feat: note scoring 60/6 with rating award + Minecraft theme

- Score max changed from 100 to 60, rating max from 10 to 6
- Note scoring now awards actual rating points (delta-based)
- Re-scoring only awards/deducts the difference
- Rating history shows note_score entries with problem link
- LLM prompt includes problem statement context for better evaluation
- LLM scoring dimensions: 题意理解/思路算法/代码记录/踩坑反思 (15 each)
- Minecraft-themed UI: 矿石鉴定, 探索笔记, 存入宝典, etc.
- Fallback scoring adjusted for 60-point scale
- Handle LLM markdown code fence wrapping in response

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

scripts/analyze_learning_note.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-"""Score a learning note (0-100) and map to rating (1-10) via LLM with fallback."""
+"""Score a learning note (0-60) and map to rating (0-6) via LLM with fallback."""
 
 from __future__ import annotations
 
@@ -27,28 +27,28 @@ def load_input(path: str) -> Dict[str, Any]:
 
 def fallback(note: str) -> Dict[str, Any]:
     n = note.strip()
-    score = 40
-    if len(n) >= 300:
-        score += 15
-    if len(n) >= 800:
+    score = 20
+    if len(n) >= 200:
         score += 10
+    if len(n) >= 500:
+        score += 5
     if "```" in n:
-        score += 15
+        score += 10
     if "踩坑" in n or "错误" in n or "debug" in n.lower():
-        score += 10
+        score += 8
     if "总结" in n or "注意" in n:
-        score += 10
-    score = min(100, score)
-    rating = max(1, min(10, round(score / 10)))
+        score += 7
+    score = min(60, score)
+    rating = max(0, min(6, round(score / 10)))
     feedback_md = (
-        "### 笔记评分（规则兜底）\n"
-        f"- 评分：**{score}/100**，评级：**{rating}/10**\n"
-        "\n### 你做得好的地方\n"
-        "- 记录了学习过程（已检测到一定的笔记内容）。\n"
-        "\n### 建议补充\n"
-        "- 写清：本节课**学习目标**、**关键概念**、**代码模板**。\n"
-        "- 至少写 1 个你遇到的坑（如输入输出格式、编译报错）以及解决方案。\n"
-        "- 最后用 3-5 行做总结，方便复习。\n"
+        "### ⛏️ 矿石鉴定报告（规则兜底）\n"
+        f"- 品质：**{score}/60** ⚡ 经验值：**+{rating}**\n"
+        "\n### 🏆 已获成就\n"
+        "- 记录了探索过程（检测到一定的笔记内容）。\n"
+        "\n### 📜 升级指南\n"
+        "- 写清本次**探索目标**、**核心知识点**、**代码配方**。\n"
+        "- 至少记录 1 个你遇到的陷阱（如格式、编译报错）以及修复方案。\n"
+        "- 最后用 3-5 行做总结，铸造你的知识宝典。\n"
     )
     return {"score": score, "rating": rating, "feedback_md": feedback_md, "model_name": "fallback-rules"}
 
@@ -60,23 +60,35 @@ def call_llm(payload: Dict[str, Any]) -> Dict[str, Any]:
     if not api_url:
         raise RuntimeError("missing OI_LLM_API_URL")
 
+    problem_title = payload.get("problem_title", "")
+    problem_statement = payload.get("problem_statement", "")
+    # Truncate long statements to save tokens
+    if len(problem_statement) > 2000:
+        problem_statement = problem_statement[:2000] + "\n...(truncated)"
+
     system = (
-        "你是一位面向小学生的C++竞赛教练，请对学习笔记打分。"
-        "评分满分100分，并给出10分制评级（rating=round(score/10)，范围1-10）。"
-        "评分维度：覆盖度30、正确性30、可操作性20、反思总结20。"
-        "输出必须是JSON，不要输出其他任何文字。"
+        "你是一位 Minecraft 风格的C++竞赛教练（矿石鉴定大师），请结合题目内容对学习笔记打分。\n"
+        "评分满分60分，经验值(rating)=round(score/10)，范围0-6。\n"
+        "评分维度：\n"
+        "- 题意理解 15分：是否正确理解题目要求\n"
+        "- 思路与算法 15分：解题思路是否清晰、算法是否正确\n"
+        "- 代码记录 15分：是否有代码片段/模板/关键实现\n"
+        "- 踩坑反思 15分：是否记录了坑点、调试过程、经验教训\n"
+        "请用 Minecraft 游戏风格的语言给出反馈，使用⛏️🏆📜💎⚡等图标。\n"
+        "输出必须是纯JSON（不要markdown代码块），不要输出其他任何文字。"
     )
     user = {
-        "task": "对学习笔记评分并给出改进建议",
+        "task": "结合题目对学习笔记评分并给出改进建议",
         "problem": {
             "id": payload.get("problem_id"),
-            "title": payload.get("problem_title"),
+            "title": problem_title,
+            "statement": problem_statement,
         },
         "note": payload.get("note", ""),
         "output_json_schema": {
-            "score": "integer 0-100",
-            "rating": "integer 1-10",
-            "feedback_md": "markdown string",
+            "score": "integer 0-60",
+            "rating": "integer 0-6",
+            "feedback_md": "markdown string, Minecraft style",
             "model_name": "string",
         },
     }
@@ -105,14 +117,21 @@ def call_llm(payload: Dict[str, Any]) -> Dict[str, Any]:
                 raise RuntimeError(f"HTTP {resp.status_code}")
             data = resp.json()
             content = data["choices"][0]["message"]["content"]
-            parsed = json.loads(content)
+            # Strip markdown code fence if present
+            c = content.strip()
+            if c.startswith("```"):
+                c = c.split("\n", 1)[-1]
+                if c.endswith("```"):
+                    c = c[:-3]
+                c = c.strip()
+            parsed = json.loads(c)
             if not isinstance(parsed, dict):
                 raise ValueError("model output not object")
             score = int(parsed.get("score", 0))
-            score = max(0, min(100, score))
+            score = max(0, min(60, score))
             rating = int(parsed.get("rating", round(score / 10)))
-            rating = max(1, min(10, rating))
-            feedback_md = str(parsed.get("feedback_md", "")).strip() or "### 笔记评分\n- 请补充更多内容（学习目标/代码/总结）。"
+            rating = max(0, min(6, rating))
+            feedback_md = str(parsed.get("feedback_md", "")).strip() or "### ⛏️ 矿石鉴定报告\n- 请补充更多内容（探索目标/代码配方/总结）。"
             model_name = str(parsed.get("model_name", model)).strip() or model
             return {"score": score, "rating": rating, "feedback_md": feedback_md, "model_name": model_name}
         except Exception as e:  # noqa: BLE001
@@ -132,7 +151,7 @@ def main() -> int:
     if not note.strip():
         print(
             json.dumps(
-                {"score": 0, "rating": 1, "feedback_md": "### 笔记为空\n请先写笔记再评分。", "model_name": "validator"},
+                {"score": 0, "rating": 0, "feedback_md": "### ⛏️ 空白卷轴\n请先写下你的探索笔记再进行鉴定。", "model_name": "validator"},
                 ensure_ascii=False,
             )
         )