Update app.py

app.py

@@ -1,50 +1,53 @@
 # -*- coding: utf-8 -*-
-"""app.ipynb
-
-Automatically generated by Colab.
-
-Original file is located at
-    https://colab.research.google.com/drive/1vb2j78WT7l9XQiXUBvl1VAtELSAzOUZJ
+"""
+HarmoniFind – Semantic Spotify Search
+HF Spaces app.py
 """
 
 import os
 import random
+from difflib import SequenceMatcher
+
 import numpy as np
 import pandas as pd
 import faiss
 import gradio as gr
+import html as html_lib
+
 from sentence_transformers import SentenceTransformer
 from huggingface_hub import InferenceClient
 import spotipy
 from spotipy.oauth2 import SpotifyClientCredentials
-from difflib import SequenceMatcher
-import html as html_lib 
 
+# ---------- Paths to precomputed data ----------
 
-# ---------- Load data ----------
 CLEAN_CSV_PATH = "df_combined_clean.csv"
 EMB_PATH       = "df_embed.npz"
 INDEX_PATH     = "hnsw.index"
 
+# ---------- Load data ----------
+
 df_combined = pd.read_csv(CLEAN_CSV_PATH)
 emb_data = np.load(EMB_PATH)
 df_embeddings = emb_data["df_embeddings"].astype("float32")
 index = faiss.read_index(INDEX_PATH)
 
-# ---------- Secrets ----------
+# ---------- Secrets from env (HF Space secrets) ----------
 
 HF_TOKEN = os.getenv("HF_TOKEN")
 
+# HF Space secrets should be named SPOTIPY_CLIENT_ID / SPOTIPY_CLIENT_SECRET
 SPOTIFY_CLIENT_ID = os.getenv("SPOTIPY_CLIENT_ID")
 SPOTIFY_CLIENT_SECRET = os.getenv("SPOTIPY_CLIENT_SECRET")
 
-
 # ---------- Models ----------
+
+# Query encoder (same as notebook)
 query_embedder = SentenceTransformer("all-mpnet-base-v2")
 
+# LLaMA-2 for query expansion (remote HF Inference)
 LLAMA_MODEL_ID = "meta-llama/Llama-2-7b-chat-hf"
 
-# Create a generic client; we'll pass model per call
 hf_client = None
 if HF_TOKEN:
     try:
@@ -53,25 +56,33 @@ if HF_TOKEN:
         print("⚠️ Could not initialize HF Inference client:", repr(e))
         hf_client = None
 
+# Spotify client
 sp = None
 if SPOTIFY_CLIENT_ID and SPOTIFY_CLIENT_SECRET:
-    auth = SpotifyClientCredentials(client_id=SPOTIFY_CLIENT_ID, client_secret=SPOTIFY_CLIENT_SECRET)
-    sp = spotipy.Spotify(auth_manager=auth)
+    try:
+        auth = SpotifyClientCredentials(
+            client_id=SPOTIFY_CLIENT_ID,
+            client_secret=SPOTIFY_CLIENT_SECRET,
+        )
+        sp = spotipy.Spotify(auth_manager=auth)
+    except Exception as e:
+        print("⚠️ Could not initialize Spotify client:", repr(e))
+        sp = None
 
-# ---------- Helper functions ----------
-def encode_query(text):
+# ---------- Core helpers ----------
+
+def encode_query(text: str) -> np.ndarray:
     return query_embedder.encode([text], convert_to_numpy=True).astype("float32")
 
+
 def expand_with_llama(query: str) -> str:
     """
     Enrich the query using LLaMA via HF Inference.
 
