Create-knowledge-find-decode/scripts/build_query_demo.py

"""四种打法生成 query 的 demo：只到「生成 query」，写 data/queries/demo.json（不落库）。

用法：PYTHONPATH=. python scripts/build_query_demo.py
③ 联想 + ④ LLM 走真实接口，须在能访问 crawler.aiddit.com / OpenRouter 的环境跑。
"""
from __future__ import annotations

import json
import time
from pathlib import Path

from acquisition.query import (
    ACTIONS, BLOCK, CARRIER_POS_GROUPED, DEMAND, KTYPE_SUFFIX, MODALITIES, STAGES,
    sample_nodes, tactic2_form_llm, tactic4_llm, tactic_multiaxis,
)
from core.config import Settings

ROOT = Path(__file__).resolve().parent.parent
OUT = ROOT / "data" / "queries" / "demo.json"


def main() -> None:
    settings = Settings.from_env()
    topics = sample_nodes("实质", depths=(3,), limit=12, under="理念")   # 实质：理念分支(事件/知识/情感…)
    forms = sample_nodes("形式", depths=(4,), limit=20, under="架构")    # 形式：架构分支(叙事/修辞/创意…)
    print(f"实质 {len(topics)}: {topics}")
    print(f"形式 {len(forms)}: {forms}")

    t2 = tactic2_form_llm(forms, settings)
    # ③ 暂留空：小红书/抖音无真·搜索联想接口（keyword_v2 返回的是帖子不是联想词），待接入真 sug 再补
    print(f"① LLM 实质: {topics[:5]}")
    t4 = tactic4_llm(topics[:5], settings)
    t5 = tactic_multiaxis(n=36)   # ④ 机械多轴组合（无 LLM）

    data = {
        "generated_at": int(time.time()),
        "block": BLOCK,
        "tactic1": {"name": "实质 × 创作阶段 × 需求", "需求点": DEMAND, "items": t4},
        "tactic2": {"name": "形式 + 载体位置", "载体位置": CARRIER_POS_GROUPED, "items": t2},
        "tactic3": {"name": "搜索词sug扩展", "items": [],
                    "待接入": "小红书/抖音都没有真正的「输入框搜索联想」接口；keyword_v2 名字像联想、实际返回的是帖子。此页暂空，待接入真 sug 接口后再补。"},
        "tactic4": {"name": "多轴正交组合", "items": t5,
                    "人工轴": {"阶段": STAGES, "动作": ACTIONS, "知识类型": KTYPE_SUFFIX, "模态": MODALITIES}},
    }
    OUT.parent.mkdir(parents=True, exist_ok=True)
    OUT.write_text(json.dumps(data, ensure_ascii=False, indent=1), encoding="utf-8")
    print(f"\n① {len(t4)} 条  ② {len(t2)} 条  ③ 待接入  ④ {len(t5)} 条(机械组合) → {OUT}")


if __name__ == "__main__":
    main()