else None
+    return BotSnapshot(time.time() - t0, b["connected"], b["playing"], song)
+
+
+def run(minutes: float, host: str, port: int) -> int:
+    print(f"[setup] duration={minutes}min host={host}:{port}")
+
+    cleanup_existing(TEST_BOT_NAMES)
+
+    bot_ids = [
+        create_bot(TEST_BOT_NAMES[0], TEST_BOT_NICKS[0], host, port),
+        create_bot(TEST_BOT_NAMES[1], TEST_BOT_NICKS[1], host, port),
+    ]
+
+    # Start both, allowing a small stagger to avoid handshake collision
+    for i, bid in enumerate(bot_ids):
+        start_bot(bid)
+        print(f"[start] bot{i+1} started")
+        time.sleep(1.5)
+
+    # Wait until both are connected (or bail after 15s)
+    deadline = time.time() + 15
+    while time.time() < deadline:
+        bots = {b["id"]: b for b in api_get("/api/bot/")["bots"]}
+        if all(bots[b]["connected"] for b in bot_ids):
+            print("[start] both bots connected")
+            break
+        time.sleep(0.5)
+    else:
+        print("[fatal] bots did not both come online in 15s")
+        cleanup_existing(TEST_BOT_NAMES)
+        return 2
+
+    # Kick off playback on both
+    for i, bid in enumerate(bot_ids):
+        play(bid, QUERIES[i])
+        print(f"[play] bot{i+1} -> {QUERIES[i]!r}")
+
+    t0 = time.time()
+    end = t0 + minutes * 60
+    first_drop: dict[str, float] = {}
+    last_state: dict[str, BotSnapshot] = {}
+
+    print(f"[monitor] polling every {POLL_INTERVAL}s for {minutes} min...")
+    print(f"{'time':>7}  {'bot1':<40}  {'bot2':<40}")
+
+    def fmt(snap: BotSnapshot) -> str:
+        flag = ("C" if snap.connected else "-") + ("P" if snap.playing else "-")
+        song = (snap.song or "").replace("\n", " ")[:30]
+        return f"{flag} {song}"
+
+    try:
+        while time.time() < end:
+            snaps = [snapshot(bid, t0) for bid in bot_ids]
+            elapsed = int(time.time() - t0)
+            row = f"{elapsed:>6}s  {fmt(snaps[0]):<40}  {fmt(snaps[1]):<40}"
+            # Only print when state changes or every 10s
+            changed = False
+            for bid, s in zip(bot_ids, snaps):
+                prev = last_state.get(bid)
+                if (prev is None
+                        or prev.connected != s.connected
+                        or prev.playing != s.playing
+                        or prev.song != s.song):
+                    changed = True
+                last_state[bid] = s
+                if not s.connected and bid not in first_drop:
+                    first_drop[bid] = s.t
+            if changed or elapsed % 10 == 0:
+                print(row)
+
+            # If both stopped playing but are still connected, re-queue the same song
+            for i, (bid, s) in enumerate(zip(bot_ids, snaps)):
+                if s.connected and not s.playing:
+                    try:
+                        play(bid, QUERIES[i])
+                    except Exception as e:
+                        print(f"[warn] re-play bot{i+1} failed: {e}")
+
+            time.sleep(POLL_INTERVAL)
+    except KeyboardInterrupt:
+        print("\n[abort] interrupted")
+
+    # Summary
+    total = time.time() - t0
+    print()
+    print("=" * 60)
+    print(f"Total observed time: {total:.1f}s")
+    for i, bid in enumerate(bot_ids):
+        drop = first_drop.get(bid)
+        if drop is None:
+            print(f"  bot{i+1} ({TEST_BOT_NICKS[i]}): stayed connected the whole run")
+        else:
+            print(f"  bot{i+1} ({TEST_BOT_NICKS[i]}): FIRST DISCONNECT at t+{drop:.1f}s")
+    print("=" * 60)
+
+    # Cleanup
+    cleanup_existing(TEST_BOT_NAMES)
+    print("[cleanup] done")
+
+    return 0 if not first_drop else 1
+
+
+def main() -> int:
+    p = argparse.ArgumentParser()
+    p.add_argument("--minutes", type=float, default=3.0)
+    p.add_argument("--host", default="127.0.0.1")
+    p.add_argument("--port", type=int, default=9987)
+    args = p.parse_args()
+    try:
+        return run(args.minutes, args.host, args.port)
+    except requests.HTTPError as e:
+        print(f"[http-error] {e} body={e.response.text[:200] if e.response else ''}")
+        return 3
+
+
+if __name__ == "__main__":
+    sys.exit(main())
@@ -1,105 +1,119 @@
-"""Reproduce the player-bar-not-appearing bug."""
+"""Reproduce / regression-check the player-bar-not-appearing bug.
+
+Captures the bot's initial playback state and restores it on exit so the
+test never leaves the user with surprise music or a cleared queue.
+"""
 import time
 import requests
 from playwright.sync_api import sync_playwright

