idZddlZddlZddlZddlmZmZmZmZddl m Z ddl m Z ejeZejdj!ej"daGddZeZy) uj Cliente para Neo4j - GraphRAG. Maneja conexiones y queries Cypher para el grafo de conocimiento médico. N)ListDictAnyOptional) GraphDatabase)settingszneo4j.notificationsceZdZdZdZdZdefdZdefdZdZ d Z d Z dAd ed e e eefde edee eeffdZdBdededee eeffdZdBdededee eeffdZ dCdededededeedeededede eeffdZdDdededee eeffdZdDdededee eeffdZdede eeffdZdEdededee eeffdZ dFdedeedee eeffdZdBdedee eeffd Zdefd!Zdeefd"Zd#edefd$Zd#edefd%Z dEd#ededee eeffd&Z!defd'Z"defd(Z#dGd)ed*edeefd+Z$d,ee eefde eeffd-Z%de eeffd.Z&de'fd/Z(defd0Z)defd1Z*defd2Z+defd3Z,d4eedefd5Z-d6Z.d7Z/dGd8edefd9Z0d8edefd:Z1d;ee eefde eeffd<Z2d;ee eefd8ed=ede eeffd>Z3dHded8ededee eeffd?Z4d,ee eefde eeffd@Z5y )I Neo4jClientz3Cliente para interactuar con Neo4j GraphRAG Retail.xctj|_tj|_tj |_tj|_d|_ d|_ tj|_ ddlm}||_y)N)embeddings_service)r NEO4J_URIuri NEO4J_USERuserNEO4J_PASSWORDpasswordNEO4J_DATABASEdatabase_driver_keepalive_thread threadingEvent_keepalive_stopr)selfrs 8/opt/lhia/marcimex/agent/app/rag_service/neo4j_client.py__init__zNeo4jClient.__init__s_%%''  //  //  37(0:"4rcB||_tjd|y)zCambia la base de datos activa.u🔄 Base de datos cambiada a: N)rloggerinfo)rrs r set_databasezNeo4jClient.set_database(s   5hZ@Arreturnc|jS)z-Retorna el nombre de la base de datos actual.rrs rget_current_databasez Neo4jClient.get_current_database-s }}rc  tj|j|j|jfddddd|_|j j tjd|jd|jd|jy#t$r"}tjd |Yd }~y d }~wwxYw) u4Establece conexión con Neo4j e inicia el keepalive. 2T)authconnection_acquisition_timeoutmax_connection_lifetimeconnection_timeoutmax_connection_pool_size keep_aliveu✅ Conectado a Neo4j en z (DB: )u❌ Error conectando a Neo4j: NF) rdriverrrrrverify_connectivityr!r"r_start_keepalive Exceptionerror)res rconnectzNeo4jClient.connect1s (//ii//1(+#%)+DL LL , , . KK3DHH:VDMM?RST U  ! ! #  LL9!= > sBB C 'CC cjjjjr6jjrjj djj fd}t j|dd_jjy)uVInicia un hilo daemon que ejecuta RETURN 1 cada 2 min para mantener la conexión viva.timeoutcjjjsjsy jj j 5}|j djdddtjdjjjsyy#1swYGxYw#t$r"}tjd|Yd}~[d}~wwxYw)Nr=r&zRETURN 1u💓 Neo4j keepalive OKu⚠️ Neo4j keepalive falló: ) rwait_KEEPALIVE_INTERVALrsessionrrunconsumer!debugr7warning)rBr9rs r _ping_loopz0Neo4jClient._start_keepalive.._ping_loopNs**//8P8P/Q||J--t}}-E: J/779:LL!:; **//8P8P/Q::!JNN%DQC#HIIJs/&C C<CC C C9C44C9Tzneo4j-keepalive)targetdaemonnameN) rsetris_alivejoinclearrThreadstart)rrGs` rr6zNeo4jClient._start_keepaliveEs   "  ! !d&<&<&E&E&G  " " ' ' ' 2 ""$ J"+!1!1DWh!i $$&rc|jj|jr0|jjtj dyy)Nu🔌 Desconectado de Neo4j)rrKrcloser!r"r's rrRzNeo4jClient.close\s=   " << LL   KK4 5 rNquery parametersc|js|jsgS|xs |j}tdD]`} |jj |5}|j ||xsid}|Dcgc]}|j c}cdddcSgScc}w#1swYnxYww#t$r} t| } tjd| |dk(rd| jvs$d| jvsd | jvrftjd  |jjn#t$rYnwxYwd|_|js gcYd} ~ cSYd} ~ DgcYd} ~ cSd} ~ wwxYw) Nr<r&r=u❌ Error query: rdefunctzfailed to read connectionu<⚠️ Conexión Neo4j defunta detectada — reconectando...)rr:rrangerBrCdatar7strr!r8lowerrFrR) rrSrTr db_to_useattemptrBresultrecordr9err_strs r execute_queryzNeo4jClient.execute_querybs~||<<>"9- Qx G \\))9)=@$[[ 0@b"[MF8>?fFKKM?@@ ( !@@@@ a& 045a!   sxB1B$7BB$ B1B$$B- )B11 F :A3F.E F EFEF0F >F?F F  query_textlimitcd}d} |j|||d}tjdt|d|Dcgc]&}|j dd|j ddf(c}|Scc}w#|j|||d}tjd t|d|Dcgc]&}|j dd|j ddf(ncc}wc}|cYSxYw) uJ Búsqueda difusa de productos por nombre o descripción. a? CALL db.index.fulltext.queryNodes('productFulltext', $query_text) YIELD node, score MATCH (node:Product) OPTIONAL MATCH (node)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (node)-[:MADE_BY]->(b:Brand) RETURN node.id as id, node.sku_id as sku_id, node.name as name, node.price as price, node.description as description, node.link as link, node.image_link as image_link, c.name as category, b.name as brand, score, node.promotion as promotion ORDER BY score DESC LIMIT $limit a MATCH (p:Product) WHERE toLower(p.name) CONTAINS