Upload 3 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+Spotify_Youtube.csv filter=lfs diff=lfs merge=lfs -text

Spotify_Youtube.csv

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a08bda951ec1c62a2fbaeb5547141ba23dd7b16347f0e83bd043b1ba9cf168cd
+size 30782016

app.py

@@ -0,0 +1,144 @@
+import gradio as gr
+import pandas as pd
+import datetime
+
+# --- Load your dataset ---
+df_clean = pd.read_csv("your_dataset.csv")  # Replace with your actual filename
+audio_feature_cols = ["Valence", "Danceability", "Energy", "Tempo", "Instrumentalness"]  # Customize if needed
+
+# --- Global feedback memory ---
+feedback_df = pd.DataFrame(columns=["Query", "Title", "Artist", "Feedback"])
+
+# --- Dummy parsing logic (replace with your LLM function if needed) ---
+def parse_user_input(user_input):
+    return {
+        "mood": "happy",
+        "context": "",
+        "preference": "Likes",
+        "reference_song": "",
+        "genre": "",
+        "instrumental": "",
+        "tempo": "",
+        "artist": "",
+        "gender": "",
+        "limit": "5"
+    }
+
+# --- Dummy explanation generator ---
+def generate_explanation(user_input, row):
+    return f"{row['Title']} by {row['Artist']} fits your input: {user_input}"
+
+# --- Dummy recommender fallback ---
+def recommend_by_reference_song(parsed, df, feature_cols, llm, prompt):
+    return df.head(int(parsed["limit"]))  # Replace with actual similarity logic
+
+def recommend_from_dataframe(parsed, df):
+    mood = parsed["mood"].lower()
+    top_n = int(parsed.get("limit", "5"))
+    val_range = (0.3, 0.7)
+    dance_range = (0.3, 0.7)
+
+    if mood == "happy":
+        val_range = (0.6, 1.0)
+        dance_range = (0.6, 1.0)
+    elif mood == "sad":
+        val_range = (0.0, 0.4)
+        dance_range = (0.0, 0.4)
+
+    df_filtered = df[
+        (df["Valence"] >= val_range[0]) & (df["Valence"] <= val_range[1]) &
+        (df["Danceability"] >= dance_range[0]) & (df["Danceability"] <= dance_range[1])
+    ]
+
+    sort_col = parsed.get("preference", "Likes")
+    if sort_col not in df.columns:
+        sort_col = "Likes"
+
+    return df_filtered.sort_values(by=sort_col, ascending=False).head(top_n)
+
+# --- Feedback ---
+def submit_feedback(query, title, artist, feedback_type):
+    global feedback_df
+    new_feedback = pd.DataFrame([{
+        "Query": query,
+        "Title": title,
+        "Artist": artist,
+        "Feedback": feedback_type
+    }])
+    feedback_df = pd.concat([feedback_df, new_feedback], ignore_index=True)
+    return f"✅ Feedback recorded for {title}: {feedback_type}"
+
+def save_results(results_df):
+    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    path = f"recommendations_{timestamp}.csv"
+    results_df.to_csv(path, index=False)
+    return path
+
+def save_feedback():
+    timestamp = datetime.datetime.now().strftime("%Y%m%d_%H%M%S")
+    path = f"feedback_log_{timestamp}.csv"
+    feedback_df.to_csv(path, index=False)
+    return path
+
+# --- Main handler ---
+def handle_query(input_text):
+    parsed = parse_user_input(input_text)
+    if parsed["reference_song"].strip().lower() not in ["", "...", "none", "n/a"]:
+        results = recommend_by_reference_song(parsed, df_clean, audio_feature_cols, None, None)
+    else:
+        results = recommend_from_dataframe(parsed, df_clean)
+    
+    results = results.copy()
+    results["Explanation"] = results.apply(lambda row: generate_explanation(input_text, row), axis=1)
+    return results[["Title", "Artist", "Explanation"]], results
+
+# --- Gradio Interface ---
+with gr.Blocks() as demo:
+    gr.Markdown("## 🎧 Song Recommender with Feedback")
+
+    user_input = gr.Textbox(label="What kind of song are you looking for?")
+    btn_run = gr.Button("🔍 Recommend Songs")
+
+    song_output = gr.Dataframe(label="🎵 Recommended Songs")
+    results_state = gr.State()
+
+    download_btn = gr.Button("⬇️ Download Recommendations CSV", visible=False)
+    download_file = gr.File()
+    btn_run.click(fn=handle_query, inputs=user_input, outputs=[song_output, results_state])
+    btn_run.click(lambda: gr.update(visible=True), None, outputs=[download_btn])
+    download_btn.click(fn=save_results, inputs=[results_state], outputs=download_file)
+
+    gr.Markdown("### 🧠 Optional Feedback")
+    toggle_feedback_btn = gr.Button("✍️ Give Feedback")
+    feedback_group = gr.Group(visible=False)
+
+    with feedback_group:
+        with gr.Row():
+            feedback_title = gr.Textbox(label="Song Title")
+            feedback_artist = gr.Textbox(label="Artist")
+
+        with gr.Row():
+            btn_like = gr.Button("👍 Relevant")
+            btn_dislike = gr.Button("👎 Not Relevant")
+
+        feedback_response = gr.Textbox(label="Feedback Message")
+
+    toggle_feedback_btn.click(lambda: gr.update(visible=True), None, outputs=[feedback_group])
+
+    btn_like.click(
+        fn=submit_feedback,
+        inputs=[user_input, feedback_title, feedback_artist, gr.State("👍")],
+        outputs=feedback_response
+    )
+
+    btn_dislike.click(
+        fn=submit_feedback,
+        inputs=[user_input, feedback_title, feedback_artist, gr.State("👎")],
+        outputs=feedback_response
+    )
+
+    download_feedback_btn = gr.Button("⬇️ Download Feedback Log CSV", visible=True)
+    download_feedback_file = gr.File()
+    download_feedback_btn.click(fn=save_feedback, outputs=download_feedback_file)
+
+demo.launch()

requirements.txt

@@ -0,0 +1,7 @@
+gradio
+pandas
+transformers
+scikit-learn
+accelerate
+torch
+gender-guesser