 BASE = "http://localhost:3000"

-def get_bot_id():
-    r = requests.get(f"{BASE}/api/bot/")
-    return r.json()["bots"][0]["id"]

-def stop(bot_id):
-    requests.post(f"{BASE}/api/player/{bot_id}/stop")
-    requests.post(f"{BASE}/api/player/{bot_id}/clear")
+def api(path, method="GET", **kw):
+    fn = getattr(requests, method.lower())
+    r = fn(f"{BASE}{path}", timeout=10, **kw)
+    r.raise_for_status()
+    return r.json() if r.text else None

-def play(bot_id, query="test"):
-    requests.post(f"{BASE}/api/player/{bot_id}/play", json={"query": query, "platform": "netease"})

-with sync_playwright() as p:
-    browser = p.chromium.launch(headless=True)
-    ctx = browser.new_context()
-    ctx.add_init_script("""
-      (() => {
-        const OrigWS = window.WebSocket;
-        window.__wsMessages = [];
-        window.WebSocket = function(...args) {
-          const ws = new OrigWS(...args);
-          ws.addEventListener('message', (ev) => {
-            try {
-              const d = JSON.parse(ev.data);
-              const summary = {type: d.type, botId: d.botId};
-              if (d.status) summary.playing = d.status.playing;
-              if (d.status?.currentSong) summary.song = d.status.currentSong.name;
-              window.__wsMessages.push(summary);
-              console.log('WS_MSG ' + JSON.stringify(summary));
-            } catch(e) {}
-          });
-          return ws;
-        };
-        Object.assign(window.WebSocket, OrigWS);
-      })();
-    """)
-    page = ctx.new_page()
-    logs = []
-    page.on("console", lambda msg: logs.append(f"[{msg.type}] {msg.text}"))
+def get_bot(bot_id):
+    return next(b for b in api("/api/bot/")["bots"] if b["id"] == bot_id)

-    bot_id = get_bot_id()
-    stop(bot_id)
-    time.sleep(0.6)

-    page.goto(BASE)
-    page.wait_for_load_state("networkidle")
-    time.sleep(0.8)
+def capture_state(bot_id):
+    b = get_bot(bot_id)
+    return {
+        "playing": b["playing"],
+        "paused": b["paused"],
+        "song": (b["currentSong"] or {}).get("name"),
+    }

-    # Hook into WebSocket messages from the page to confirm they arrive
-    page.evaluate("""
-      () => {
-        const origWS = window.WebSocket;
-        // Already connected via useWebSocket — tap Pinia store directly
-        // Expose store state via window
-        const store = window.__pinia?._s?.get('player');
-        window.__getStore = () => ({
-          bots: JSON.parse(JSON.stringify(store?.bots || [])),
-          activeBotId: store?.activeBotId,
-          currentSong: store?.currentSong ? JSON.parse(JSON.stringify(store.currentSong)) : null,
-          isPlaying: store?.isPlaying,
-        });
-      }
-    """)

-    snap_before = page.evaluate("() => window.__getStore?.()")
-    print("store before:", snap_before)
+def main():
+    bots = api("/api/bot/")["bots"]
+    if not bots:
+        print("[skip] no bots")
+        return
+    bot_id = bots[0]["id"]
+    initial = capture_state(bot_id)
+    print(f"[init] initial state: {initial}")

