# -*- coding: utf-8 -*- """ STEP105 Human Profile Engine 중개사별 고정 작성 습관 + 매물종류별 tail template 선택. """ import json import random import hashlib from pathlib import Path BASE_DIR = Path(__file__).resolve().parent PROFILE_DIR = BASE_DIR / "human_profiles" TEMPLATE_DIR = BASE_DIR / "tail_comment_templates" DEFAULT_PROFILE = { "profile_id": "default", "name": "기본형", "typing": { "min_delay_ms": 35, "max_delay_ms": 125, "pause_chance": 0.055, "long_pause_chance": 0.018, "typo_chance": 0.002 }, "review": { "short_review_min_ms": 5000, "short_review_max_ms": 15000, "before_publish_min_sec": 30, "before_publish_max_sec": 120 }, "tail": { "min_chars": 260, "max_chars": 720, "style_weights": { "friendly": 4, "checklist": 2, "question": 2, "concise": 2 } } } def _safe_json_load(path): try: return json.loads(Path(path).read_text(encoding="utf-8")) except Exception: return None def list_profiles(): if not PROFILE_DIR.exists(): return [] rows = [] for path in sorted(PROFILE_DIR.glob("profile_*.json")): data = _safe_json_load(path) if data: rows.append(data) return rows def stable_profile_index(realtor_id, profile_count): if profile_count <= 0: return 0 key = str(realtor_id or "0").encode("utf-8") digest = hashlib.sha256(key).hexdigest() return int(digest[:8], 16) % profile_count def get_human_profile(realtor_id=None, profile_id=None): profiles = list_profiles() if profile_id: for profile in profiles: if str(profile.get("profile_id")) == str(profile_id): return profile if profiles: idx = stable_profile_index(realtor_id, len(profiles)) return profiles[idx] return DEFAULT_PROFILE def get_typing_config(profile): profile = profile or DEFAULT_PROFILE cfg = dict(DEFAULT_PROFILE["typing"]) cfg.update(profile.get("typing") or {}) return cfg def get_review_config(profile): profile = profile or DEFAULT_PROFILE cfg = dict(DEFAULT_PROFILE["review"]) cfg.update(profile.get("review") or {}) return cfg def normalize_property_type(value): text = str(value or "").strip().lower() if text in ["apt", "apartment", "아파트"]: return "apartment" if text in ["villa", "빌라", "연립", "다세대"]: return "villa" if text in ["officetel", "오피스텔"]: return "officetel" if text in ["store", "상가", "상업시설", "사무실"]: return "store" if text in ["land", "토지", "대지", "임야", "전", "답"]: return "land" if text in ["building", "건물", "다가구", "단독", "상가주택"]: return "building" if "아파트" in text: return "apartment" if "오피스텔" in text: return "officetel" if any(x in text for x in ["빌라", "연립", "다세대"]): return "villa" if any(x in text for x in ["상가", "사무실"]): return "store" if any(x in text for x in ["토지", "대지", "임야"]): return "land" if any(x in text for x in ["다가구", "단독", "건물", "상가주택"]): return "building" return "default" def detect_property_type_from_context(draft=None, queue=None): draft = draft or {} queue = queue or {} candidates = [ draft.get("real_estate_type"), draft.get("real_estate_type_name"), draft.get("property_type"), draft.get("draft_title"), queue.get("real_estate_type"), queue.get("publish_title"), queue.get("article_no"), ] source = draft.get("source_json") if isinstance(source, dict): detail = source.get("detail") or {} if isinstance(detail, dict): candidates.extend([ detail.get("real_estate_type"), detail.get("real_estate_type_name"), detail.get("article_name"), detail.get("building_name"), ]) merged = " ".join([str(x or "") for x in candidates]) return normalize_property_type(merged) def _weighted_choice(weight_map): items = [] for key, weight in (weight_map or {}).items(): try: w = int(weight) except Exception: w = 1 items.extend([key] * max(1, w)) if not items: return "friendly" return random.choice(items) def load_tail_templates(property_type, style=None): property_type = normalize_property_type(property_type) style = str(style or "").strip() dirs = [ TEMPLATE_DIR / property_type, TEMPLATE_DIR / "default", ] files = [] for d in dirs: if not d.exists(): continue if style: files.extend(sorted(d.glob(f"*_{style}_*.txt"))) files.extend(sorted(d.glob(f"{style}_*.txt"))) files.extend(sorted(d.glob("*.txt"))) unique = [] seen = set() for f in files: if str(f) in seen: continue seen.add(str(f)) unique.append(f) return unique def render_tail_template(text, context=None): context = context or {} values = { "title": context.get("title") or "오늘 소개드린 매물", "office_name": context.get("office_name") or "중개사무소", "trade_type": context.get("trade_type") or "거래", "property_type": context.get("property_type") or "매물", "price": context.get("price") or "", "region": context.get("region") or "", } for key, value in values.items(): text = text.replace("{" + key + "}", str(value or "")) return text.strip() def build_profile_tail_comment(realtor_id=None, profile=None, property_type="default", context=None): profile = profile or get_human_profile(realtor_id=realtor_id) tail_cfg = dict(DEFAULT_PROFILE["tail"]) tail_cfg.update(profile.get("tail") or {}) style = _weighted_choice(tail_cfg.get("style_weights") or {}) files = load_tail_templates(property_type=property_type, style=style) if not files: return "" path = random.choice(files) text = path.read_text(encoding="utf-8", errors="replace").strip() text = render_tail_template(text, context=context or {}) min_chars = int(tail_cfg.get("min_chars") or 260) max_chars = int(tail_cfg.get("max_chars") or 720) if len(text) > max_chars: text = text[:max_chars].rsplit(" ", 1)[0].strip() if len(text) < min_chars: extra_files = load_tail_templates(property_type="default", style="concise") if extra_files: extra = random.choice(extra_files).read_text(encoding="utf-8", errors="replace").strip() extra = render_tail_template(extra, context=context or {}) text = (text + "\n\n" + extra).strip() return text def profile_summary(profile): profile = profile or DEFAULT_PROFILE return { "profile_id": profile.get("profile_id"), "name": profile.get("name"), "typing": get_typing_config(profile), "review": get_review_config(profile), "tail": profile.get("tail") or {}, }