#!/usr/bin/env python3 """ 408考研学习 v7 - 思路核对版 先做后对：做习题→AI推理→思路核对→答案+思路都对才✅ 支持Word(.docx)和文本(.txt)两种格式 """ import os, json, time, urllib.request, re, subprocess from datetime import datetime, timedelta STATE_FILE = "/root/downloads/.study_state.json" SENT_FILE = "/root/downloads/.sent_questions.txt" BASE_DIR = "/root/downloads" LAST_STUDY_FILE = "/root/downloads/.last_study_time.txt" TAVILY_KEY = "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" ZHENTI_PDF = f"{BASE_DIR}/09-24高清无水印无错误408真题合集.pdf" MINIMAX_URL = "http://guizhouyun.site:2177/v1/chat/completions" MINIMAX_KEY = "sk-QgA0QbvP3lkJ9Q5R2bCeC92f0bE14f30B84812677aFb37Db" # 章节目录：(科目, 文件路径, 文件类型, ChTag, ChName, 习题关键词, 答案关键词) # 文件类型: "docx" 或 "txt" CHAPTERS = [ # 计算机网络(Word) ("计算机网络", f"{BASE_DIR}/王道计网.docx", "docx", "1.1", "计算机网络概述", "1.1.7", "1.1.8"), ("计算机网络", f"{BASE_DIR}/王道计网.docx", "docx", "1.2", "计算机网络体系结构", "1.2.4", "1.2.5"), ("计算机网络", f"{BASE_DIR}/王道计网.docx", "docx", "2.1", "物理层", "2.1.4", "2.1.5"), # 计算机组成原理(Text) ("计算机组成原理", f"{BASE_DIR}/王道计组.txt", "txt", "1.1", "计算机系统概述", "习题精选", "答案"), ] NOTE_FILES = { "计算机网络": f"{BASE_DIR}/计算机网络_笔记.md", "数据结构": f"{BASE_DIR}/数据结构_笔记.md", "操作系统": f"{BASE_DIR}/操作系统_笔记.md", "计算机组成原理":f"{BASE_DIR}/计算机组成原理_笔记.md", } WRONG_FILES = { "计算机网络": f"{BASE_DIR}/错题本_计算机网络.md", "计算机组成原理":f"{BASE_DIR}/错题本_计算机组成原理.md", "数据结构": f"{BASE_DIR}/错题本_数据结构.md", "操作系统": f"{BASE_DIR}/错题本_操作系统.md", } for f in list(NOTE_FILES.values()) + list(WRONG_FILES.values()): if not os.path.exists(f) or os.path.getsize(f) < 5: with open(f, "w", encoding="utf-8") as ff: ff.write(f"# {os.path.basename(f).replace('.md','')}\n\n") def load_state(): return json.load(open(STATE_FILE)) if os.path.exists(STATE_FILE) \ else {"ch_idx": 0, "learned": {}} def save_state(s): open(STATE_FILE,"w").write(json.dumps(s)) # ============== 文件读取 ============== def read_docx_para(path, start, end): from docx import Document doc = Document(path) return "\n".join([p.text.strip() for p in doc.paragraphs[start:end] if p.text.strip()]) def read_txt_chapters(path, kw_start, kw_end=None): """txt文件：搜索关键词范围""" with open(path, encoding="utf-8") as f: lines = f.readlines() paras = [l.rstrip("\n") for l in lines] # 找起始位置 start_idx = None for i, p in enumerate(paras): if kw_start in p and len(p) < 40: start_idx = i break if start_idx is None: return "" # 找结束位置 if kw_end: end_idx = None for i, p in enumerate(paras[start_idx+1:], start_idx+1): if kw_end in p and len(p) < 40: end_idx = i break end_idx = end_idx or len(paras) else: end_idx = start_idx + 200 return "\n".join(paras[start_idx:end_idx]) def find_para_indices_docx(path, kw): """docx: 找关键词所在段落索引""" from docx import Document doc = Document(path) for i, p in enumerate(doc.paragraphs): if kw in p.text and len(p.text) < 40: return i return None def find_para_indices_txt(path, kw): """txt: 找关键词所在行索引""" with open(path, encoding="utf-8") as f: lines = f.readlines() for i, l in enumerate(lines): if kw in l and len(l.strip()) < 40: return i return None # ============== MiniMax推理 ============== def call_minimax(prompt, max_tokens=400): payload = json.dumps({ "model": "MiniMax-M2.7", "messages": [{"role": "user", "content": prompt}], "max_tokens": max_tokens, "temperature": 0.3, }).encode() try: req = urllib.request.Request(MINIMAX_URL, data=payload, headers={"Authorization": f"Bearer {MINIMAX_KEY}", "Content-Type": "application/json"}) with urllib.request.urlopen(req, timeout=20) as resp: result = json.loads(resp.read()) msg = result["choices"][0]["message"] return msg.get("content","").strip() or msg.get("reasoning_content","").strip()[:300] except Exception as e: return f"[API错误: {e}]" def reason_answer(question, knowledge): prompt = f"""知识点：{knowledge[:500]} 题目：{question[:200]} 请直接给出答案和关键推导（一句话）。 