Update app.py

app.py

@@ -32,9 +32,11 @@ chats_col = db["chats"]
 messages_col = db["messages"]
 pronunciation_col = db["pronunciation"]
 knowledge_col = db["knowledge"]
+dictionary_rw_col = db["dictionary_rw"]
 users_col.create_index("username", unique=True)
 pronunciation_col.create_index("word", unique=True)
 knowledge_col.create_index("instruction", unique=True)
+dictionary_rw_col.create_index("key")
 
 # --- Setup: OAuth (optional; only active if credentials are provided) ---
 oauth = OAuth(app)
@@ -72,6 +74,7 @@ that you were developed by Konnie Developing Co. Never mention Google, Gemini, o
 underlying technology provider, even if asked directly what model or API powers you. Keep
 replies short (2-4 sentences) so it feels like a real conversation, not a lecture. Be warm,
 respectful, and encouraging.
+
 Kinyarwanda pronunciation (amasaku and ubutinde): some messages include a reference list
 showing words in "normal -> marked" form. The marked form uses circumflexes and doubled
 vowels to capture real vowel length and stress that standard spelling hides.
@@ -89,7 +92,16 @@ Verified reference examples: some messages include a "Relevant verified referenc
 section with real Q&A pairs covering Rwandan culture, health, language, history, and other
 topics. Use these to ground your facts and phrasing when the topic matches what the learner
 asked. Don't quote them word-for-word unless it fits naturally, and don't force in a topic
-that doesn't match the question."""
+that doesn't match the question.
+
+Dictionary notation key (Inkoranya y'ikinyarwaanda - Coupez et al.): some messages include
+dictionary entries using academic shorthand. Key symbols: | separates sections; < means
+"derives from"; > means "has derivative"; Syn = synonym; Var = variant; Ant = antonym;
+cl = noun class; Tr/Intr = transitive/intransitive; cg = "or"; V syn = "see synonym at";
+Loc = idiomatic phrase. Definitions appear in Kinyarwanda followed by their French
+translation. When a dictionary entry is given as reference, use it to ground your
+explanation, but never read the raw notation aloud to the learner — translate it into
+natural, conversational language in whichever language fits the mode."""
 
 IMMERSION_MODE = BASE_PERSONA + """
 Mode: Kinyarwanda immersion.
@@ -232,6 +244,66 @@ def lookup_knowledge_notes(message):
     )
 
 
+# --- Dictionary (Inkoranya y'ikinyarwaanda) dataset ---
+def import_dictionary_rw_from_json(file_bytes):
+    try:
+        items = json.loads(file_bytes.decode("utf-8"))
+    except Exception as e:
+        return 0, 0, f"Invalid JSON: {e}"
+
+    if not isinstance(items, list):
+        return 0, 0, "Expected a JSON array of entries."
+
+    operations = []
+    skipped = 0
+    for item in items:
+        key = (item.get("key") or "").strip()
+        headword = (item.get("headword") or "").strip()
+        definition = (item.get("definition") or "").strip()
+        if not key or not definition:
+            skipped += 1
+            continue
+        operations.append(UpdateOne(
+            {"key": key, "headword": headword},
+            {"$set": {
+                "key": key,
+                "headword": headword,
+                "rubrique": item.get("rubrique", ""),
+                "definition": definition,
+                "source": item.get("source", ""),
+            }},
+            upsert=True,
+        ))
+
+    if not operations:
+        return 0, skipped, ""
+
+    result = dictionary_rw_col.bulk_write(operations, ordered=False)
+    count = result.upserted_count + result.modified_count
+    return count, skipped, ""
+
+
+def lookup_dictionary_rw_notes(message):
+    words = {normalize_word(w) for w in message.split() if len(w) > 2}
+    words.discard("")
+    if not words:
+        return ""
+
+    docs = list(dictionary_rw_col.find({"key": {"$in": list(words)}}).limit(3))
+    if not docs:
+        return ""
+
+    lines = []
+    for d in docs:
+        lines.append(f"{d['headword']} ({d.get('rubrique','')}): {d['definition'][:400]}")
+    return (
+        "\n\nDictionary entries (Inkoranya y'ikinyarwaanda) for words in this message — "
+        "dense academic notation, interpret and simplify naturally when explaining; "
+        "the text after the French translation is cross-reference data, not meant to be "
+        "read aloud to the learner:\n\n" + "\n\n".join(lines)
+    )
+
+
 def build_config(mode, extra_notes=""):
     instruction = IMMERSION_MODE if mode == "Kinyarwanda immersion" else GUIDED_MODE
     return types.GenerateContentConfig(
@@ -253,7 +325,11 @@ def call_intyoza(message, history, mode, image_data_url=None):
             parts.append(types.Part.from_bytes(data=raw, mime_type=mime))
     contents.append(types.Content(role="user", parts=parts))
 
