from __future__ import annotations import json import shutil from copy import deepcopy from pathlib import Path import psycopg from psycopg.rows import dict_row from psycopg.types.json import Jsonb from app.config import settings from common_paths import PROJECT_ROOT, TRAVEL_KG_EXPORT_ROOT PROJECT_ID = "travel_agency_2_0_test" TENANT_ID = "travel_agency" GRAPH_NAME = "travel_agency_2_0_test" NAMESPACE = "travel_agency_2_0" SCHEMA_DIR = PROJECT_ROOT / "schema搭建/travel_agency_2_0_test" DOWNLOAD_DIR = TRAVEL_KG_EXPORT_ROOT / "travel_agency_2_0_test_旅行社2.0测试" CURRENT_JSON = SCHEMA_DIR / "travel_agency_2_0_schema.current.json" def add_fields(entity: dict, fields: list[str], after: str | None = None) -> None: current = list(entity.get("fields") or []) insert_at = len(current) if after and after in current: insert_at = current.index(after) + 1 for field in fields: if field not in current: current.insert(insert_at, field) insert_at += 1 entity["fields"] = current def schema_to_dsl(payload: dict) -> str: lines: list[str] = [ "```text", f"namespace {payload.get('namespace', NAMESPACE)}", f"version {payload.get('version', '')}", "", "// 2.4：面向客服快速问答，保留固定线路/nearby 资源主干，新增轻量 BusinessRule 规则层。", ] for name, meta in payload.get("entity_types", {}).items(): lines.append("") lines.append(f"entity {name} // {meta.get('definition', '')}") if meta.get("primary_key"): lines.append(f" primary_key {meta['primary_key']}") for field in meta.get("fields", []): lines.append(f" property {field}: Text") for name, meta in payload.get("relation_types", {}).items(): lines.append("") lines.append( f"relation {name}: {meta.get('from', '')} -> {meta.get('to', '')} // {meta.get('definition', '')}" ) for prop in meta.get("properties", []): lines.append(f" property {prop}: Text") lines.extend(["", "// design_principles"]) for item in payload.get("design_principles", []): lines.append(f"- {item}") lines.extend(["", "// revision_notes"]) for item in payload.get("revision_notes", []): lines.append(f"- {item}") lines.extend(["", "// query_recipes"]) for name, query in payload.get("query_recipes", {}).items(): lines.append(f"{name}:") lines.append(query) lines.append("```") return "\n".join(lines) + "\n" def build_schema() -> dict: schema = deepcopy(json.loads(CURRENT_JSON.read_text(encoding="utf-8"))) schema["version"] = "2.4" schema["display_name"] = "旅行社2.4客服固定线路资源图谱 Schema" schema["purpose"] = ( "面向旅行社客服与小包团销售：固定路线像公交线路一样可查；RouteStop 串联景区；" "Hotel/Restaurant 作为 POI 通过 nearby 推荐；门票、小交通、保险、车辆等非 POI 放 TravelItem；" "客服高频的退改、付款、成团、老人儿童、车型和费用包含通过轻量 BusinessRule 查询。" ) enums = schema.setdefault("enums", {}) enums["BusinessRule.rule_type"] = [ "InquiryIntake", "ProductRecommendation", "Payment", "Refund", "GroupSize", "VehicleCapacity", "HotelPolicy", "MealPolicy", "TicketDiscount", "ScenicFee", "ElderChildCare", "ShoppingPromise", "ServicePromise", "RiskNotice", "SupplierConfirmation", "Other", ] enums["BusinessRule.scope"] = ["Global", "Product", "ScenicAttraction", "TravelItem", "Hotel", "Restaurant"] entity_types = schema.setdefault("entity_types", {}) add_fields( entity_types["TourProduct"], [ "sales_priority", "supported_customer_scenarios", "group_size_rule_text", "deposit_policy", "payment_policy", "refund_policy", "meal_policy", "shopping_policy", "service_promise", "elder_child_policy", "quote_required_fields", ], after="group_size_max", ) add_fields( entity_types["ProductDay"], [ "breakfast_included", "lunch_included", "dinner_included", "meal_status", "hotel_area", "hotel_grade_default", "customer_day_summary", ], after="meal_plan", ) add_fields( entity_types["RouteStop"], [ "customer_visible", "can_replace_resource", "walk_note", "elderly_note", "child_note", "expense_summary", ], after="route_display_text", ) add_fields( entity_types["ScenicAttraction"], [ "must_pay_items_summary", "optional_items_summary", "reservation_required", "id_card_required", ], after="free_ticket_policy", ) add_fields( entity_types["Hotel"], [ "has_elevator", "low_floor_available", "family_room_available", "room_count_policy", "booking_note", ], after="breakfast_policy", ) add_fields( entity_types["Restaurant"], [ "meal_standard_text", "suitable_group_size", "booking_note", "classification_status", ], after="meal_type", ) add_fields( entity_types["TravelItem"], [ "customer_visible", "item_scope", "default_status_hint", "status_source", "dedupe_key", "canonical_item_id", ], after="normalized_type_reason", ) entity_types["BusinessRule"] = { "definition": "客服高频业务规则与话术依据。只承载可复用规则，不把客户单次选择写进图谱。", "primary_key": "rule_id", "fields": [ "rule_id", "rule_type", "scope", "target_type", "target_id", "title", "content", "customer_reply", "trigger_keywords", "applies_to_days", "applies_to_group_type", "priority", "source_file", "source_text", "data_quality", "updated_at", ], } relation_types = schema.setdefault("relation_types", {}) relation_types["RULE_APPLIES_TO"] = { "from": "BusinessRule", "to": "TourProduct|ScenicAttraction|TravelItem|Hotel|Restaurant", "definition": "业务规则适用到产品、景区、费用项、酒店或餐厅。全局规则可只保留 target_type/target_id 属性，不强制连边。", "properties": ["target_type", "target_id", "scope", "priority", "source_file", "evidence_level"], } schema["design_principles"] = [ "主干不变：TourProduct -> ProductDay -> RouteStop -> ScenicAttraction，路线固定且有序。", "附近资源不写死到产品：Hotel/Restaurant 通过 ScenicAttraction -> ATTRACTION_NEARBY_RESOURCE 动态推荐。", "默认资源不污染资源本体：RouteStop/Product 到资源的默认、包含、必付、可选状态仍放关系属性。", "非 POI 统一 TravelItem：门票、小交通、保险、车辆、赠送服务、视频讲解等不再拆一堆实体。", "客服规则单独轻量化：退改、付款、成团、老人儿童、车型等放 BusinessRule，避免在长文本里临时找答案。", "不引入 CustomerSelection：客户选择属于订单/会话侧，暂不进入基础知识图谱。", "行政区继续属性化：当前业务按景区/线路/资源查询，行政区不再额外建层级关系。", "数据不确定必须标注：模拟价格、供应商待确认、政策冲突必须写 data_quality/requires_supplier_confirm。", ] schema["revision_notes"] = [ "从 2.3 升级到 2.4，新增 BusinessRule 和 RULE_APPLIES_TO。", "补充 TourProduct/ProductDay/RouteStop 的客服展示字段，支撑住哪、吃几餐、是否可替换、费用关注点。", "补充 Hotel/Restaurant 的客服筛选字段，支撑电梯、低楼层、家庭房、餐标和团型适配。", "补充 TravelItem 的去重与状态提示字段，解决同一观光车/扶梯被多个文本口径重复抽取的问题。", "继续保留 HotelRoomType 和 ResourcePriceRule；当前图谱数据中这两类节点为空，需要后续从住宿资源库补入。", ] schema["query_recipes"] = { **schema.get("query_recipes", {}), "customer_route_recommendation": "MATCH (p:TourProduct) WHERE toInteger(p.duration_days)=$days AND coalesce(p.small_group_supported,false)=true RETURN p ORDER BY toInteger(coalesce(p.sales_priority,'999')), toFloat(coalesce(p.base_price_min_adult,'999999')) LIMIT 5", "customer_day_detail": "MATCH (p:TourProduct)-[:PRODUCT_HAS_DAY]->(d:ProductDay)-[:DAY_HAS_STOP]->(s:RouteStop) WHERE p.product_id=$product_id RETURN p,d,s ORDER BY toInteger(d.day_no), toInteger(s.order_no)", "route_fee_summary": "MATCH (p:TourProduct)-[:PRODUCT_HAS_DAY]->(d:ProductDay)-[:DAY_HAS_STOP]->(s:RouteStop)-[:STOP_VISITS_ATTRACTION]->(a:ScenicAttraction)-[:ATTRACTION_HAS_ITEM]->(item:TravelItem) WHERE p.product_id=$product_id RETURN a.name,item.type,item.name,item.price,item.default_status_hint,item.mandatory_fee_policy ORDER BY a.name,item.type,item.price", "nearby_replaceable_resources": "MATCH (s:RouteStop)-[:STOP_VISITS_ATTRACTION]->(a:ScenicAttraction)-[near:ATTRACTION_NEARBY_RESOURCE]->(resource) WHERE