-    On HF Spaces, the Inference provider can sometimes be unavailable
-    or misconfigured (giving the StopIteration error you saw). In that
-    case, we log and fall back to the raw query so the UI keeps working.
+    If anything fails (no client, provider issues, rate limits, etc.),
+    we log and fall back to the raw query so the app keeps working.
     """
     if hf_client is None or not HF_TOKEN:
-        # No client/token -> behave like "no expansion"
         return query
 
     prompt = f"""You are helping someone search a lyrics catalog.
@@ -90,14 +101,13 @@ Output (no explanation, just titles or keywords):"""
     try:
         response = hf_client.text_generation(
             prompt,
-            model=LLAMA_MODEL_ID,          
+            model=LLAMA_MODEL_ID,
             max_new_tokens=96,
             temperature=0.2,
             repetition_penalty=1.05,
         )
     except Exception as e:
-        
-        print("LLaMA expansion failed on HF, using raw query:", repr(e))
+        print("⚠️ LLaMA expansion failed on HF, using raw query:", repr(e))
         return query
 
     keywords = str(response).strip().replace("\n", " ")
@@ -105,252 +115,101 @@ Output (no explanation, just titles or keywords):"""
     return expanded
 
 
-def distances_to_similarity_pct(dists):
-    if len(dists) == 0: return np.array([])
+def distances_to_similarity_pct(dists: np.ndarray) -> np.ndarray:
+    if len(dists) == 0:
+        return np.array([])
     dmin, dmax = dists.min(), dists.max()
-    if dmax - dmin == 0: return np.ones_like(dists) * 100
+    if dmax - dmin == 0:
+        return np.ones_like(dists) * 100
     sims = 100 * (1 - (dists - dmin) / (dmax - dmin))
     return sims
 
-def label_vibes(sim):
-    if sim >= 90: return "dead-on"
-    elif sim >= 80: return "strong vibes"
-    elif sim >= 70: return "adjacent"
-    elif sim >= 60: return "stretch but related"
-    else: return "pretty random"
 
-def semantic_search(query, k=10, random_extra=0, use_llama=True):
-    if not query.strip():
-        return pd.DataFrame(columns=["artist","song","similarity_pct","vibes","is_random"])
+def label_vibes(sim: float) -> str:
+    if sim >= 90:
+        return "dead-on"
+    elif sim >= 80:
+        return "strong vibes"
+    elif sim >= 70:
+        return "adjacent"
+    elif sim >= 60:
+        return "stretch but related"
+    else:
+        return "pretty random"
+
+
+def semantic_search(query: str, k: int = 10, random_extra: int = 0, use_llama: bool = True) -> pd.DataFrame:
+    if not query or not query.strip():
+        return pd.DataFrame(columns=["artist", "song", "similarity_pct", "vibes", "is_random"])
+
     q_text = expand_with_llama(query) if use_llama else query
     q_vec = encode_query(q_text)
     dists, idxs = index.search(q_vec, k)
+
     sem_df = df_combined.iloc[idxs[0]].copy()
     sem_df["similarity_pct"] = distances_to_similarity_pct(dists[0])
     sem_df["vibes"] = sem_df["similarity_pct"].apply(label_vibes)
     sem_df["is_random"] = False
+
     rand_df = pd.DataFrame()
     if random_extra > 0:
-        chosen = np.random.choice(len(df_combined), size=min(random_extra, len(df_combined)), replace=False)
+        chosen = np.random.choice(
+            len(df_combined),
+            size=min(random_extra, len(df_combined)),
+            replace=False,
+        )
         rand_df = df_combined.iloc[chosen].copy()
         rand_df["similarity_pct"] = np.nan
         rand_df["vibes"] = "pure random"
         rand_df["is_random"] = True
+
     results = pd.concat([sem_df, rand_df], ignore_index=True)
     return results
 
-def lookup_spotify_track_smart(artist, song):
-    if not sp: return None, None
+
+def lookup_spotify_track_smart(artist: str, song: str):
+    if not sp:
+        return None, None
     q = f"track:{song} artist:{artist}"
     try:
         results = sp.search(q, type="track", limit=3)
-        if not results["tracks"]["items"]:
+        items = results.get("tracks", {}).get("items", [])
+        if not items:
             return None, None
-        best = max(results["tracks"]["items"],
-                   key=lambda t: SequenceMatcher(None, t["name"].lower(), song.lower()).ratio())
-        return best["external_urls"]["spotify"], best["album"]["images"][0]["url"]
-    except Exception:
+        best = max(
+            items,
+            key=lambda t: SequenceMatcher(None, t["name"].lower(), song.lower()).ratio(),
+        )
+        url = best["external_urls"]["spotify"]
+        images = best["album"]["images"]
+        img_url = images[0]["url"] if images else None
+        return url, img_url
+    except Exception as e:
+        print("⚠️ Spotify search failed:", repr(e))
         return None, None
 