-    # Player bar should be absent right now (nothing playing)
-    initial = page.locator(".player-wrapper").count()
-    print(f"initial .player-wrapper count: {initial}")
+    try:
+        # Clear slate
+        api(f"/api/player/{bot_id}/stop", method="POST")
+        time.sleep(0.6)

-    # Trigger play via API (simulates any play trigger)
-    play(bot_id, "test")
+        with sync_playwright() as p:
+            browser = p.chromium.launch(headless=True)
+            try:
+                ctx = browser.new_context()
+                ctx.add_init_script(
+                    """
+                    (() => {
+                      const OrigWS = window.WebSocket;
+                      window.__wsMessages = [];
+                      window.WebSocket = function(...args) {
+                        const ws = new OrigWS(...args);
+                        ws.addEventListener('message', (ev) => {
+                          try {
+                            const d = JSON.parse(ev.data);
+                            window.__wsMessages.push({type: d.type, botId: d.botId});
+                          } catch(e) {}
+                        });
+                        return ws;
+                      };
+                      Object.assign(window.WebSocket, OrigWS);
+                    })();
+                    """
+                )
+                page = ctx.new_page()
+                page.goto(BASE)
+                page.wait_for_load_state("networkidle")
+                time.sleep(0.8)

-    # Poll for up to 6s to see if player bar appears automatically
-    appeared_at = None
-    for i in range(60):
-        if page.locator(".player-wrapper").count() > 0:
-            appeared_at = i * 0.1
-            break
-        page.wait_for_timeout(100)
+                assert page.locator(".player-wrapper").count() == 0, (
+                    "player bar should be hidden before playback"
+                )

-    print(f"player bar appeared after: {appeared_at}")
-    snap_after = page.evaluate("() => window.__getStore?.()")
-    print("store after:", snap_after)
+                # Trigger play via API (simulates any play trigger)
+                api(
+                    f"/api/player/{bot_id}/play",
+                    method="POST",
+                    json={"query": "the mass", "platform": "netease"},
+                )

-    # After force reload, does it appear?
-    if appeared_at is None:
-        page.reload()
-        page.wait_for_load_state("networkidle")
-        time.sleep(0.8)
-        after_reload = page.locator(".player-wrapper").count()
-        print(f"after reload .player-wrapper count: {after_reload}")
+                # Poll for up to 6s to see if player bar appears automatically
+                appeared_at = None
+                for i in range(60):
+                    if page.locator(".player-wrapper").count() > 0:
+                        appeared_at = i * 0.1
+                        break
+                    page.wait_for_timeout(100)

-    print("--- console logs ---")
-    for line in logs[-20:]:
-        print(line)
+                if appeared_at is None:
+                    msgs = page.evaluate("() => window.__wsMessages")
+                    print(f"[FAIL] player bar never appeared; WS msgs: {msgs}")
+                    raise AssertionError("player bar did not auto-show on stateChange")
+                print(f"[PASS] player bar appeared after {appeared_at:.1f}s")
+            finally:
+                browser.close()
+    finally:
+        # Restore: stop the "test" song we triggered, then re-apply initial
+        # state as best we can. We can't re-queue the user's previous song,
+        # but we can at least stop ours and leave the bot idle if it was idle.
+        try:
+            api(f"/api/player/{bot_id}/stop", method="POST")
+        except Exception as e:
+            print(f"[warn] failed to stop test song on cleanup: {e}")
+        post = capture_state(bot_id)
+        print(f"[restore] bot now idle (was playing={initial['playing']} song={initial['song']!r})")
+        if initial["playing"] and initial["song"]:
+            print(
+                f"[note] initial bot was playing {initial['song']!r}; "
+                "this test cannot resume arbitrary tracks — you may need to restart playback"
+            )

