Track Chief

Connect Spotify

Track Chief sorts your library into your own genre playlists by learning from what you've already sorted, and feeds your mood playlists. Everything runs in this browser — no server, no password, nothing leaves your phone.

Log in with Spotify

Phase 0: Capability probe

Testing every endpoint this app depends on before building anything on top of it.

Continue to setup

Which playlists are which?

Spotify's API can't see your folders, so confirm my guesses: which playlists live in your genre folder, which in your mood folder, and which to ignore entirely.

Save & sync library

▶ Triage mode

Dry run (show what would happen, write nothing) Apply checked placements

SPARK

◉ READ THE ROOM NEW HEAT // NEVER-PLAYED PLAY THROUGH THIS APP ENGAGE SHOCK MODE

Library

Re-sync library Rebuild model

Data export import

Drop your Spotify data export ZIP here (Privacy Settings → Download your data). Parsed entirely on this device. Unlocks skip-aware recommendations, time-of-day mood matching, and session similarity.

Playlist buckets

Edit genre / mood / ignore mapping

Appearance

Theme

Primary drives the glow, buttons and highlights. Secondary tints every background and panel. The whole app rebuilds itself around your two picks, live.

Headline font

Shock mode

SPARK's evil twin. Flip this and the booth belongs to SHOCK — way filthier mouth, zero chill, roasts with the safety off. House rules even in the gutter: no slurs, no hate. Everything else is fair game. Flip back anytime.

ENGAGE SHOCK MODE

SPARK voice

Pick the voice SPARK speaks with. Your phone's built-in voices are free; an ElevenLabs key (free tier) unlocks a human-sounding one — spoken lines get cached on your device so the free quota goes far.

Built-in voice

Best available (auto)

Neural voice (optional)

SPARK's brain (optional)

Give SPARK a real AI brain: a free Hugging Face account gets you an access token (Settings → Access Tokens → New token, with Inference Providers permission). Paste it here and SPARK stops reading from his script — he generates fresh replies to whatever you say, in full character, SHOCK included. Costs fractions of a cent per reply after the free monthly credits. No token? He falls back to his script, same as now.

Qwen 2.5 72B — sharp, good default Qwen 2.5 7B — cheapest, lighter wit Llama 3.3 70B — accept its license on HF first Custom model id…

Advanced: use a different provider (Groq, Together, OpenRouter…)

Any OpenAI-compatible endpoint works. Example: OpenRouter's free tier — key from openrouter.ai, endpoint https://openrouter.ai/api/v1/chat/completions, and a model id ending in :free (e.g. nvidia/nemotron-3-super-120b-a12b:free — the current free list lives at openrouter.ai/models). Groq's console (console.groq.com) also works where their network allows it.

Last.fm (optional)

Free API key expands mood recommendations beyond your own library via similar-track lookups.

Save key

Backup & transfer

Everything Track Chief knows — buckets, subgenre tree, corrections, caches, undo history — in one file. Import it on any device to clone your setup exactly.

Export setup file Import setup file

Undo

Diagnostics

Re-run endpoint probe

Playback log

—

Swap Client ID (keeps all your data)

If Spotify rate-limits this app for hours, create a second app in the developer dashboard (same Redirect URI) and paste its Client ID here. You'll re-approve access once; buckets, sync cache, and corrections are kept.

Swap & re-login

Log out & wipe local data