-def search_pipeline(query, k=10, random_extra=0, use_llama=True):
+
+def search_pipeline(query: str, k: int = 10, random_extra: int = 0, use_llama: bool = True) -> pd.DataFrame:
     res = semantic_search(query, k, random_extra, use_llama)
-    if res.empty or not sp:
+    if res.empty or sp is None:
         res["spotify_url"], res["album_image"] = None, None
         return res
+
     urls, imgs = [], []
     for _, r in res.iterrows():
-        u, i = lookup_spotify_track_smart(r["artist"], r["song"])
-        urls.append(u); imgs.append(i)
+        u, i = lookup_spotify_track_smart(str(r["artist"]), str(r["song"]))
+        urls.append(u)
+        imgs.append(i)
     res["spotify_url"], res["album_image"] = urls, imgs
     return res
 
-# ---------- HTML builders ----------
-def make_bg_style_html(query=None):
-    base_top, base_bottom = "#1e293b", "#020617"
-    return f"<style>:root {{--hf-bg-top:{base_top};--hf-bg-bottom:{base_bottom};}}</style>"
-
-def results_to_lux_html(results, query):
-    ...
-# ---------- CSS ----------
-app_css = """
-@import url('https://fonts.googleapis.com/css2?family=Inter...
-...
-# ---------- Background palette + helper ----------
-
-BG_PALETTE = [
-    ("#1e293b", "#020617"),
-    ("#0f172a", "#020617"),
-    ("#0b1120", "#020617"),
-    ("#111827", "#020617"),
-    ("#1f2937", "#020617"),
-]
-
-def make_bg_style_html():
-    """Pick a gradient pair from the palette and emit a <style> that sets CSS vars."""
-    top, bottom = random.choice(BG_PALETTE)
-    return f"<style>:root {{ --hf-bg-top: {top}; --hf-bg-bottom: {bottom}; }}</style>"
-
-# ---------- Theming + helpers ----------
-
-def infer_theme(query: str):
-    q = (query or "").lower()
-    if any(w in q for w in ["night", "drive", "highway", "city", "neon"]):
-        return {"name": "Midnight Drive", "emoji": "🌃"}
-    if any(w in q for w in ["party", "dance", "club", "crowd", "festival"]):
-        return {"name": "Nightclub Neon", "emoji": "🎉"}
-    if any(w in q for w in ["shower", "bathroom", "mirror", "getting ready"]):
-        return {"name": "Mirror Concert", "emoji": "🚿"}
-    if any(w in q for w in ["dog", "pet", "cat", "bloopers"]):
-        return {"name": "Pet Bloopers", "emoji": "🐶"}
-    # default
-    return {"name": "", "emoji": "🎧"}
-
-# ---------- DataFrame -> HTML ----------
-
-def results_to_lux_html(results: pd.DataFrame, query: str) -> str:
-    if results is None or results.empty:
-        return """
-        <div id="lux-wrapper">
-          <div id="lux-header">
-            <div class="lux-subline">HarmoniFind • Semantic playlist</div>
-            <h1>🎧 Describe a vibe to start</h1>
-            <p style="font-size:0.9rem;color:rgba(156,163,175,0.95);margin-top:8px;">
-              Type a brief above, or click <strong>🎲</strong> for a fun prompt.
-            </p>
-          </div>
-        </div>
-        """
-
-    theme = infer_theme(query)
-    query_safe = html_lib.escape(query or "")
-    emoji = theme["emoji"]
-
-    cards_html = ""
-    tracks_plain = []
-
-    for _, row in results.iterrows():
-        raw_artist = str(row.get("artist", ""))
-        raw_song   = str(row.get("song", ""))
-
-        artist = html_lib.escape(raw_artist)
-        song   = html_lib.escape(raw_song)
-
-        # for clipboard list
-        tracks_plain.append(f"{raw_song} — {raw_artist}")
 
-        is_random = bool(row.get("is_random", False))
+def get_random_vibe() -> str:
+    topics = ["late-night drives", "dog bloopers", "breakups", "sunset beaches", "college nostalgia"]
+    perspectives = ["first-person", "third-person", "group", "inner monologue"]
+    tones = ["dreamy", "chaotic", "romantic", "melancholic"]
+    return f"Lyrics about {random.choice(topics)}, told in {random.choice(perspectives)}, {random.choice(tones)}."
 