-    browser.close()
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,100 @@
+"""E2E: the new power button in the Bot Selector dropdown toggles bot connected state.
+
+Captures the target bot's initial connected state and restores it on exit
+(including on assertion failure), so running this test never pollutes the
+user's current bot setup.
+"""
+import time
+import requests
+from playwright.sync_api import sync_playwright
+
+BASE = "http://localhost:3000"
+
+
+def get_bot(bot_id):
+    return next(b for b in requests.get(f"{BASE}/api/bot/").json()["bots"] if b["id"] == bot_id)
+
+
+def wait_for_connected(bot_id, want: bool, timeout_s: float = 12.0) -> bool:
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        if get_bot(bot_id)["connected"] is want:
+            return True
+        time.sleep(0.2)
+    return False
+
+
+def set_connected(bot_id, want: bool) -> None:
+    """Force the bot into the given connected state via API."""
+    current = get_bot(bot_id)["connected"]
+    if current == want:
+        return
+    endpoint = "start" if want else "stop"
+    requests.post(f"{BASE}/api/bot/{bot_id}/{endpoint}")
+    wait_for_connected(bot_id, want)
+
+
+def main():
+    bots = requests.get(f"{BASE}/api/bot/").json()["bots"]
+    if not bots:
+        print("[skip] no bots registered, nothing to test")
+        return
+    target = bots[0]
+    bot_id = target["id"]
+    initial_connected = target["connected"]
+    print(f"[init] target bot {target['name']} ({bot_id[:8]}), initial connected={initial_connected}")
+
+    try:
+        # Force bot disconnected before the test
+        set_connected(bot_id, False)
+        assert not get_bot(bot_id)["connected"], "bot should be disconnected at start"
+
+        with sync_playwright() as p:
+            browser = p.chromium.launch(headless=True)
+            try:
+                page = browser.new_page(viewport={"width": 1440, "height": 900})
+                page.goto(BASE)
+                page.wait_for_load_state("networkidle")
+                time.sleep(0.6)
+
+                # Open dropdown
+                page.locator(".bot-selector-btn").click()
+                page.wait_for_selector(".bot-power-btn")
+
+                # Click the power button to start
+                page.locator(".bot-power-btn").first.click()
+                print("[ui] clicked power (start)")
+
+                assert wait_for_connected(bot_id, True), "bot should be connected after clicking start"
+                print("[api] bot connected = True")
+
+                # Let UI catch up via WS then re-open the dropdown to re-check class
+                time.sleep(1.0)
+                page.locator(".bot-selector-btn").click()  # close
+                time.sleep(0.2)
+                page.locator(".bot-selector-btn").click()  # reopen
+                page.wait_for_selector(".bot-power-btn.online", timeout=3000)
+                print("[ui] power button now shows .online class")
+
+                # Click again to stop
+                page.locator(".bot-power-btn.online").first.click()
+                print("[ui] clicked power (stop)")
+
+                assert wait_for_connected(bot_id, False), "bot should be disconnected after clicking stop"
+                print("[api] bot connected = False")
+
+                print("[PASS] power button toggles bot connection")
+            finally:
+                browser.close()
+    finally:
+        # Always restore the initial state so the test never leaves the bot
+        # in an unexpected place
+        set_connected(bot_id, initial_connected)
+        final = get_bot(bot_id)["connected"]
+        print(f"[restore] bot connected={final} (initial was {initial_connected})")
+        if final != initial_connected:
+            print("[warn] failed to restore initial connected state")
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,130 @@
+"""Regression for the 'connected=False but playing=True' stuck-state bug.
+
+After rapid disconnect/reconnect, the library could drop the connection
+(TS3 server anti-flood or a hung handshake). The bot then ended up in an
+inconsistent state: player.state='playing' but tsClient disconnected.
+
+This test verifies three fixes:
+  Bug A — startBot() has a 15s timeout instead of hanging forever on a
+          stalled handshake. /start returns a clean 500 instead of blocking.
+  Bug B — play/add/etc commands are rejected when the bot is not connected.
+  Bug C — the tsClient 'disconnected' handler always clears player state,