toLower($query_text) OR toLower(p.description) CONTAINS toLower($query_text) OPTIONAL MATCH (p)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (p)-[:MADE_BY]->(b:Brand) RETURN p.id as id, p.sku_id as sku_id, p.name as name, p.price as price, p.description as description, p.link as link, p.image_link as image_link, c.name as category, b.name as brand, p.promotion as promotion LIMIT $limit rcrdu'🔍 Neo4j fulltext_search resultados (): id?rJu'🔍 Neo4j fallback_search resultados ()rbr!r"lenget)rrcrdrSfallback_queryresultsrs rfulltext_search_productsz$Neo4jClient.fulltext_search_products~s3     ((zTY0Z[G KKA#g,sCJTK}~UVUZUZ[_`cUdfgfkfklrsvfwTxTKSLM NNTK )).^c:deW [[B3w<.PSDKUL~VWV[V[\`adVeghglglmstwgxUyULULTMNO^s(5A6+A1& A61A667C+-+CC+c B |jj|}|s'tjd|j ||Sd}|j |||d}tj dt|d|Dcgc]&}|jdd|jddf(c}|s'tjd |j ||S|Scc}w#t$r4}tjd ||j ||cYd }~Sd }~wwxYw) u Búsqueda semántica usando embeddings. Fallback a fulltext si el índice no existe o retorna vacío. ,⚠️ Embedding vacío, fallback a fulltextab CALL db.index.vector.queryNodes('productSearch', $limit, $embedding) YIELD node, score MATCH (node:Product) OPTIONAL MATCH (node)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (node)-[:MADE_BY]->(b:Brand) RETURN node.id as id, node.sku_id as sku_id, node.name as name, node.price as price, node.description as description, node.link as link, node.image_link as image_link, c.name as category, b.name as brand, score, node.promotion as promotion ORDER BY score DESC, node.id ASC  embeddingrdu%🔍 Neo4j vector_search resultados (rgrhrirJu8⚠️ Vector search sin resultados, fallback a fulltextu❌ Error vector search: N) r embed_queryr!rFrorbr"rjrkr7r8)rrcrdrsrSrmrnr9s rvector_search_productsz"Neo4jClient.vector_search_productss3  D//;;JGI NO55j%HH E((iRW0XYG KK?G ~SAHRI{|STSXSXY]^aSbdedidijpqtduRvRIQJK LYZ44ZGGN RI  D LL4QC8 900UC C Ds<AC!7C!=+C(1C!C!C!! D*)DDDskip min_scoreexcluded_categoriesrspromotion_filterorder_by_promotionc tdz at} tjd| d|d|d|tj} |dtj} |jj |}tjd|ddd tj| z d d |s-tj d |j||} | d gdS|g}||zt|dzzdz} |rdnd}d|d}|dz}tj} |j||| |||||d} tjdtj| z d d t| |kD}| d|}tjd|d|d|dt|d|dtj| z d d |||dS#t$r<}tjd|d|j||} | d gdcYd}~Sd}~wwxYw) u Búsqueda vectorial con paginación, filtrado de score y categorías. Acepta un embedding pre-calculado para garantizar consistencia entre páginas. promotion_filter=True → solo retorna productos con promotion=true (filtro duro). order_by_promotion=True → muestra todos los resultados, promocionados primero. Returns: Dict con 'products' (lista) y 'has_more' (bool) r u7📡 [Neo4j] vector_search_products_paginated LLAMADA #z q='z' skip=z limit=Nu⏱️ [Neo4j] embed_query('(z'): z.2fsrqF)productshas_morersr*z5ORDER BY node.promotion DESC, score DESC, node.id ASCz ORDER BY score DESC, node.id ASCa CALL db.index.vector.queryNodes('productSearch', $fetch_count, $embedding) YIELD node, score WHERE score >= $min_score MATCH (node:Product) OPTIONAL MATCH (node)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (node)-[:MADE_BY]->(b:Brand) WITH node, score, c, b WHERE (c IS NULL OR NOT c.name IN $excluded_categories) AND (NOT $promotion_filter OR node.promotion = true) RETURN node.id AS id, node.sku_id AS sku_id, node.name AS name, node.price AS price, node.description AS description, node.summary AS summary, node.link AS link, node.image_link AS image_link, node.in_stock AS in_stock, node.availability AS availability, node.color AS color, node.specifications AS specifications, node.global_category AS global_category, node.sub_category AS sub_category, node.product_type AS product_type, node.promotion AS promotion, node.sync_status AS sync_status, c.name AS category, b.name AS brand, score zB SKIP $skip LIMIT $page_limit )rs fetch_countrwrxrv page_limitryu(⏱️ [Neo4j] execute_query vectorial: u$🔍 [Neo4j] paginated_search (skip=z, limit=z , min_score=rgz productos, has_more=u — total: u#❌ Error paginated vector search: Texc_info) _vector_search_call_countr!r"timerrtrFrorjrbr7r8)rrcrvrdrwrxrsryrzcall_nt_startt0rmr order_clauserSrrr~r9s r vector_search_products_paginatedz,Neo4jClient.vector_search_products_paginatedst, "Q&!