testtest3

app.py

@@ -1,21 +1,27 @@
 import os
+import time
 import gradio as gr
 import pandas as pd
+import numpy as np
 import joblib
 import xgboost as xgb
-import freesound
+import requests
+from requests.adapters import HTTPAdapter
+from urllib3.util.retry import Retry
 
 
 # =========================
 # CONFIG
 # =========================
-API_TOKEN = "zE9NjEOgUMzH9K7mjiGBaPJiNwJLjSM53LevarRK"  # <-- remplace ici
+API_TOKEN = "A ECRIRE"  # <-- remplace ici
 
 MIN_EFFECT, MAX_EFFECT = 0.5, 3.0
 MIN_MUSIC, MAX_MUSIC = 10.0, 60.0
 
 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
 
+FREESOUND_API_BASE = "https://freesound.org/apiv2"
+
 
 # =========================
 # UI (CSS)
@@ -79,10 +85,6 @@ def html_result(badge_text, duration, rating_text, downloads_text, extra_html=""
 # INTERPRETATION
 # =========================
 def interpret_results(avg_class: int, dl_class: int) -> str:
-    """
-    avg_class: 0=Missed info, 1=Low, 2=Medium, 3=High (déduit du label texte)
-    dl_class: 0=Low, 1=Medium, 2=High (sortie num_downloads)
-    """
     if avg_class == 0:
         return (
             "ℹ️ <b>Interprétation</b> :<br>"
@@ -121,13 +123,9 @@ def interpret_results(avg_class: int, dl_class: int) -> str:
 
 
 def avg_label_to_class(avg_label: str) -> int:
-    """
-    Convertit label texte du label encoder en 0..3
-    """
     if avg_label is None:
         return 0
     s = str(avg_label).strip().lower()
-
     if "miss" in s or "missing" in s or "none" in s or "no" in s:
         return 0
     if "high" in s or "élev" in s or "eleve" in s:
@@ -140,23 +138,62 @@ def avg_label_to_class(avg_label: str) -> int:
 
 
 # =========================
-# FreeSound client
+# HTTP SESSION (retries)
 # =========================
-client = freesound.FreesoundClient()
-client.set_token(API_TOKEN, "token")
+def make_session():
+    session = requests.Session()
+    retry = Retry(
+        total=5,
+        backoff_factor=0.8,
+        status_forcelist=[429, 500, 502, 503, 504],
+        allowed_methods=["GET"],
+        raise_on_status=False,
+    )
+    adapter = HTTPAdapter(max_retries=retry)
+    session.mount("https://", adapter)
+    session.mount("http://", adapter)
+    return session
+
+
+SESSION = make_session()
+
+
+def fetch_sound_metadata_by_id(sound_id: int, fields: str) -> dict:
+    """
+    Appel API FreeSound directement (plus stable) + retries + timeout.
+    """
+    url = f"{FREESOUND_API_BASE}/search/text/"
+    headers = {"Authorization": f"Token {API_TOKEN}"}
+
+    params = {
+        "query": "",
+        "filter": f"id:{sound_id}",
+        "fields": fields,
+        "page_size": 1,
+    }
+
+    # timeout séparé (connect, read)
+    resp = SESSION.get(url, headers=headers, params=params, timeout=(6, 20))
+    if resp.status_code == 401:
+        raise RuntimeError("Token invalide ou non autorisé (401).")
+    if resp.status_code >= 400:
+        raise RuntimeError(f"Erreur HTTP {resp.status_code}: {resp.text[:200]}")
+
+    data = resp.json()
+    results = data.get("results", [])
+    if not results:
+        raise RuntimeError("Sound not found (aucun résultat pour cet ID).")
+    return results[0]
 
 
 # =========================
-# Charger les modèles (NOMS EXACTS DU REPO)
+# Charger modèles (NOMS EXACTS)
 # =========================
-
-# Music
 music_num_model = joblib.load(os.path.join(BASE_DIR, "music_model_num_downloads.joblib"))
 music_feat_list = joblib.load(os.path.join(BASE_DIR, "music_model_features_list.joblib"))
 music_avg_model = joblib.load(os.path.join(BASE_DIR, "music_xgb_avg_rating.joblib"))
 music_avg_le = joblib.load(os.path.join(BASE_DIR, "music_xgb_avg_rating_label_encoder.joblib"))
 
-# Effect sound
 effect_num_model = joblib.load(os.path.join(BASE_DIR, "effectSound_model_num_downloads.joblib"))
 effect_feat_list = joblib.load(os.path.join(BASE_DIR, "effect_model_features_list.joblib"))
 effect_avg_model = joblib.load(os.path.join(BASE_DIR, "effectSound_xgb_avg_rating.joblib"))
@@ -172,19 +209,29 @@ def safe_float(v):
         return 0.0
 
 
-def predict_with_model(model, features_dict, feat_list, le=None):
+def build_feature_df(sound: dict, feat_list: list) -> pd.DataFrame:
+    """
+    Tableau lisible des features utilisées (valeur API + NaN si absent).
+    """
+    rows = []
+    for col in feat_list:
+        val = sound.get(col, np.nan)
+        if val is None or isinstance(val, (list, dict)):
+            val = np.nan
+        rows.append({"feature": col, "value": val})
+    return pd.DataFrame(rows)
+
+
+def predict_with_model(model, sound: dict, feat_list: list, le=None):
     row = []
     for col in feat_list:
-        val = features_dict.get(col, 0)
+        val = sound.get(col, 0)
         if val is None or isinstance(val, (list, dict)):
             val = 0
         row.append(safe_float(val))
 