+          even when connect() never completed (so !this.connected).
+
+We also sanity-check that the bot recovers (can start a fresh cycle) after
+a transient failure.
+"""
+import time
+import requests
+
+BASE = "http://localhost:3000"
+
+
+def api(path, method="GET", **kw):
+    return getattr(requests, method.lower())(f"{BASE}{path}", timeout=30, **kw)
+
+
+def get_bot(bot_id):
+    return next(b for b in api("/api/bot/").json()["bots"] if b["id"] == bot_id)
+
+
+def wait_connected(bot_id, want, timeout=15):
+    end = time.time() + timeout
+    while time.time() < end:
+        if get_bot(bot_id)["connected"] is want:
+            return True
+        time.sleep(0.15)
+    return False
+
+
+def main():
+    bots = api("/api/bot/").json()["bots"]
+    if not bots:
+        print("[skip] no bots")
+        return
+    bot_id = bots[0]["id"]
+    initial_connected = bots[0]["connected"]
+    print(f"[init] bot={bot_id[:8]} initial connected={initial_connected}")
+
+    try:
+        # Start from a clean slate
+        api(f"/api/bot/{bot_id}/stop", method="POST")
+        wait_connected(bot_id, False)
+
+        # --- Bug B: play while disconnected must be rejected ---
+        r = api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json={"query": "rejection test", "platform": "netease"},
+        )
+        assert r.status_code >= 400, (
+            f"play while disconnected should fail, got {r.status_code} {r.text[:120]}"
+        )
+        assert "not connected" in r.text.lower(), (
+            f"expected 'not connected' error, got: {r.text[:200]}"
+        )
+        b = get_bot(bot_id)
+        assert not b["playing"], f"player shouldn't be playing after rejected /play: {b}"
+        print("[PASS] Bug B — /play rejected while bot disconnected; state untouched")
+
+        # --- Bug C: normal start→play→stop leaves player state clean ---
+        r = api(f"/api/bot/{bot_id}/start", method="POST")
+        assert r.status_code == 200, f"start failed {r.text[:120]}"
+        assert wait_connected(bot_id, True), "bot did not connect within 15s"
+
+        r = api(
+            f"/api/player/{bot_id}/play",
+            method="POST",
+            json={"query": "the mass", "platform": "netease"},
+        )
+        assert r.status_code == 200, f"play failed {r.text[:120]}"
+        time.sleep(1.2)
+        b = get_bot(bot_id)
+        assert b["connected"] and b["playing"], f"should be connected+playing: {b}"
+
+        api(f"/api/bot/{bot_id}/stop", method="POST")
+        assert wait_connected(bot_id, False, timeout=5), "bot did not disconnect"
+        b = get_bot(bot_id)
+        assert not b["playing"], (
+            f"player should have stopped after bot disconnect (Bug C): {b}"
+        )
+        print("[PASS] Bug C — stop clears both connected and playing state")
+
+        # --- Bug A: startBot has a deadline and returns a clean error if connect hangs ---
+        # We can't easily force a hang in-process, but we can sanity-check that
+        # startBot returns promptly (well under the 15s cap) on a normal run.
+        t0 = time.time()
+        r = api(f"/api/bot/{bot_id}/start", method="POST")
+        elapsed = time.time() - t0
+        assert r.status_code == 200, f"start failed {r.text[:120]}"
+        assert elapsed < 10, f"start should be prompt, took {elapsed:.1f}s"
+        assert wait_connected(bot_id, True), "bot did not connect"
+        print(
+            f"[PASS] Bug A — startBot completed in {elapsed:.2f}s "
+            "(deadline is 15s, would throw on hang)"
+        )
+
+        # --- Recovery: after any failure, another start should work ---
+        api(f"/api/bot/{bot_id}/stop", method="POST")
+        wait_connected(bot_id, False)
+        # Give TS3 server a moment to forget us (anti-flood grace)
+        time.sleep(2)
+        r = api(f"/api/bot/{bot_id}/start", method="POST")
+        assert r.status_code == 200, f"recovery start failed {r.text[:120]}"
+        assert wait_connected(bot_id, True), "bot did not recover"
+        print("[PASS] recovery — bot reconnects cleanly after a cycle")
+
+        print("ALL GREEN")
+
+    finally:
+        if initial_connected:
+            api(f"/api/bot/{bot_id}/start", method="POST")
+            wait_connected(bot_id, True)
+        else:
+            api(f"/api/bot/{bot_id}/stop", method="POST")
+            wait_connected(bot_id, False)
+        print(f"[restore] bot connected={get_bot(bot_id)['connected']} (was {initial_connected})")
+
+
+if __name__ == "__main__":
+    main()