* EfXNWTF'% : ))+Y M YY[ 33?? K  ::cr?:K4PTPYPYP[^`P`adOeefghMN77 EJ$+RPP"*&(#,-@)AB)FFKK&H7   <N =!EHJB((&*&':($41G KKB499;QSCSTWBXXYZ [7|e+HvH KK6tfHUGqcBTL R33JFG 'UL L Ms&BGC(G H  1H>H H c|jj|}|s'tjd|j ||Sd} |j |||d}tj dt|d|d|S#t$r6}tjd|d |j ||cYd }~Sd }~wwxYw) u Búsqueda vectorial que retorna TODOS los atributos del nodo Product excepto el embedding. Usada por el endpoint de diagnóstico /products/search. rqa CALL db.index.vector.queryNodes('productSearch', $limit, $embedding) YIELD node, score MATCH (node:Product) OPTIONAL MATCH (node)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (node)-[:MADE_BY]->(b:Brand) RETURN node.id AS id, node.sku_id AS sku_id, node.name AS name, node.price AS price, node.description AS description, node.summary AS summary, node.link AS link, node.image_link AS image_link, node.in_stock AS in_stock, node.availability AS availability, node.color AS color, node.specifications AS specifications, node.global_category AS global_category, node.sub_category AS sub_category, node.product_type AS product_type, node.promotion AS promotion, node.sync_status AS sync_status, c.name AS category, b.name AS brand, score ORDER BY score DESC LIMIT $limit rru🔍 [vector_search_full]  productos para ''u❌ Error vector_search_full: TrN) rrtr!rFrorbr"rjr7r8)rrcrdrsrSrmr9s rvector_search_products_fullz'Neo4jClient.vector_search_products_full<s ++77 C  NNI J00UC C < D((iRW0XYG KK4S\NBST^S__`a bN D LL9!=L M00UC C Ds;B C +B>8C>Ccd} |j|||d}tjdt|d|d|S#t$r&}tj d|dgcYd }~Sd }~wwxYw) us Búsqueda fulltext que retorna TODOS los atributos del nodo Product excepto el embedding. a CALL db.index.fulltext.queryNodes('productFulltext', $query_text) YIELD node, score MATCH (node:Product) OPTIONAL MATCH (node)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (node)-[:MADE_BY]->(b:Brand) RETURN node.id AS id, node.sku_id AS sku_id, node.name AS name, node.price AS price, node.description AS description, node.summary AS summary, node.link AS link, node.image_link AS image_link, node.in_stock AS in_stock, node.availability AS availability, node.color AS color, node.specifications AS specifications, node.global_category AS global_category, node.sub_category AS sub_category, node.product_type AS product_type, node.promotion AS promotion, node.sync_status AS sync_status, c.name AS category, b.name AS brand, score ORDER BY score DESC LIMIT $limit rfu🔍 [fulltext_search_full] rru ❌ Error fulltext_search_full: TrN)rbr!r"rjr7r8)rrcrdrSrmr9s rfulltext_search_products_fullz)Neo4jClient.fulltext_search_products_fullls|  < ((zTY0Z[G KK6s7|nDUV`Uaabc dN  LL;A3?$L OI s;A A/ A*$A/*A/ product_idc@d}|j|d|i}|r|dSiS)zF Obtiene detalles completos de un producto por su ID. ar MATCH (p:Product {id: $product_id}) OPTIONAL MATCH (p)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (p)-[:MADE_BY]->(b:Brand) RETURN p.id as id, p.sku_id as sku_id, p.name as name, p.price as price, p.description as description, p.summary as summary, p.link as link, p.image_link as image_link, p.in_stock as in_stock, p.availability as availability, p.color as color, p.specifications as specifications, p.promotion as promotion, p.sync_status as sync_status, c.name as category, b.name as brand rrrb)rrrSrms rget_product_contextzNeo4jClient.get_product_contexts3  $$U\:,FG$wqz,",rc0d}|j|||dS)u- Busca políticas de tienda. z MATCH (pol:Policy) WHERE toLower(pol.text) CONTAINS toLower($query_text) OR toLower(pol.title) CONTAINS toLower($query_text) RETURN pol as policy LIMIT $limit rfr)rrcrdrSs rsearch_policieszNeo4jClient.search_policiess# !!