SongToSpot — IR + AI Pipeline (Dev Demo)

Lazy OAuth (Login Only at “Send”)

# /api/playlist (excerpt)
user_sp = get_sp_user()
if not user_sp:
    oauth = spotify_oauth()
    return jsonify({
        "ok": False,
        "need_login": True,
        "login_url": oauth.get_authorize_url()
    }), 401

if (res.status === 401) {
  const body = await res.clone().json().catch(() => ({}));
  if (body?.need_login && body.login_url) {
    const ok = await popupLoginAndWait(body.login_url);
    if (ok) return postPlaylist(payload); // retry
  }
}

Progressive Search (Optimistic → Verified)

def _resolve_spotify_url(artist, title, market=None):
    sp = get_sp_client_credentials()
    for q in (f'track:"{title}" artist:"{artist}"', f"{artist} {title}"):
        res = sp.search(q=q, type="track", limit=1, market=market or SPOTIFY_MARKET)
        items = (res or {}).get("tracks", {}).get("items", [])
        if items:
            return items[0]["external_urls"]["spotify"]
    return None

// Light up 🎧 links as they resolve
for (const row of candidates) {
  const res = await fetch('/api/quicksearch', { method:'POST',
    headers:{'Content-Type':'application/json'},
    body: JSON.stringify({ artist: row.artist, title: row.title })
  });
  const j = await res.json();
  if (j.found && j.spotify_url) updateRowWithLink(row, j.spotify_url);
}

Parallel ID Resolution (Order‑Preserving)

from concurrent.futures import ThreadPoolExecutor, as_completed

def _bulk_resolve_track_ids(sp, urls, max_workers=10):
    ids = [None] * len(urls)

    def _fetch_one(idx_url):
        idx, url = idx_url
        try:
            return idx, sp.track(url)["id"]
        except Exception:
            return idx, None

    with ThreadPoolExecutor(max_workers=max_workers) as ex:
        for f in as_completed([ex.submit(_fetch_one, (i,u)) for i,u in enumerate(urls)]):
            idx, tid = f.result()
            ids[idx] = tid
    return [t for t in ids if t]

SongToSpot — From Freeform Prompts to Playable Playlists

Abstract

Pipeline Sandbox (Interactive)

Semantic Decomposition

Candidate Expansion

Fallback Strategies (Heuristics)

Tracks (from API if available)

Key Implementation Patterns (Interactive Code)

IR + AI Pipeline Diagram