#!/usr/bin/env python3 """Build full Markdown pages and keyword routes for the thinking-model library.""" from __future__ import annotations import csv import json import re from pathlib import Path ROOT = Path(__file__).resolve().parents[3] WORK_DIR = ROOT / "work" / "2026-05-29-thinking-models-knowledge-base" STABLE_DIR = ROOT / "standards-stack" / "llm-wiki" / "thinking-models" MODELS_DIR = STABLE_DIR / "models" CATALOG_PATH = STABLE_DIR / "thinking-model-catalog.csv" ROUTER_PATH = STABLE_DIR / "thinking-model-router.csv" SLIDES_PATH = WORK_DIR / "extracted" / "slides.json" ROUTE_MODEL_ALIASES = { "PDCA": "PDCA原则&经验积累型学习", "MECE": "MECE分析法", "5Why": "5Why思考法", "5W2H": "5W2H问题分析法", "SWOT": "SWOT分析", "3C": "3C分析", "PEST": "PEST分析模型", "五力分析": "五力分析模型", "波特三战略": "波特的3个基本战略", "HOOK": "上瘾（HOOK）模型", "MVP": "最小可行性产品（MVP）", "情绪ABC": "情绪ABC理论", "GROW": "GROW教练模型", "麦肯锡7S": "麦肯锡7S模型", "团队绩效模型": "德雷克斯勒-西贝特团队绩效模型", } def slugify(value: str) -> str: value = re.sub(r"[\\/:*?\"<>|]+", "-", value) value = re.sub(r"\s+", "-", value.strip()) value = value.strip("-") return value or "model" def read_catalog() -> list[dict[str, str]]: with CATALOG_PATH.open(newline="", encoding="utf-8") as fh: return list(csv.DictReader(fh)) def read_router() -> list[dict[str, str]]: with ROUTER_PATH.open(newline="", encoding="utf-8") as fh: return list(csv.DictReader(fh)) def read_slides() -> dict[int, dict[str, object]]: data = json.loads(SLIDES_PATH.read_text(encoding="utf-8")) return {int(item["slide_no"]): item for item in data} def clean_paragraphs(paragraphs: list[str]) -> list[str]: cleaned: list[str] = [] previous = None for item in paragraphs: text = re.sub(r"\s+", " ", item).strip() if not text: continue if text == previous: continue cleaned.append(text) previous = text return cleaned def route_keywords(row: dict[str, str]) -> str: parts = [ row["model"], row["scenario"], row["when_to_use"], row["expected_output"], row["answer_shape"], ] text = " ".join(parts) tokens = [] for token in re.split(r"[\s,，;；、/()（）&]+", text): token = token.strip("：:。.") if len(token) >= 2 and token not in tokens: tokens.append(token) return ";".join(tokens[:18]) def model_markdown(row: dict[str, str], slide: dict[str, object]) -> str: paragraphs = clean_paragraphs(slide.get("paragraphs", [])) # type: ignore[arg-type] lines = [ f"# {row['model_id']} {row['model']}", "", "## 元数据", "", f"- model_id: `{row['model_id']}`", f"- part: `{row['part']}`", f"- source_slide: `{row['slide_no']}`", f"- scenario: `{row['scenario']}`", f"- when_to_use: {row['when_to_use']}", f"- expected_output: {row['expected_output']}", f"- answer_shape: {row['answer_shape']}", f"- source_confidence: {row['source_confidence']}", "", "## 问答词路", "", f"- route_keywords: {route_keywords(row)}", f"- prompt_path: 先确认问题是否属于 `{row['scenario']}`，再用 `{row['model']}` 产出 `{row['expected_output']}`。", f"- answer_contract: {row['answer_shape']}", "", "## PPT 原文抽取", "", ] if paragraphs: lines.extend(f"- {item}" for item in paragraphs) else: lines.append("- no text extracted") lines.append("") return "\n".join(lines) def write_model_pages(catalog: list[dict[str, str]], slides: dict[int, dict[str, object]]) -> list[dict[str, str]]: MODELS_DIR.mkdir(parents=True, exist_ok=True) index_rows = [] for row in catalog: slide_no = int(row["slide_no"]) slide = slides[slide_no] filename = f"{row['model_id']}-{slugify(row['model'])}.md" path = MODELS_DIR / filename path.write_text(model_markdown(row, slide), encoding="utf-8") index_rows.append({**row, "markdown_path": f"models/{filename}"}) return index_rows def write_full_markdown(index_rows: list[dict[str, str]], slides: dict[int, dict[str, object]]) -> None: lines = [ "# 150 个思维模型全量 Markdown", "", "本文件是 `150个思维模型&麦肯锡战略思维.pptx` 的稳定 Markdown 真源。