-        sim_pct = row.get("similarity_pct", None)
-        if pd.isna(sim_pct) or is_random:
-            sim_display = "—"
-            score_bg = "rgba(148,163,184,0.2)"
-            vibes = "pure random"
-        else:
-            sim_display = f"{float(sim_pct):.1f}%"
-            score_bg = "rgba(34,197,94,0.14)"
-            vibes = html_lib.escape(str(row.get("vibes", "")))
-
-        url = row.get("spotify_url", None)
-        img = row.get("album_image", None)
-
-        if isinstance(img, str) and img:
-            cover = f'<div class="lux-cover"><img src="{html_lib.escape(img)}"></div>'
-        else:
-            cover = '<div class="lux-cover">♪</div>'
-
-        if isinstance(url, str) and url:
-            play_btn = f'<a class="lux-play-btn" href="{html_lib.escape(url)}" target="_blank">▶︎ Play on Spotify</a>'
-        else:
-            play_btn = ""
-
-        random_chip = ""
-        if is_random:
-            random_chip = '<span class="lux-chip">🎲 random pick</span>'
-
-        cards_html += f"""
-        <div class="lux-card">
-          {cover}
-          <div class="lux-main">
-            <div class="lux-title-row">
-              <div>
-                <div class="lux-title">{song}</div>
-                <div class="lux-artist">{artist}</div>
-              </div>
-              <div class="lux-score">
-                <div class="lux-score-badge" style="background:{score_bg};">{sim_display}</div>
-                <div class="lux-vibes">{vibes}</div>
-              </div>
-            </div>
-            <div class="lux-bottom-row">
-              {play_btn}
-              {random_chip}
-            </div>
-          </div>
-        </div>
-        """
-
-    # Build track list text for clipboard
-    header_line = f"HarmoniFind results for: {query or ''}".strip()
-    if not header_line:
-        header_line = "HarmoniFind results"
-    list_text = header_line + "\n\n" + "\n".join(tracks_plain)
-    # escape for JS string
-    js_text = (
-        list_text
-        .replace("\\", "\\\\")
-        .replace("'", "\\'")
-        .replace("\n", "\\n")
-    )
-
-    meta_html = f"""
-    <p>Semantic matches first, plus optional 🎲 discovery if you enabled it.</p>
-    <div class="lux-meta">
-      <span class="lux-badge">Tracks: {len(results)}</span>
-      <a class="lux-pill" href="javascript:void(0);" onclick="navigator.clipboard.writeText('{js_text}');">
-        🔗 Copy Your HarmoniFinds
-      </a>
-    </div>
-    """
-
-    html = f"""
-    <div id="lux-wrapper">
-      <div id="lux-header">
-        <div class="lux-subline">HarmoniFind • Semantic playlist</div>
-        <h1>{emoji} {query_safe or "Untitled vibe"}</h1>
-        {meta_html}
-      </div>
-      <div class="lux-playlist-wrapper">
-        {cards_html}
-      </div>
-    </div>
-    """
-    return html
-
-# ---------- Search + bg wrapper ----------
-
-def core_search_html(query, k, random_extra):
-    # LLaMA expansion always ON
-    results = search_pipeline(
-        query=query or "",
-        k=int(k),
-        random_extra=int(random_extra),
-        use_llama=True,
-    )
-    return results_to_lux_html(results, query or "")
-
-def search_with_bg(query, k, random_extra):
-    """Return playlist HTML + a new background style snippet."""
-    playlist_html = core_search_html(query, k, random_extra)
-    bg_style_html = make_bg_style_html()
-    return playlist_html, bg_style_html
-
-def surprise_brief():
-    return get_random_vibe()
-
-def clear_all():
-    # reset query, results (empty state), and bg
-    empty_html = results_to_lux_html(None, "")
-    return "", empty_html, make_bg_style_html()
 
