From 275563252427555b0e19e839d8301e68460050f9 Mon Sep 17 00:00:00 2001
From: admin <tim.leikauf@gmail.com>
Date: Sun, 26 Apr 2026 10:27:12 +0200
Subject: [PATCH] Optimize generate_questions: word cache + gunicorn timeout
 300s

- Cache all words globally across levels (avoids 500+ Directus calls)
- Add Procfile with --timeout 300 to prevent gunicorn killing long requests

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 Procfile |  1 +
 app.py   | 75 +++++++++++++++++++++++++-------------------------------
 2 files changed, 35 insertions(+), 41 deletions(-)
 create mode 100644 Procfile

diff --git a/Procfile b/Procfile
new file mode 100644
index 0000000..e2d771e
--- /dev/null
+++ b/Procfile
@@ -0,0 +1 @@
+web: gunicorn app:app --bind 0.0.0.0:8000 --timeout 300 --workers 2
diff --git a/app.py b/app.py
index 1b7dd5c..dc9fb14 100644
--- a/app.py
+++ b/app.py
@@ -1029,7 +1029,7 @@ def generate_questions(obj_id: str):
     if not levels:
         return jsonify({"error": "No levels in AI response"}), 500
 
-    # Junction-Collections sicherstellen
+    # Junction-Collections sicherstellen (einmalig)
     for col, f1, f2 in [
         ("words_objects", "words_id", "objects_id"),
         ("questions_objects", "questions_id", "objects_id"),
@@ -1044,6 +1044,31 @@ def generate_questions(obj_id: str):
         "questions_linked": 0,
     }
 
+    # Alle eindeutigen Wörter aus allen Leveln vorab sammeln und einmalig laden
+    all_words_by_level: dict[str, int] = {}  # title_de → first level seen
+    for lvl in levels:
+        level = int(lvl.get("level") or 1)
+        for w in lvl.get("words", []) + lvl.get("distractor_words", []) + [lvl.get("short_answer", "")]:
+            w = str(w).strip()
+            if w and w not in all_words_by_level:
+                all_words_by_level[w] = level
+
+    # Wörter einmalig anlegen / finden (globaler Cache über alle Level)
+    global_word_map: dict[str, str] = {}  # title_de → id
+    for w, lvl_num in all_words_by_level.items():
+        try:
+            wid, is_new = _find_or_create_word(w, lvl_num, token)
+            global_word_map[w] = wid
+            _ensure_link(
+                "words_objects",
+                {"words_id": wid, "objects_id": obj_id},
+                {"words_id": wid, "objects_id": obj_id},
+                token,
+            )
+            stats["words_created" if is_new else "words_linked"] += 1
+        except Exception as e:
+            print(f"[generate_questions] word error '{w}': {e}")
+
     for lvl in levels:
         level = int(lvl.get("level") or 1)
         q_de = (lvl.get("question") or "").strip()
@@ -1055,39 +1080,7 @@ def generate_questions(obj_id: str):
         if not q_de:
             continue
 
-        # Kurzantwort-Wort
-        short_answer_id = None
-        if short_text:
-            try:
-                wid, is_new = _find_or_create_word(short_text, level, token)
-                short_answer_id = wid
-                _ensure_link(
-                    "words_objects",
-                    {"words_id": wid, "objects_id": obj_id},
-                    {"words_id": wid, "objects_id": obj_id},
-                    token,
-                )
-                stats["words_created" if is_new else "words_linked"] += 1
-            except Exception as e:
-                print(f"[generate_questions] short_answer error: {e}")
-
-        # Alle einzigartigen Wörter verarbeiten
-        all_unique = list(dict.fromkeys(words_list + distractor_list))
-        word_map: dict[str, str] = {}  # title_de → id
-
-        for w in all_unique:
-            try:
-                wid, is_new = _find_or_create_word(w, level, token)
-                word_map[w] = wid
-                _ensure_link(
-                    "words_objects",
-                    {"words_id": wid, "objects_id": obj_id},
-                    {"words_id": wid, "objects_id": obj_id},
-                    token,
-                )
-                stats["words_created" if is_new else "words_linked"] += 1
-            except Exception as e:
-                print(f"[generate_questions] word error '{w}': {e}")
+        short_answer_id = global_word_map.get(short_text) if short_text else None
 
         # Frage anlegen / verknüpfen
         try:
@@ -1106,23 +1099,23 @@ def generate_questions(obj_id: str):
             token,
         )
 
-        # related_words (bestehende Junction questions_words)
+        # related_words
         for w in words_list:
-            if w in word_map:
+            if w in global_word_map:
                 _ensure_link(
                     "questions_words",
-                    {"questions_id": q_id, "words_id": word_map[w]},
-                    {"questions_id": q_id, "words_id": word_map[w]},
+                    {"questions_id": q_id, "words_id": global_word_map[w]},
+                    {"questions_id": q_id, "words_id": global_word_map[w]},
                     token,
                 )
 
         # distractor_words
         for w in distractor_list:
-            if w in word_map:
+            if w in global_word_map:
                 _ensure_link(
                     "questions_distractor_words",
-                    {"questions_id": q_id, "words_id": word_map[w]},
-                    {"questions_id": q_id, "words_id": word_map[w]},
+                    {"questions_id": q_id, "words_id": global_word_map[w]},
+                    {"questions_id": q_id, "words_id": global_word_map[w]},
                     token,
                 )