% U)STTrc|g}d}|j|||d}tjdt|d|S)u Retorna productos con promotion=true sin búsqueda vectorial. Se usa cuando el usuario pide ver las ofertas/promociones de forma genérica. a MATCH (p:Product) WHERE p.promotion = true AND p.sync_status <> 'eliminado' OPTIONAL MATCH (p)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (p)-[:MADE_BY]->(b:Brand) WITH p, c, b WHERE c IS NULL OR NOT c.name IN $excluded_categories RETURN p.id AS id, p.sku_id AS sku_id, p.name AS name, p.price AS price, p.description AS description, p.summary AS summary, p.link AS link, p.image_link AS image_link, p.in_stock AS in_stock, p.availability AS availability, p.color AS color, p.specifications AS specifications, p.global_category AS global_category, p.sub_category AS sub_category, p.product_type AS product_type, p.promotion AS promotion, p.sync_status AS sync_status, c.name AS category, b.name AS brand LIMIT $limit )rxrdu 🏷️ [get_promoted_products] u productos en promociónrbr!r"rj)rrdrxrSrms rget_promoted_productsz!Neo4jClient.get_promoted_productssX  &"$  8$$U#6-    6s7|nD\]^rcvd}|j|d|i}tjdt|d|S)u} Obtiene productos destacados/aleatorios para mostrar cuando el usuario hace una consulta genérica. a MATCH (p:Product) WHERE p.sync_status <> 'eliminado' OPTIONAL MATCH (p)-[:BELONGS_TO]->(c:Category) OPTIONAL MATCH (p)-[:MADE_BY]->(b:Brand) RETURN p.id as id, p.sku_id as sku_id, p.name as name, p.price as price, p.description as description, p.link as link, p.image_link as image_link, c.name as category, b.name as brand, 1.0 as score, p.promotion as promotion ORDER BY rand() LIMIT $limit rdu🌟 Productos destacados: productosr)rrdrSrms rget_featured_productsz!Neo4jClient.get_featured_productssA  $$UWe,<= 1#g,zJKrcd}|j|}|r|djddnd}tjd|d|S)z Asigna promotion=false a todos los productos que no tengan el campo. Retorna la cantidad de productos actualizados. z MATCH (p:Product) WHERE p.promotion IS NULL SET p.promotion = false RETURN count(p) as updated rupdatedu"🏷️ promotion inicializado en rrbrkr!r")rrSrmrs rinitialize_promotion_fieldz&Neo4jClient.initialize_promotion_fieldsN   $$U+29'!*..A.q 8 LMrc|d}|j|}|Dcgc]}|jds|dc}Scc}w)uD Obtiene las categorías disponibles con productos. z MATCH (p:Product)-[:BELONGS_TO]->(c:Category) WHERE p.sync_status <> 'eliminado' RETURN DISTINCT c.name as category, count(p) as count ORDER BY count DESC LIMIT 10 categoryrbrkrrSrmrns rget_available_categoriesz$Neo4jClient.get_available_categories s> $$U+'.D!!%% 2C* DDD99 customer_idc|r|syd} |j|||dy#t$r"}tjd|Yd}~yd}~wwxYw)z(Registra que un usuario vio un producto.Nz MERGE (u:Customer {id: $cid}) MERGE (p:Product {id: $pid}) MERGE (u)-[r:VIEWED]->(p) SET r.timestamp = datetime() )cidpidu⚠️ Error logging view: rbr7r!rF)rrrrSr9s r log_user_viewzNeo4jClient.log_user_viewsQ*f  =   e[%L M = >>7s; < < = A AA c|r|syd} |j|||dy#t$r"}tjd|Yd}~yd}~wwxYw)u Registra qué buscó el usuario.Nz MERGE (u:Customer {id: $cid}) MERGE (s:SearchTerm {text: toLower($text)}) MERGE (u)-[r:SEARCHED]->(s) SET r.timestamp = datetime() )rtextu⚠️ Error logging search: r)rrrcrSr9s rlog_user_searchzNeo4jClient.log_user_search'sT*f   @   uk:&N O @ NN:1#> ? ? @rcn|sgSd}|j|||d}|sd}|j|||d}|S)u Genera recomendaciones basadas en productos vistos previamente. Lógica: "Usuarios que vieron lo que tú viste, también vieron..." as MATCH (current:Customer {id: $cid})-[:VIEWED]->(p:Product) MATCH (other:Customer)-[:VIEWED]->(p) WHERE other.id <> $cid MATCH (other)-[:VIEWED]->(reco:Product) WHERE NOT (current)-[:VIEWED]->(reco) RETURN reco.id as id, reco.name as name, count(*) as frequency ORDER BY frequency DESC LIMIT $limit )rrda1 MATCH (u:Customer {id: $cid})-[:VIEWED]->(p:Product)-[:BELONGS_TO]->(c:Category) MATCH (reco:Product)-[:BELONGS_TO]->(c) WHERE NOT (u)-[:VIEWED]->(reco) AND reco.id <> p.id RETURN reco.id as id, reco.name as name LIMIT $limit r)rrrdrSrm query_contents