 # ---------- CSS ----------
 
@@ -585,10 +444,187 @@ button.secondary-btn:hover {
 }
 """
 
+# ---------- Background palette + helper ----------
+
+BG_PALETTE = [
+    ("#1e293b", "#020617"),
+    ("#0f172a", "#020617"),
+    ("#0b1120", "#020617"),
+    ("#111827", "#020617"),
+    ("#1f2937", "#020617"),
+]
+
+def make_bg_style_html() -> str:
+    """Pick a gradient pair from the palette and emit a <style> that sets CSS vars."""
+    top, bottom = random.choice(BG_PALETTE)
+    return f"<style>:root {{ --hf-bg-top: {top}; --hf-bg-bottom: {bottom}; }}</style>"
+
+# ---------- Theming + helpers ----------
+
+def infer_theme(query: str):
+    q = (query or "").lower()
+    if any(w in q for w in ["night", "drive", "highway", "city", "neon"]):
+        return {"name": "Midnight Drive", "emoji": "🌃"}
+    if any(w in q for w in ["party", "dance", "club", "crowd", "festival"]):
+        return {"name": "Nightclub Neon", "emoji": "🎉"}
+    if any(w in q for w in ["shower", "bathroom", "mirror", "getting ready"]):
+        return {"name": "Mirror Concert", "emoji": "🚿"}
+    if any(w in q for w in ["dog", "pet", "cat", "bloopers"]):
+        return {"name": "Pet Bloopers", "emoji": "🐶"}
+    # default
+    return {"name": "", "emoji": "🎧"}
+
+# ---------- DataFrame -> HTML ----------
+
+def results_to_lux_html(results: pd.DataFrame, query: str) -> str:
+    if results is None or results.empty:
+        return """
+        <div id="lux-wrapper">
+          <div id="lux-header">
+            <div class="lux-subline">HarmoniFind • Semantic playlist</div>
+            <h1>🎧 Describe a vibe to start</h1>
+            <p style="font-size:0.9rem;color:rgba(156,163,175,0.95);margin-top:8px;">
+              Type a brief above, or click <strong>🎲</strong> for a fun prompt.
+            </p>
+          </div>
+        </div>
+        """
+
+    theme = infer_theme(query)
+    query_safe = html_lib.escape(query or "")
+    emoji = theme["emoji"]
+
+    cards_html = ""
+    tracks_plain = []
+
+    for _, row in results.iterrows():
+        raw_artist = str(row.get("artist", ""))
+        raw_song   = str(row.get("song", ""))
+
+        artist = html_lib.escape(raw_artist)
+        song   = html_lib.escape(raw_song)
+
+        # for clipboard list
+        tracks_plain.append(f"{raw_song} — {raw_artist}")
+
+        is_random = bool(row.get("is_random", False))
+
+        sim_pct = row.get("similarity_pct", None)
+        if pd.isna(sim_pct) or is_random:
+            sim_display = "—"
+            score_bg = "rgba(148,163,184,0.2)"
+            vibes = "pure random"
+        else:
+            sim_display = f"{float(sim_pct):.1f}%"
+            score_bg = "rgba(34,197,94,0.14)"
+            vibes = html_lib.escape(str(row.get("vibes", "")))
+
+        url = row.get("spotify_url", None)
+        img = row.get("album_image", None)
+
+        if isinstance(img, str) and img:
+            cover = f'<div class="lux-cover"><img src="{html_lib.escape(img)}"></div>'
+        else:
+            cover = '<div class="lux-cover">♪</div>'
+
+        if isinstance(url, str) and url:
+            play_btn = f'<a class="lux-play-btn" href="{html_lib.escape(url)}" target="_blank">▶︎ Play on Spotify</a>'
+        else:
+            play_btn = ""
+
+        random_chip = ""
+        if is_random:
+            random_chip = '<span class="lux-chip">🎲 random pick</span>'
+
+        cards_html += f"""
+        <div class="lux-card">
+          {cover}
+          <div class="lux-main">
+            <div class="lux-title-row">