s.stop_id=$stop_id AND labels(resource)[0] IN ['Hotel','Restaurant'] RETURN a,near,resource ORDER BY near.resource_type, toInteger(coalesce(near.rank,'999')) LIMIT 20", "vehicle_capacity_for_people": "MATCH (v:TravelItem) WHERE v.type='Vehicle' AND toInteger(coalesce(v.safe_passenger_capacity,v.seat_count,'0')) >= $people RETURN v ORDER BY toInteger(coalesce(v.safe_passenger_capacity,v.seat_count,'999')) LIMIT 5", "customer_service_rules": "MATCH (r:BusinessRule) WHERE r.rule_type IN $rule_types OR any(k IN split(coalesce(r.trigger_keywords,''),'|') WHERE $question CONTAINS k) RETURN r ORDER BY toInteger(coalesce(r.priority,'999')) LIMIT 10", } return schema def write_schema_files(schema: dict) -> dict[str, str]: SCHEMA_DIR.mkdir(parents=True, exist_ok=True) DOWNLOAD_DIR.mkdir(parents=True, exist_ok=True) paths = { "current_json": SCHEMA_DIR / "travel_agency_2_0_schema.current.json", "current_dsl": SCHEMA_DIR / "travel_agency_2_0_schema.current.dsl.md", "v_json": SCHEMA_DIR / "travel_agency_2_0_schema.v2_4.json", "v_dsl": SCHEMA_DIR / "travel_agency_2_0_schema.v2_4.dsl.md", "sample_json": SCHEMA_DIR / "travel_agency_2_0_schema.v2_4.original_sample.json", "sample_dsl": SCHEMA_DIR / "travel_agency_2_0_schema.v2_4.original_sample.dsl.md", } json_text = json.dumps(schema, ensure_ascii=False, indent=2) dsl_text = schema_to_dsl(schema) for path in paths.values(): path.write_text(json_text if path.suffix == ".json" else dsl_text, encoding="utf-8") shutil.copy2(path, DOWNLOAD_DIR / path.name) (DOWNLOAD_DIR / "当前schema文件位置.md").write_text( "\n".join( [ "# 当前旅行社2.0 Schema 文件位置", "", f"- DSL 源码：`{paths['v_dsl']}`", f"- JSON：`{paths['v_json']}`", f"- 系统页面：`http://localhost:8102/admin/modeling/schema`", "", "当前版本：2.4。新增客服业务规则层 BusinessRule；不新增 CustomerSelection。", ] ) + "\n", encoding="utf-8", ) return {key: str(path) for key, path in paths.items()} def publish_schema(schema: dict) -> int: with psycopg.connect(settings.database_url, row_factory=dict_row) as conn: with conn.cursor() as cur: cur.execute( f""" UPDATE {settings.db_schema}.ontology_schemas SET status='published', updated_at=now() WHERE tenant_id=%s AND project_id=%s AND status='active' """, (TENANT_ID, PROJECT_ID), ) cur.execute( f""" INSERT INTO {settings.db_schema}.ontology_schemas ( tenant_id, project_id, namespace, version, display_name, description, status, schema_jsonb, created_by, published_by, published_at, updated_at ) VALUES (%s,%s,%s,%s,%s,%s,'active',%s,%s,%s,now(),now()) RETURNING id """, ( TENANT_ID, PROJECT_ID, NAMESPACE, 6, schema["display_name"], schema["purpose"], Jsonb(schema), "codex", "codex", ), ) schema_id = cur.fetchone()["id"] cur.execute( f""" UPDATE {settings.db_schema}.graph_releases SET schema_id=%s, graph_release_id=%s, graph_name=%s, status='active', activated_at=now(), updated_at=now(), metadata_jsonb = coalesce(metadata_jsonb, '{{}}'::jsonb) || %s::jsonb WHERE tenant_id=%s AND project_id=%s AND alias='active' """, ( schema_id, "travel_agency_2_0_test_v2_4", GRAPH_NAME, Jsonb({"schema_version": "2.4", "schema_id": schema_id}), TENANT_ID, PROJECT_ID, ), ) conn.commit() return schema_id def main() -> None: schema = build_schema() files = write_schema_files(schema) schema_id = publish_schema(schema) print( json.dumps( { "schema_id": schema_id, "schema_version": schema["version"], "entity_count": len(schema["entity_types"]), "relation_count": len(schema["relation_types"]), "files": files, }, ensure_ascii=False, indent=2, ) ) if __name__ == "__main__": main()