rget_user_recommendationsz$Neo4jClient.get_user_recommendations6sV 2I  $$UK%,PQ]))-W\9]^Wrc@d}|j|}|r|ddSdS)u9Cuenta el número total de productos en la base de datos.z*MATCH (p:Product) RETURN count(p) as totalrtotalr)rrSrms rcount_productszNeo4jClient.count_productsYs-<$$U+&-wqz'"414rc(|jdk(S)u7Verifica si la base de datos de productos está vacía.r)rr's ris_database_emptyzNeo4jClient.is_database_empty_s""$))rr dimensionscddl}|tj}|j|j j }g}t |D]6}||t|z}|dz dz }|jt|d8|S)u} Genera un embedding fake para pruebas. Usa un hash del texto para generar valores determinísticos. rNg_@g?) hashlibrEMBEDDING_DIMENSIONSsha256encodedigestrYrjappendround) rrrr hash_bytesfake_embeddingibyte_val normalizeds rgenerate_fake_embeddingz#Neo4jClient.generate_fake_embeddingcs   !66J^^DKKM299; z" 8A!!c*o"56H"U*c1J  ! !% A"6 7  8 rr~c |sddddSd}t|Dcgc]\}}|jdddk(s||f!}}}t|Dcgc]\}}|jdddk7s||f!}}}i}|r|Dcgc]?\}}|jddd |jd dd |jd dA} }}tjd t | d |j j | } t|D]]\} \} }| t | kr| | r | | || <%tjd| |j j| | || <_d} d}d'dtttfdttdtttffd}d}d}|Dcgc]\}}|||j|g }}}tdt ||D]l}||||z} |j| d|i}|r|ddnd}||z }|t ||z z }tjd|d|t |zd|dn|Dcgc] \}}||}}}tdt ||D]l}||||z} |j|d|i}|r|ddnd}||z }|t ||z z }tjd|d|t |zd|dntjd |d!|d"||t |d#|d$|d%d&Scc}}wcc}}wcc}}wcc}}w#t $r4}tj#d|d||t |z }Yd}~d}~wwxYwcc}}w#t $r4}tj#d|d||t |z }Yd}~8d}~wwxYw)(u5 Crea o actualiza productos en batch en Neo4j. Solo genera embeddings para productos NUEVOS (sync_status='nuevo'). Los productos existentes (sync_status='normal') mantienen su embedding. Optimizaciones: - Embeddings en batch (1 llamada OpenAI para todos los nuevos, no 1 por producto) - Escrituras en Neo4j por chunks con UNWIND (no 1 query por producto) Args: products: Lista de diccionarios con datos de productos Returns: Dict con estadísticas de la operación rzNo products to createcreatederrorsmessager+ sync_statusnuevotitle descriptionru(🔢 Generando embeddings en batch para z productos nuevos...u1⚠️ Fallback embed_query para producto en pos a} UNWIND $products AS prod MERGE (p:Product {id: prod.product_id}) SET p.name = prod.title, p.description = prod.description, p.summary = prod.summary, p.price = prod.price, p.link = prod.link, p.image_link = prod.image_link, p.sku_id = prod.sku_id, p.rpx_code = prod.rpx_code, p.in_stock = prod.in_stock, p.availability = prod.availability, p.status = prod.status, p.color = prod.color, p.specifications = prod.specifications, p.global_category = prod.global_category, p.sub_category = prod.sub_category, p.product_type = prod.product_type, p.promotion = prod.promotion, p.sync_status = prod.sync_status, p.embedding = prod.embedding, p.updated_at = datetime() WITH p, prod FOREACH (_ IN CASE WHEN prod.category IS NOT NULL THEN [1] ELSE [] END | MERGE (c:Category {name: prod.category}) MERGE (p)-[:BELONGS_TO]->(c) ) FOREACH (_ IN CASE WHEN prod.sub_category IS NOT NULL THEN [1] ELSE [] END | MERGE (sc:SubCategory {name: prod.sub_category}) MERGE (p)-[:IN_SUBCATEGORY]->(sc) ) FOREACH (_ IN CASE WHEN prod.brand IS NOT NULL THEN [1] ELSE [] END | MERGE (b:Brand {name: prod.brand}) MERGE (p)-[:MADE_BY]->(b) ) RETURN count(p) AS processed aM UNWIND $products AS prod MATCH (p:Product {id: prod.product_id}) SET p.name = prod.title, p.description = prod.description, p.summary = prod.summary, p.price = prod.price, p.link = prod.link, p.image_link = prod.image_link, p.sku_id = prod.sku_id, p.rpx_code = prod.rpx_code, p.in_stock = prod.in_stock, p.availability = prod.availability, p.status = prod.status, p.color = prod.color, p.specifications = prod.specifications, p.global_category = prod.global_category, p.sub_category = prod.sub_category, p.product_type = prod.product_type, p.promotion = prod.promotion, p.sync_status = prod.sync_status, p.updated_at = datetime() WITH p, prod FOREACH (_ IN CASE WHEN prod.category IS NOT NULL THEN [1] ELSE [] END | MERGE (c:Category {name: