Update conversation_logic.py

conversation_logic.py

@@ -266,12 +266,15 @@ def _is_direct_solve_request(text: str, intent: str) -> bool:
 def should_retrieve(intent: str, solved: bool, raw_user_text: str, category: Optional[str] = None) -> bool:
     normalized_category = normalize_category(category)
 
+    if intent in {"walkthrough", "step_by_step", "method", "explain", "concept", "definition", "instruction"}:
+        return True
+
+    if intent == "hint":
+        return solved is False or normalized_category in {"Quantitative", "Verbal", "DataInsight"}
+
     if _is_direct_solve_request(raw_user_text, intent):
         return (not solved) and normalized_category in {"Verbal", "DataInsight"}
 
-    if intent in RETRIEVAL_ALLOWED_INTENTS:
-        return True
-
     if not solved and normalized_category in {"Verbal", "DataInsight"}:
         return True
 
@@ -526,33 +529,88 @@ def _is_bad_generated_reply(text: str, user_text: str = "") -> bool:
 
     return False
 
+def _clean_teaching_text(text: str) -> str:
+    text = normalize_spaces((text or "").replace("\n", " ").strip())
+    text = re.sub(r"^[\-\•\*\d\.\)\s]+", "", text)
+    if len(text) > 160:
+        text = text[:157].rstrip() + "..."
+    return text
+
+
 def _pick_teaching_line(
     chunks: List[RetrievedChunk],
     current_reply: str,
+    question_text: str,
 ) -> Optional[str]:
     if not chunks:
         return None
 
     reply_keywords = _extract_keywords(current_reply)
+    question_keywords = _extract_keywords(question_text)
     best_line = None
-    best_score = -1
+    best_score = float("-inf")
+
+    principle_markers = {
+        "remember", "key", "idea", "rule", "method", "approach", "strategy",
+        "useful", "helps", "means", "convert", "rewrite", "isolate", "check",
+    }
 
     for chunk in chunks:
-        text = (chunk.text or "").strip().replace("\n", " ")
+        raw_text = (chunk.text or "").strip()
+        if not raw_text:
+            continue
+
+        text = _clean_teaching_text(raw_text)
         if not text:
             continue
-        if len(text) > 180:
-            text = text[:177].rstrip() + "..."
 
-        chunk_keywords = _extract_keywords(text)
-        novelty = len(chunk_keywords - reply_keywords)
-        overlap = len(chunk_keywords & reply_keywords)
+        lower_text = text.lower()
+        chunk_keywords = _extract_keywords(lower_text)
+
+        novelty_vs_reply = len(chunk_keywords - reply_keywords)
+        novelty_vs_question = len(chunk_keywords - question_keywords)
+        overlap_with_reply = len(chunk_keywords & reply_keywords)
+        overlap_with_question = len(chunk_keywords & question_keywords)
+
+        principle_bonus = 0.0
+        if any(marker in lower_text for marker in principle_markers):
+            principle_bonus += 1.5
+
+        if any(
+            phrase in lower_text
+            for phrase in [
+                "convert percent to decimal",
+                "undo operations in reverse order",
+                "set up an equation",
+                "part-whole relationship",
+                "isolate the variable",
+            ]
+        ):
+            principle_bonus += 2.0
+
+        too_similar_penalty = 0.0
+        if overlap_with_reply >= max(3, novelty_vs_reply + 1):
+            too_similar_penalty += 2.0
+
+        score = (
+            1.8 * novelty_vs_reply
+            + 1.0 * novelty_vs_question
+            + principle_bonus
+            - 0.5 * overlap_with_question
+            - 0.8 * overlap_with_reply
+            - too_similar_penalty
+        )
+
+        if len(text.split()) < 5:
+            score -= 2.0
 
-        score = novelty - 0.3 * overlap
         if score > best_score:
             best_score = score
             best_line = text
 
+    if best_score < 1.0:
+        return None
+
     return best_line
 
 class ConversationEngine:
@@ -676,10 +734,14 @@ class ConversationEngine:
                 result.used_retrieval = True
                 result.teaching_chunks = filtered
 
-        if selected_chunks and resolved_help_mode in {"walkthrough", "step_by_step", "method", "explain", "concept"}:
-            teaching_line = _pick_teaching_line(selected_chunks, reply)
-            if teaching_line:
-                reply = f"{reply}\n\nKey idea: {teaching_line}"
+            if selected_chunks and resolved_help_mode in {"walkthrough", "step_by_step", "method", "explain", "concept"}:
+                teaching_line = _pick_teaching_line(
+                    chunks=selected_chunks,
+                    current_reply=reply,
+                    question_text=solver_input,
+            )
+                if teaching_line:
+                    reply = f"{reply}\n\nKey idea: {teaching_line}"
 
         should_try_generator = (
             self.generator is not None