+              <div>
+                <div class="lux-title">{song}</div>
+                <div class="lux-artist">{artist}</div>
+              </div>
+              <div class="lux-score">
+                <div class="lux-score-badge" style="background:{score_bg};">{sim_display}</div>
+                <div class="lux-vibes">{vibes}</div>
+              </div>
+            </div>
+            <div class="lux-bottom-row">
+              {play_btn}
+              {random_chip}
+            </div>
+          </div>
+        </div>
+        """
+
+    # Build track list text for clipboard
+    header_line = f"HarmoniFind results for: {query or ''}".strip()
+    if not header_line:
+        header_line = "HarmoniFind results"
+    list_text = header_line + "\n\n" + "\n".join(tracks_plain)
+    # escape for JS string
+    js_text = (
+        list_text
+        .replace("\\", "\\\\")
+        .replace("'", "\\'")
+        .replace("\n", "\\n")
+    )
+
+    meta_html = f"""
+    <p>Semantic matches first, plus optional 🎲 discovery if you enabled it.</p>
+    <div class="lux-meta">
+      <span class="lux-badge">Tracks: {len(results)}</span>
+      <a class="lux-pill" href="javascript:void(0);" onclick="navigator.clipboard.writeText('{js_text}');">
+        🔗 Copy Your HarmoniFinds
+      </a>
+    </div>
+    """
+
+    html = f"""
+    <div id="lux-wrapper">
+      <div id="lux-header">
+        <div class="lux-subline">HarmoniFind • Semantic playlist</div>
+        <h1>{emoji} {query_safe or "Untitled vibe"}</h1>
+        {meta_html}
+      </div>
+      <div class="lux-playlist-wrapper">
+        {cards_html}
+      </div>
+    </div>
+    """
+    return html
+
+# ---------- Search + bg wrapper ----------
+
+def core_search_html(query, k, random_extra):
+    # LLaMA expansion always ON now
+    results = search_pipeline(
+        query=query or "",
+        k=int(k),
+        random_extra=int(random_extra),
+        use_llama=True,
+    )
+    return results_to_lux_html(results, query or "")
+
+def search_with_bg(query, k, random_extra):
+    """Return playlist HTML + a new background style snippet."""
+    playlist_html = core_search_html(query, k, random_extra)
+    bg_style_html = make_bg_style_html()
+    return playlist_html, bg_style_html
+
+def surprise_brief():
+    return get_random_vibe()
+
+def clear_all():
+    # reset query, results (empty state), and bg
+    empty_html = results_to_lux_html(None, "")
+    return "", empty_html, make_bg_style_html()
+
 # ---------- Gradio UI ----------
 
 with gr.Blocks(title="HarmoniFind") as demo:
-    # inject CSS manually (HF version of Gradio may not support css=...)
+    # Inject CSS manually (HF Gradio version may not support css=... kwarg)
     gr.HTML(f"<style>{app_css}</style>")
 
     # dynamic bg style holder (updated on each search)
@@ -619,7 +655,7 @@ with gr.Blocks(title="HarmoniFind") as demo:
                 surprise_btn = gr.Button("🎲 Surprise me", elem_classes=["secondary-btn"])
                 clear_btn    = gr.Button("Clear", elem_classes=["secondary-btn"])
 
-        # Sliders only (LLaMA always on)
+        # Sliders only (LLaMA is always on; no checkbox)
         with gr.Accordion("Search settings", open=False):
             with gr.Row():
                 k_slider    = gr.Slider(5, 50, value=10, step=1, label="# semantic matches")
@@ -659,4 +695,3 @@ with gr.Blocks(title="HarmoniFind") as demo:
 if __name__ == "__main__":
     port = int(os.getenv("PORT", 7860))
     demo.launch(server_name="0.0.0.0", server_port=port)
-