prod.category}) MERGE (p)-[:BELONGS_TO]->(c) ) FOREACH (_ IN CASE WHEN prod.sub_category IS NOT NULL THEN [1] ELSE [] END | MERGE (sc:SubCategory {name: prod.sub_category}) MERGE (p)-[:IN_SUBCATEGORY]->(sc) ) FOREACH (_ IN CASE WHEN prod.brand IS NOT NULL THEN [1] ELSE [] END | MERGE (b:Brand {name: prod.brand}) MERGE (p)-[:MADE_BY]->(b) ) RETURN count(p) AS processed Nproductrsr$c |jdd}id|jdd|jdd|jdd|jdd|jdd|jdd |jd d |jd d |jd d |jd d |jd d|jdd|jdd|jdd|jdd|jdd|jd|jd|jd|jdd|d}|||d<|S)Nrrrrrsummarypricelink image_linksku_idrpx_coderglobal_category sub_category product_typebrandin_stock availabilitystatuscolorspecifications promotionF)rrrrrs)rk)rrsrds r _build_paramz6Neo4jClient.bulk_create_products.._build_params!++mW=K\!:W!5]!;Y!7  W!5  V!4 \!:X!6Z!87;;z#:"7;;/@#A7;;~#>7;;~#>7;;w#7Z!8 ^!"?!([%!@!,+A.$!*+Hrr~ processedu ✅ Nuevos [u–z]: z procesadosu❌ Error batch nuevos chunk : u ✅ Actualizados [u❌ Error batch updates chunk u✅ Sync completado: z procesados,  erroreszSync completed: z processed,  errorsrrrrN) enumeraterkr!r"rjrembed_documentsrFrtrr[rrfloatrYrbr7r8)rr~ CHUNK_SIZErp new_productsupdate_productsembeddings_map_textsbatch_embeddingsjidxquery_new_batchquery_update_batchrrr new_params chunk_startchunkr_rr9 update_paramss rbulk_create_productsz Neo4jClient.bulk_create_productsws]" A:QR R .7x-@mTQAEE-Y`DaelDl1a&m m.7.AndaQUU=ZaEbfmEmAq6nn24 )Aq55"%&amR(@'A155UWCXBYZE KKB3u:,Nbc d#66FFuM (6 X 8Cs+,,1A!1D*:1*=N3'NN%VWXVY#Z[*.*A*A*M*MeTUh*WN3'  X% P$ L $sCx. T%[ TXY\^aYaTb :NZZTQl1n&8&8B&?@Z Z C OZ@ %K{K*,DEE %++Oj%=PQ6U@ST6AL;2#M:A!M"NA!N  N)M??N O )OO c|j}|dk(rdddSd}|j|}|r|ddnd}d}|j|tjd|d|d |d dS) u Elimina todos los productos y sus relaciones de la base de datos. Returns: Dict con estadísticas de eliminación rzNo hay productos para eliminar)deletedrz| MATCH (p:Product) WITH p, p.id as pid DETACH DELETE p RETURN count(pid) as deleted rz~ MATCH (n) WHERE (n:Category OR n:SubCategory OR n:Brand) AND NOT (n)<-[]-() DELETE n 🗑️ Eliminados z productos de Neo4jz Eliminados z productos correctamente)rrbr!r")r count_beforerSrmr cleanup_querys rdelete_all_productszNeo4jClient.delete_all_products5s**, 1 ;    $$U++2'!*Y' =) )'2EFG$WI-EF  rc|d}|j|}|Dchc]}|jds|dc}Scc}w)zBObtiene todos los IDs de productos existentes en la base de datos.z#MATCH (p:Product) RETURN p.id as idrhrrs rget_all_product_idszNeo4jClient.get_all_product_ids^s95$$U+!(8AAEE$K$888rcvd}|j|}|r|ddnd}tjd|d|S)z Marca todos los productos existentes como PENDING. Returns: Cantidad de productos marcados zn MATCH (p:Product) SET p.sync_status = 'pendiente' RETURN count(p) as updated rru📋 Marcados z productos como PENDINGrbr!r"rrSrmcounts rmark_all_products_as_pendingz(Neo4jClient.mark_all_products_as_pendingdsH  $$U+)0 9%a nUG+BCD rcvd}|j|}|r|ddnd}tjd|d|S)u Marca como DELETED todos los productos que quedaron en estado PENDING. Estos son productos que existían en Neo4j pero no están en el nuevo XML. Returns: Cantidad de productos marcados como eliminados z MATCH (p:Product) WHERE p.sync_status = 'pendiente' SET p.sync_status = 'eliminado' RETURN count(p) as deleted rru🗑️ Marcados z. productos como DELETED (no estaban en el XML)r r s rmark_pending_as_deletedz#Neo4jClient.mark_pending_as_deletedusI  $$U+)0 9%a 'w.\]^ rcvd}|j|}|r|ddnd}tjd|d|S)u Elimina físicamente de Neo4j todos los productos con sync_status='eliminado'. Usa DETACH DELETE para eliminar también sus relaciones. Returns: Cantidad de productos eliminados z MATCH (p:Product) WHERE p.sync_status = 'eliminado' WITH p, p.id as pid DETACH DELETE p RETURN count(pid) as deleted rru 🗑️ Eliminados físicamente z! productos con estado 'eliminado'r r s rdelete_eliminated_productsz&Neo4jClient.delete_eliminated_productssI $$U+)0 9%a 6ug=^_` rcd}|j|}|r|djddnd}tjd|d|S)u Pone promotion=false en todos los productos que estaban en true. Se llama antes de aplicar