它保留每个模型对应 PPT 页的文本抽取，并补充问答路由字段，供 Codex 后续按模型回答问题。", "", "## 使用方式", "", "- 查单个模型：进入 `models/.md`。", "- 查场景路由：看 `thinking-model-question-routes.md`。", "- 做问答：先读 `thinking-model-keyword-routes.csv` 或 `thinking-model-router.csv` 选模型，再读对应模型 Markdown 的 PPT 原文。", "", "## 全量模型目录", "", ] for row in index_rows: lines.append(f"- [{row['model_id']} {row['model']}]({row['markdown_path']}) - {row['part']} / {row['scenario']}") lines.append("") lines.append("## 全量正文") lines.append("") for row in index_rows: slide = slides[int(row["slide_no"])] lines.append(model_markdown(row, slide)) lines.append("---") lines.append("") (STABLE_DIR / "full-content.md").write_text("\n".join(lines), encoding="utf-8") def write_model_index(index_rows: list[dict[str, str]]) -> None: fields = [ "model_id", "slide_no", "part", "model", "scenario", "when_to_use", "expected_output", "answer_shape", "markdown_path", "source_confidence", ] with (STABLE_DIR / "thinking-model-index.csv").open("w", encoding="utf-8", newline="") as fh: writer = csv.DictWriter(fh, fieldnames=fields) writer.writeheader() writer.writerows(index_rows) def write_keyword_routes(index_rows: list[dict[str, str]]) -> None: fields = [ "model_id", "model", "part", "scenario", "route_keywords", "question_triggers", "prompt_path", "expected_output", "markdown_path", ] rows = [] for row in index_rows: rows.append( { "model_id": row["model_id"], "model": row["model"], "part": row["part"], "scenario": row["scenario"], "route_keywords": route_keywords(row), "question_triggers": row["when_to_use"], "prompt_path": f"用 `{row['model']}` 先判断适用场景，再按 `{row['answer_shape']}` 输出。", "expected_output": row["expected_output"], "markdown_path": row["markdown_path"], } ) with (STABLE_DIR / "thinking-model-keyword-routes.csv").open("w", encoding="utf-8", newline="") as fh: writer = csv.DictWriter(fh, fieldnames=fields) writer.writeheader() writer.writerows(rows) def match_route_model(name: str, by_name: dict[str, dict[str, str]]) -> dict[str, str] | None: if name in by_name: return by_name[name] alias = ROUTE_MODEL_ALIASES.get(name) if alias and alias in by_name: return by_name[alias] for model, row in by_name.items(): if name == model or name in model or model in name: return row return None def write_question_routes(router: list[dict[str, str]], index_rows: list[dict[str, str]]) -> None: by_name = {row["model"]: row for row in index_rows} lines = [ "# 思维模型问答词路", "", "本文件把用户问题映射到模型组合。回答时不要只报模型名，要进入对应 `models/*.md` 或 `full-content.md` 读取 PPT 原文，再用模型推导结论。", "", "## 总规则", "", "1. 先用用户问题中的动词和对象判断场景。", "2. 普通问题选 1-3 个模型；复杂问题最多 3-5 个。", "3. 模型提供结构，不替代业务证据。", "4. 输出必须包含：选用模型、为什么选、模型推导、结论/动作、待验证项。", "", ] for route in router: lines.append(f"## {route['scenario']}") lines.append("") lines.append(f"- target_result: {route['target_result']}") lines.append(f"- preferred_models: {route['preferred_models']}") lines.append("- model_paths:") for name in route["preferred_models"].split(";"): name = name.strip() matched = match_route_model(name, by_name) if matched: lines.append(f" - `{matched['model']}` -> `{matched['markdown_path']}`") else: lines.append(f" - `{name}` -> check `thinking-model-keyword-routes.csv`") lines.append("") (STABLE_DIR / "thinking-model-question-routes.md").write_text("\n".join(lines), encoding="utf-8") def main() -> int: catalog = read_catalog() router = read_router() slides = read_slides() index_rows = write_model_pages(catalog, slides) write_full_markdown(index_rows, slides) write_model_index(index_rows) write_keyword_routes(index_rows) write_question_routes(router, index_rows) print(json.dumps({"models": len(index_rows), "stable_dir": str(STABLE_DIR)}, ensure_ascii=False)) return 0 if __name__ == "__main__": raise SystemExit(main())