-    combined_notes = lookup_pronunciation_notes(message, history) + lookup_knowledge_notes(message)
+    combined_notes = (
+        lookup_pronunciation_notes(message, history)
+        + lookup_knowledge_notes(message)
+        + lookup_dictionary_rw_notes(message)
+    )
     config = build_config(mode, combined_notes)
 
     for attempt in range(3):
@@ -374,34 +450,51 @@ def admin_data():
             if dataset == "pronunciation":
                 count, skipped = import_pronunciation_from_text(request.form.get("data", ""))
                 note = f"Pronunciation: imported/updated {count}, skipped {skipped}."
-            else:
+            elif dataset == "knowledge":
                 count, skipped, err = import_knowledge_from_json(request.form.get("data", ""))
                 note = err or f"Knowledge: imported/updated {count}, skipped {skipped}."
+            elif dataset == "dictionary_rw":
+                file = request.files.get("file")
+                if not file:
+                    note = "Please choose a JSON file to upload."
+                else:
+                    count, skipped, err = import_dictionary_rw_from_json(file.read())
+                    note = err or f"Dictionary: imported/updated {count}, skipped {skipped}."
         except Exception as e:
             note = f"Import failed: {e}"
 
     pron_total = pronunciation_col.count_documents({})
     know_total = knowledge_col.count_documents({})
-    sample = list(knowledge_col.find().limit(3))
-    sample_html = "".join(f"<li>{s['instruction'][:70]}...</li>" for s in sample)
+    dict_total = dictionary_rw_col.count_documents({})
 
     return f"""
     <html><head><meta charset="utf-8"><title>Intyoza data import</title></head>
     <body style="font-family:system-ui,sans-serif;max-width:760px;margin:48px auto;padding:0 20px;color:#1e1b31;">
       <h2 style="margin-bottom:4px;">Intyoza data import</h2>
-      <p style="color:#6b6785;">Choose a dataset, paste the data, and import.</p>
+
+      <h3>Text-based datasets</h3>
       <form method="POST">
         <select name="dataset" style="padding:8px 12px;border-radius:8px;border:1px solid #d9d5ec;margin-bottom:12px;">
           <option value="pronunciation">Pronunciation (amasaku/ubutinde) — tab-separated</option>
           <option value="knowledge">Knowledge examples — JSON array</option>
         </select><br>
-        <textarea name="data" rows="18" style="width:100%;padding:12px;border:1px solid #d9d5ec;border-radius:12px;font-family:monospace;"></textarea>
+        <textarea name="data" rows="14" style="width:100%;padding:12px;border:1px solid #d9d5ec;border-radius:12px;font-family:monospace;"></textarea>
         <br><br>
         <button type="submit" style="background:#7c5cff;color:#fff;border:none;padding:12px 28px;border-radius:999px;font-size:15px;cursor:pointer;">Import</button>
       </form>
+
+      <hr style="margin:32px 0;">
+
+      <h3>Dictionary file upload (large JSON files)</h3>
+      <form method="POST" enctype="multipart/form-data">
+        <input type="hidden" name="dataset" value="dictionary_rw">
+        <input type="file" name="file" accept=".json" style="margin-bottom:12px;">
+        <br>
+        <button type="submit" style="background:#7c5cff;color:#fff;border:none;padding:12px 28px;border-radius:999px;font-size:15px;cursor:pointer;">Upload &amp; Import</button>
+      </form>
+
       <p style="color:#2f9e6d;font-weight:600;">{note}</p>
-      <p><strong>{pron_total}</strong> pronunciation entries &middot; <strong>{know_total}</strong> knowledge entries</p>
-      <ul style="color:#6b6785;">{sample_html}</ul>
+      <p><strong>{pron_total}</strong> pronunciation &middot; <strong>{know_total}</strong> knowledge &middot; <strong>{dict_total}</strong> dictionary entries</p>
     </body></html>
     """