el nuevo catálogo de promociones. Returns: Cantidad de productos reseteados z MATCH (p:Product) WHERE p.promotion = true SET p.promotion = false RETURN count(p) as updated rru'🔄 Promociones reseteadas a false en rrr s rreset_all_promotionsz Neo4jClient.reset_all_promotionssN  $$U+07 y!,Q =eWJOP r product_idsc|syd}|j|d|i}|r|djddnd}tjd|dt |d|S)u Activa promotion=true para los productos cuyo id esté en la lista. Args: product_ids: Lista de product_id que están en promoción Returns: Cantidad de productos actualizados a promotion=true rz UNWIND $ids AS pid MATCH (p:Product {id: pid}) SET p.promotion = true RETURN count(p) as updated idsru!🏷️ Promociones activadas en z de r)rbrkr!r"rj)rrrSrmrs rset_promotions_by_idsz!Neo4jClient.set_promotions_by_idssj  $$UUK,@A07 y!,Q 7wd3{CSBTT^_` rctj}d|d} |j|tj d|dy#t $r"}tj d|Yd}~yd}~wwxYw)u Crea el índice vectorial para productos si no existe. Nota: CREATE INDEX no acepta parámetros en Neo4j, se usa valor literal. z CREATE VECTOR INDEX productSearch IF NOT EXISTS FOR (p:Product) ON (p.embedding) OPTIONS {indexConfig: { `vector.dimensions`: H, `vector.similarity_function`: 'cosine' }} u>✅ Índice vectorial 'productSearch' verificado/creado (dims=r3u?⚠️ Error creando índice productSearch (puede ya existir): Nrrrbr!r"r7rFrrrSr9s rensure_product_vector_indexz'Neo4jClient.ensure_product_vector_indexs 22 " #-.   b   u % KKXYcXddef g b NN\]^\_` a a b*A A. A))A.ctj}d|d} |j|tj d|dy#t $r"}tj d|Yd}~yd}~wwxYw)u Crea el índice vectorial para documentos si no existe. Separado del índice de productos (productSearch). Nota: CREATE INDEX no acepta parámetros en Neo4j, se usa valor literal. z CREATE VECTOR INDEX documentSearch IF NOT EXISTS FOR (d:Document) ON (d.embedding) OPTIONS {indexConfig: { `vector.dimensions`: ru?✅ Índice vectorial 'documentSearch' verificado/creado (dims=r3u@⚠️ Error creando índice documentSearch (puede ya existir): Nrrs rensure_document_vector_indexz(Neo4jClient.ensure_document_vector_indexs 22 " #-.   c   u % KKYZdYeefg h c NN]^_]`a b b crdoc_typecr|rd}|j|d|i}nd}|j|}|r|ddSdS)z3Cuenta documentos, opcionalmente filtrado por tipo.zAMATCH (d:Document {doc_type: $doc_type}) RETURN count(d) as totalr"z+MATCH (d:Document) RETURN count(d) as totalrrr)rr"rSrms rcount_documentszNeo4jClient.count_documentssL WE((X0FGGAE((/G&-wqz'"414rcd}|j|d|i}|r|ddnd}tjd|d|d|S)u4Elimina todos los documentos de un tipo específico.zu MATCH (d:Document {doc_type: $doc_type}) DETACH DELETE d RETURN count(d) as deleted r"rrrz documentos de tipo 'rr )rr"rSrmrs rdelete_documentszNeo4jClient.delete_documentssW  $$UZ,BC+2'!*Y' )'2GzQRSTr documentsc J|sddddSd}d}d}|D]} |jddd|jdd}|jj|}|s9tj d |jd d |j |}|jd |jd |jd|jd|jd d|jdd|jdd|d}|j ||} | r|dz }n|dz }tjd|d|d||t|d|d|ddS#t$r:} tjd|jd d| |dz }Yd} ~ d} ~ wwxYw)uC Crea documentos en batch en Neo4j con embeddings. Cada documento debe tener: id, doc_type, title, content, source, chunk_index, total_chunks. Args: documents: Lista de dicts con datos de documentos Returns: Dict con estadísticas de la operación rzNo documents to createra MERGE (d:Document {id: $id}) ON CREATE SET d.created_at = datetime() SET d.doc_type = $doc_type, d.title = $title, d.content = $content, d.source = $source, d.chunk_index = $chunk_index, d.total_chunks = $total_chunks, d.embedding = $embedding, d.updated_at = datetime() RETURN d.id as id rrrcontentu'⚠️ Embedding vacío para documento rhz , usando faker"source chunk_index total_chunksr )rhr"rr)r*r+r,rsu❌ Error procesando documento rNu ✅ Documentos sync completado: z creados, rzDocuments sync: z created, rr) rkrrtr!rFrrbr7r8r"rj) rr'rrrSdoctext_for_embeddingrsparamsr_r9s rbulk_create_documentsz!Neo4jClient.bulk_create_documents s A:RS S   C (+(<'=QswwyRT?U>V%W" 33??@RS  NN%LSWWUY]O[h#ij $ < <=O PI''$- # 3 WWW-"wwy1!ggh3#&77=!#<$'GGNA$>!* ++E6:qLGaKF3 >  6wiz&QYZ[^)'*VHGL     >swwt}oRPQsST!  