格式：答案：[X] 推导：...""" return call_minimax(prompt, 300) def check_recent(): """1小时内不重复学同一章节""" if os.path.exists(LAST_STUDY_FILE): try: with open(LAST_STUDY_FILE) as f: content = f.read().strip() if content: last = datetime.fromisoformat(content) if datetime.now() - last < timedelta(hours=1): return True except (ValueError, OSError): pass return False def mark_studied(): with open(LAST_STUDY_FILE, "w") as f: f.write(datetime.now().isoformat()) def verify_reasoning(my_reasoning, ref_analysis, question, my_answer): import re # 提取参考答案中的关键短语 # 格式："01. B" 或 "答案：B" ref_ans_m = re.search(r'(?:^|\n)(?:\d+\.\s*)([A-D])', ref_analysis[:300]) ref_ans = ref_ans_m.group(1) if ref_ans_m else None # 我的答案是否正确 ans_ok = my_answer not in ['?', '', 'NONE'] and ref_ans and my_answer.upper().strip('.').replace(' ','') == ref_ans.upper() # 思路核对：提取参考答案前200字的核心关键词（去停用词） stop = {'的','是','在','和','了','有','与','为','用','对','以','也','就','这','那','都','而','但','或','及','等','当','将','可','能'} def get_words(text): return set(w for w in re.findall(r'[\u4e00-\u9fff]{2,6}', text) if w not in stop) my_words = get_words(my_reasoning) ref_words = get_words(ref_analysis[:300]) overlap = len(my_words & ref_words) total = len(my_words | ref_words) sim = overlap / total if total > 0 else 0 # 答案对+重叠词>=2个 → 思路正确 reason_ok = ans_ok and sim >= 0.1 and overlap >= 2 ans_sym = '✅' if ans_ok else '❌' reason_sym = '✅' if reason_ok else '❌' detail = '重叠' + str(overlap) + '词，相似度' + str(int(sim*100)) + '%' return '答案=' + ans_sym + ' 思路=' + reason_sym + ' | ' + detail # ============== 搜索 ============== def search_tavily(query): try: payload = json.dumps({"query": query, "search_depth": "advanced", "max_results": 2, "include_answer": True}).encode() req = urllib.request.Request("https://api.tavily.com/search", data=payload, headers={"Authorization": f"Bearer {TAVILY_KEY}", "Content-Type": "application/json"}) with urllib.request.urlopen(req, timeout=12) as resp: return json.loads(resp.read()) except: return {} def search_zhenti(keyword): try: txt = f"/tmp/zt{os.getpid()}.txt" subprocess.run(["pdftotext", "-f", "1", "-l", "40", "-search", keyword, ZHENTI_PDF, txt], capture_output=True, timeout=15) if os.path.exists(txt) and os.path.getsize(txt) > 50: with open(txt) as f: r = f.read() os.unlink(txt) return r[:2500] except: pass return "" # ============== 解析 ============== def split_questions(text): if not text: return [] parts = re.split(r'\n(?=\d{2}[.、])', text) result = [] for p in parts: p = p.strip() if len(p) > 20: result.append(p) return result def extract_ref_answers(text): # 匹配 "01. B" 或 "答案：B" 格式 ans = re.findall(r'(?:^|\n)(?:\d{2}[.、]\s*)([A-D][\.、]?)', text, re.I) return [a.replace('。','').replace('.','') for a in ans] def parse_my_answer(text): m = re.search(r'答案[：:\s]*\[?([A-D])\]?', text, re.I) if m: return m.group(1).upper() m = re.search(r'\b([A-D])\b', text[:50]) if m: return m.group(1).upper() return "?" # ============== 追加文件 ============== def append_note(subj, title, content): nf = NOTE_FILES.get(subj) ts = time.strftime("%Y-%m-%d %H:%M") with open(nf, "a", encoding="utf-8") as f: f.write(f"\n\n{'='*55}\n【{ts}】{subj} - {title}\n{'='*55}\n\n{content}") def append_wrong(subj, title, content): nf = WRONG_FILES.get(subj) if not nf: return with open(nf, "a", encoding="utf-8") as f: f.write(f"\n\n{'='*55}\n【{time.strftime('%Y-%m-%d %H:%M')}】{subj} - {title}\n{'='*55}\n\n{content}") # ============== 主逻辑 ============== def main(): # 防重复：1小时内不重复学同一章节 if check_recent(): print("[SKIP] 1小时内已学习过，跳过") return state = load_state() ch_idx = state.get("ch_idx", 0) if ch_idx >= len(CHAPTERS): ch_idx = 0 subj, fpath, ftype, ch_tag, ch_name, ex_kw, ans_kw = CHAPTERS[ch_idx] print(f"\n[{time.strftime('%H:%M')}] {subj} {ch_tag} {ch_name}") # 1. 