     X = pd.DataFrame([row], columns=feat_list)
-
-    # DMatrix (avec feature names)
     dm = xgb.DMatrix(X.values, feature_names=feat_list)
-
     pred_int = int(model.get_booster().predict(dm)[0])
 
     if le is not None:
@@ -194,44 +241,46 @@ def predict_with_model(model, features_dict, feat_list, le=None):
 
 def extract_and_predict(url: str):
     if not url or not url.strip():
-        return html_error("URL vide", "Collez une URL FreeSound du type <code>https://freesound.org/s/123456/</code>")
+        return html_error("URL vide", "Collez une URL FreeSound du type <code>https://freesound.org/s/123456/</code>"), pd.DataFrame()
 
     # Parse ID
     try:
         sound_id = int(url.rstrip("/").split("/")[-1])
     except Exception:
-        return html_error("URL invalide", "Impossible d'extraire l'ID depuis l'URL.")
+        return html_error("URL invalide", "Impossible d'extraire l'ID depuis l'URL."), pd.DataFrame()
 
-    # champs nécessaires (union music+effect)
-    all_features = list(set(music_feat_list + effect_feat_list))
+    # Fields nécessaires : union music/effect + duration
+    all_features = sorted(list(set(music_feat_list + effect_feat_list)))
     fields = "duration," + ",".join(all_features)
 
+    # Fetch API (avec retries)
     try:
-        results = client.search(query="", filter=f"id:{sound_id}", fields=fields)
+        sound = fetch_sound_metadata_by_id(sound_id, fields=fields)
     except Exception as e:
-        return html_error("Erreur API FreeSound", f"Détail : <code>{e}</code>")
-
-    if len(results.results) == 0:
-        return html_error("Son introuvable", "Aucun résultat pour cet ID.")
+        return html_error(
+            "Erreur API FreeSound",
+            f"Détail : <code>{e}</code><br><br>"
+            "Astuce : si ça arrive aléatoirement, c'est souvent un souci réseau/rate limit → réessayez."
+        ), pd.DataFrame()
 
-    sound = results.results[0]
     duration = safe_float(sound.get("duration", 0))
 
-    # Durées supportées
+    # Vérif durées
     if duration < MIN_EFFECT:
         return html_error(
             "Audio trop court",
             f"Durée : <b>{duration:.2f}s</b><br><br>"
             f"Plages : Effet sonore <b>{MIN_EFFECT}-{MAX_EFFECT}s</b> | Musique <b>{MIN_MUSIC}-{MAX_MUSIC}s</b>"
-        )
+        ), pd.DataFrame()
+
     if (MAX_EFFECT < duration < MIN_MUSIC) or duration > MAX_MUSIC:
         return html_error(
             "Audio hors plage",
             f"Durée : <b>{duration:.2f}s</b><br><br>"
             f"Plages : Effet sonore <b>{MIN_EFFECT}-{MAX_EFFECT}s</b> | Musique <b>{MIN_MUSIC}-{MAX_MUSIC}s</b>"
-        )
+        ), pd.DataFrame()
 
-    # Décision type
+    # Effect
     if MIN_EFFECT <= duration <= MAX_EFFECT:
         badge = "🔊 Effet sonore (metadata FreeSound)"
         dl_class = int(predict_with_model(effect_num_model, sound, effect_feat_list))
@@ -245,7 +294,8 @@ def extract_and_predict(url: str):
 <div class="hint">ID FreeSound : <b>{sound_id}</b></div>
 <div style="margin-top:12px; padding-top:10px; border-top:1px dashed #d1d5db">{conclusion}</div>
 """
-        return html_result(badge, duration, avg_text, dl_text, extra_html=extra)
+        df_feat = build_feature_df(sound, effect_feat_list)
+        return html_result(badge, duration, avg_text, dl_text, extra_html=extra), df_feat
 
     # Music
     badge = "🎵 Musique (metadata FreeSound)"
@@ -260,7 +310,8 @@ def extract_and_predict(url: str):
 <div class="hint">ID FreeSound : <b>{sound_id}</b></div>
 <div style="margin-top:12px; padding-top:10px; border-top:1px dashed #d1d5db">{conclusion}</div>
 """
-    return html_result(badge, duration, avg_text, dl_text, extra_html=extra)
+    df_feat = build_feature_df(sound, music_feat_list)
+    return html_result(badge, duration, avg_text, dl_text, extra_html=extra), df_feat
 
 
 # =========================
@@ -282,8 +333,11 @@ Collez une URL FreeSound. L'app récupère les <b>metadata</b> via l'API et pré
 
     url = gr.Textbox(label="URL FreeSound", placeholder="https://freesound.org/s/123456/")
     btn = gr.Button("🚀 Tester la prédiction", variant="primary")
-    out = gr.HTML()
 
-    btn.click(extract_and_predict, inputs=url, outputs=out)
+    with gr.Row():
+        out_html = gr.HTML(label="Résultat")
+        out_df = gr.Dataframe(label="Features utilisées (metadata)", interactive=False)
+
+    btn.click(extract_and_predict, inputs=url, outputs=[out_html, out_df])
 
 demo.launch(theme=theme)

requirements.txt

@@ -9,6 +9,14 @@ pydub
 opensmile
 requests
 pytz
+gradio
+pandas
+numpy
+joblib
+xgboost
+requests
+urllib3
+scikit-learn
 imblearn
 matplotlib
 git+https://github.com/MTG/freesound-python.git