sDE F"(/FF" source_urlc|j|j|}tjd|d||dkDr|j ||D] }||d<||d<|j |}||d<||d<|S)u Sincroniza documentos de un tipo específico a Neo4j. Estrategia: eliminar todos los del tipo y re-insertar (idempotente). Args: documents: Lista de documentos a sincronizar doc_type: Tipo de documento (faq, policy, etc.) source_url: URL de origen de los documentos Returns: Dict con resultado de la sincronización u📊 Documentos 'z' existentes: rr"r*previous_count)r!r$r!r"r&r0)rr'r"r1r3r-r_s rsync_documentszNeo4jClient.sync_documentsSs ))+--h7 'z?OPQ A   ! !( + 'C&C O&CM '++I6%z#1  rcr |jj|}|stjdgS|r d}||||dzd}nd}||d}|j ||}tj d|xsdd t |d |S#t$r$}tjd |gcYd }~Sd }~wwxYw) u+ Búsqueda semántica de documentos por similitud vectorial. Args: query_text: Texto de búsqueda doc_type: Filtrar por tipo (opcional) limit: Máximo de resultados Returns: Lista de documentos con score de similitud u6⚠️ Embedding vacío, no se puede buscar documentosa CALL db.index.vector.queryNodes('documentSearch', $fetch_count, $embedding) YIELD node, score WHERE node.doc_type = $doc_type RETURN node.id as id, node.doc_type as doc_type, node.title as title, node.content as content, node.source as source, score ORDER BY score DESC LIMIT $limit r*)rsr"rdrag CALL db.index.vector.queryNodes('documentSearch', $limit, $embedding) YIELD node, score RETURN node.id as id, node.doc_type as doc_type, node.title as title, node.content as content, node.source as source, score ORDER BY score DESC rru🔍 Document search (allrgz resultadosu&❌ Error en vector_search_documents: N) rrtr!rFrbr"rjr7r8) rrcr"rdrsrSr/rmr9s rvector_search_documentsz#Neo4jClient.vector_search_documentsts# //;;JGIWX  (1hQVgloqgqr(15A((7G KK01BU0C3s7|nT_` aN  LLA!E FI s#3B AB B6B1+B61B6c ||j|j}|rdn|j}tj d|rdn|d|rKtj dt |d|D]}d|d< |j |}d |d <d|d <|Stj d |d |j}|jd}d}d} |D]4}|jd} | s| dz } | |vr d|d<|dz }+d|d<|dz }6tj d|d|d|j |}|j} |j} d|d <||d <||d<||d<| |d<d|d|d| d|jddd |d<tj d|d|S)u~ Sincroniza productos desde XML a Neo4j. Lógica de sincronización: - Si la BD está vacía: sube todos con estado 'nuevo' - Si ya hay datos: 1. Marca todos los productos existentes como 'pendiente' 2. Por cada producto del XML: - Si no existe en Neo4j → estado 'nuevo' (crear) - Si ya existe en Neo4j → estado 'normal' (actualizar) 3. Los que quedaron en 'pendiente' → estado 'eliminado' Args: products: Lista de productos parseados del XML Returns: Dict con resultado de la sincronización ru📊 Estado BD: uVACÍAz productos existentesu🆕 BD vacía - Subiendo z productos como NUEVOrr initial_loadmoder3u 🔄 BD con u2 productos - Iniciando sincronización incrementalrr normalu📦 XML procesado: z nuevos, z existentesincremental_syncnewrrzSync incremental: z actualizados, z eliminados, rrru,✅ Sincronización incremental completada: ) rrrr!r"rjrr rrkrr) rr~is_empty current_countrr_ existing_ids new_count updated_countrr deleted_countphysically_deleteds rsync_products_from_xmlz"Neo4jClient.sync_products_from_xmls4& ((*))+%4+>+>+@  &8xM?RgAh&ijk  KK4S]OCXY Z$ 1)0 & 1..x8F+F6N'(F# $M KK,}o5gh i 335L  - - /IMF# #$[[6 !aKF--5GM*!Q&M.5GM*NI # KK.yk=/Q\] ^..x8F!88:M"&!@!@!B /F6N'4F# $%F5M -F9  2F9 $YKy%&mFJJx4K3LHV 9  KKFviGXFYZ [Mr)NN))rrFgffffff?NNFF)r*))rFNr)NrF)6__name__ __module__ __qualname____doc__rArr[r#r(r:r6rRrrrrrbintrorurboolrrrrrrrrrrrrrrrrrrKr rrrrrrr!r$r&r0r4r7rErrr r s= 5BSB c('.6 3HT#s(^4L_ghk_lx|~BCFHKCK~LyM8$3$s$4PTUXZ]U]P^K_$LDDSDdSVX[S[nI]DF)-!%!&#(vMvMvM vM  vM "#Y vM;vMvM!vM c3hvMp.Dc.D#.DtTXY\^aYaTbOc.D`))C)QUVZ[^`c[cVdQe)V-c-d38n-& U# Uc U$tCQTH~BV U)-,,"#Y, d38n  ,\3tDcN7K*C E$s) E = =# = @3 @C @CDQUVY[^V^Q_L`F55 *4*CSDQVK(| T$sCx.-A| d3PS8n| |' T#s(^' R9S9 c"&C(c&cs2b.c,55s5   F tDcN/CF SRUXF PT#s(^(<Y\aefiknfnaoB/#//TW/`deijmorjres`t/fWtDcN/CWSRUXWrr )rKloggingrrtypingrrrrneo4jrconfig.settingsr getLoggerrHr!setLevelERRORrr neo4j_clientrNrrrWsq  ,,$   8 $'(11'--@hhX'} r