读取知识点 if ftype == "docx": # 找知识点起始段落：ch_tag小节前一段 idx_k = find_para_indices_docx(fpath, ch_tag) idx_ex = find_para_indices_docx(fpath, ex_kw) if idx_k and idx_ex: knowledge = read_docx_para(fpath, idx_k, idx_ex) else: knowledge = "" # 读习题和答案 idx_ans = find_para_indices_docx(fpath, ans_kw) if idx_ex and idx_ans: exercises = read_docx_para(fpath, idx_ex, idx_ans) answers_raw = read_docx_para(fpath, idx_ans, idx_ans+300) else: exercises, answers_raw = "", "" else: # txt格式 knowledge = read_txt_chapters(fpath, ch_tag, ex_kw) exercises = read_txt_chapters(fpath, ex_kw, ans_kw) answers_raw = read_txt_chapters(fpath, ans_kw) print(f" [知识点] {len(knowledge)}字") print(f" [习题] {len(exercises)}字") # 2. 拆分题目 questions = split_questions(exercises) print(f" [题目] 拆分出{len(questions)}道") # 3. 真题PDF加1道 zhenti = search_zhenti(ch_name) zhenti_q = re.findall(r'(\d{2}[.、]\s*[\s\S]{20,300}?(?:A\.\s*[\s\S]{2,50}?){3})', zhenti) if zhenti_q: questions.append(f"[真题PDF] {zhenti_q[0]}") # 4. 解题技巧 tips_r = search_tavily(f"{subj} {ch_name} 解题技巧 方法") tips = "\n".join([r.get("content","")[:300] for r in tips_r.get("results",[]) if r.get("content")])[:500] # 5. 每道题：推理→思路核对 results = [] ref_answers = extract_ref_answers(answers_raw) for i, q_text in enumerate(questions[:4], 1): if len(q_text) < 15: continue print(f" [推理] 题目{i}...") my_text = reason_answer(q_text, knowledge) my_ans = parse_my_answer(my_text) print(f" [我的答案] {my_ans}") # 思路核对（用参考答案的前300字） ref_segment = answers_raw[:500] if answers_raw else "" print(f" [核对] 题目{i}思路...") verify = verify_reasoning(my_text, ref_segment, q_text, my_ans) print(f" [核对结果] {verify[:80]}") results.append((q_text, my_text, my_ans, verify)) print() # 6. 组装笔记 note = f"## 📖 知识点\n{knowledge[:2000]}\n\n" if tips: note += f"## 💡 解题技巧\n{tips}\n\n" for i, (q_text, my_text, my_ans, verify) in enumerate(results, 1): note += f"## 📝 题目{i}\n```\n{q_text[:500]}\n```\n\n" note += f"### 我的推理\n{my_text[:400]}\n\n" note += f"## 🔍 参考答案\n{answers_raw[:800]}\n\n" note += "## ✅ 核对结果\n" for i, (q_text, my_text, my_ans, verify) in enumerate(results, 1): ref_a = ref_answers[i-1].upper().replace('。','').replace('.','')[0] if i-1 < len(ref_answers) and len(ref_answers[i-1])>0 else "?" # 从verify里提结果 ans_ok = "✅" in verify[:50] reason_ok = "思路=✅" in verify note += f"- 题目{i}：答案={my_ans} {'✅' if ans_ok else '❌'} 思路={'✅' if reason_ok else '❌'}\n" note += f" 评委点评：{verify[:100]}\n" if not reason_ok: note += f" → 思路有问题，记入错题本\n" # 7. 写错题本 for i, (q_text, my_text, my_ans, verify) in enumerate(results, 1): ans_ok = "✅" in verify[:50] reason_ok = "思路=✅" in verify if not ans_ok or not reason_ok: wrong = f"## 题目{i}\n```\n{q_text[:400]}\n```\n## 我的推理\n{my_text[:300]}\n## 参考答案\n{answers_raw[:500]}\n## 评委点评\n{verify}\n## 正确思路\n（请根据参考答案补充）\n" append_wrong(subj, f"{ch_tag} {ch_name}", wrong) # 8. 保存 state["ch_idx"] = ch_idx + 1 state["learned"][subj] = state["learned"].get(subj, 0) + 1 save_state(state) append_note(subj, f"{ch_tag} {ch_name}", note) mark_studied() correct = sum(1 for _,_,_,v in results if "✅" in v[:50]) with open(f"{BASE_DIR}/.study_notify.txt", "w") as f: f.write(f"📚 **{subj}** {ch_tag}\n\n✅ 已学：{ch_name}\n📖 {len(knowledge)//100*100}字\n📝 {len(results)}道题\n✅ 思路全对：{correct}道\n💾 笔记已追加") print(f"\n完成！{len(results)}道题，{correct}道思路正确") if __name__ == "__main__": main()