{"id":4856,"date":"2025-07-23T10:37:43","date_gmt":"2025-07-23T17:37:43","guid":{"rendered":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/"},"modified":"2025-07-23T10:37:43","modified_gmt":"2025-07-23T17:37:43","slug":"polaris-multi-agent-conversational-ai","status":"publish","type":"post","link":"https:\/\/www.couchbase.com\/blog\/pt\/polaris-multi-agent-conversational-ai\/","title":{"rendered":"Polaris: Intelig\u00eancia de Dados Conversacional Impulsionada por IA para Empresas Atrav\u00e9s de uma Arquitetura Multi-Agente"},"content":{"rendered":"<p class=\"wp-block-paragraph\">No ambiente acelerado de hoje, a capacidade de acessar, compreender e agir rapidamente sobre os dados n\u00e3o \u00e9 mais um luxo, \u00e9 uma necessidade. No entanto, muitas organiza\u00e7\u00f5es descobrem que, embora sejam ricas em dados, extrair insights oportunos e acion\u00e1veis continua sendo um desafio significativo, particularmente para usu\u00e1rios de neg\u00f3cios n\u00e3o t\u00e9cnicos.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Al\u00e9m disso, os usu\u00e1rios t\u00e9cnicos precisam entender seus dados para saber quais consultas construir para obter os resultados, o que exige muito tempo e esfor\u00e7o e n\u00e3o est\u00e1 a apenas um clique de dist\u00e2ncia, permitindo que o usu\u00e1rio pergunte o que quiser em linguagem natural simples.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Al\u00e9m disso, o usu\u00e1rio ainda precisa gastar tempo entendendo os dados, mesmo com visualiza\u00e7\u00f5es em vigor. Muitas vezes h\u00e1 uma d\u00favida sobre o \u201cporqu\u00ea\u201d quando os dados s\u00e3o apresentados, e dados sem o crucial \u201cporqu\u00ea\u201d deixam uma lacuna entre a apresenta\u00e7\u00e3o dos dados e a verdadeira compreens\u00e3o. Em ess\u00eancia, o self-service de business analytics continua elusivo.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">O que \u00e9 a Polaris?<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Polaris<\/b> \u00e9 uma interface conversacional alimentada por IA multi-agente constru\u00edda para analisar dados em nosso banco de dados operacional Couchbase. O Polaris aproveita uma arquitetura multi-agente que permite aos usu\u00e1rios interagir com seus dados corporativos por meio de uma interface intuitiva e conversacional, transformando an\u00e1lises de dados complexas em um di\u00e1logo simples. Por exemplo, se uma empresa possui dados globais de vendas corporativas em v\u00e1rias regi\u00f5es e linhas de produtos, e um analista de neg\u00f3cios deseja entender <em>\u201cPor que as vendas do segundo trimestre diminu\u00edram na regi\u00e3o Nordeste para o Produto X?\u201d<\/em>, nosso aplicativo pode executar autonomamente todo o fluxo de trabalho de an\u00e1lise.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Ele recupera e filtra dados de vendas relevantes por regi\u00e3o, produto e per\u00edodo de tempo, compara tend\u00eancias de desempenho entre regi\u00f5es ou produtos compar\u00e1veis, visualiza padr\u00f5es e anomalias importantes e gera um relat\u00f3rio narrativo resumindo as causas raiz, como redu\u00e7\u00e3o de gastos com promo\u00e7\u00f5es, problemas de disponibilidade de estoque ou uma mudan\u00e7a no comportamento do cliente. Para tornar tudo ainda mais interessante, o analista de neg\u00f3cios pode fazer uma pergunta de acompanhamento para entender, em detalhes, alguma parte do relat\u00f3rio ou talvez pedir mais visualiza\u00e7\u00f5es, etc., permitindo assim uma tomada de decis\u00e3o r\u00e1pida baseada em dados.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Agora, vamos abordar o elefante na sala \u2013 os Agentes de IA.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">O que s\u00e3o Agentes de IA e quais s\u00e3o as suas capacidades?<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Agentes de IA s\u00e3o sistemas aut\u00f4nomos impulsionados por intelig\u00eancia artificial, envolvendo tipicamente M\u00f3dulos de Grande Linguagem (LLMs) que podem realizar tarefas, tomar decis\u00f5es e interagir com ambientes do mundo real, muitas vezes sem supervis\u00e3o humana constante. Ao contr\u00e1rio de chatbots tradicionais ou programas baseados em regras, os agentes de IA tamb\u00e9m aprendem com sua experi\u00eancia. O objetivo de um agente \u00e9 fazer tudo o que um operador humano faz de forma aut\u00f4noma e autom\u00e1tica. Ainda \u00e9 uma meta distante, mas a ind\u00fastria de IA est\u00e1 progredindo em dire\u00e7\u00e3o a ela. Agora, vamos analisar as capacidades dos Agentes de IA:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-large wp-image-17344\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image8-1024x241-1.png\" alt=\"What are AI Agents and what are their capabilities? \" width=\"900\" height=\"212\"><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u00a0<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Plano do Agente: Resolu\u00e7\u00e3o de Problemas Passo a Passo<\/b><b><br>\n<\/b> Os agentes de IA dividem tarefas complexas em etapas claras e gerenci\u00e1veis \u2014 identificando o problema, executando cada fase e se ajustando conforme necess\u00e1rio. Em sistemas multi-agentes, cada agente pode assumir uma tarefa espec\u00edfica, permitindo uma resolu\u00e7\u00e3o de problemas eficiente e coordenada.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Consci\u00eancia de Contexto: Gerenciamento de Mem\u00f3ria e Rastreamento de Estado<\/b><b><br>\n<\/b> Os agentes mant\u00eam o contexto entre as intera\u00e7\u00f5es, lembrando-se de entradas passadas e adaptando-se a fluxos de trabalho em andamento. Esse rastreamento de estado cria experi\u00eancias de usu\u00e1rio mais naturais, consistentes e inteligentes.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Uso de Ferramentas: Estendendo as Capacidades dos Agentes<\/b><b><br>\n<\/b> Os agentes podem interagir com ferramentas externas \u2014 APIs, bancos de dados, scripts \u2014 para realizar a\u00e7\u00f5es reais, n\u00e3o apenas oferecer sugest\u00f5es. Isso os transforma de assistentes passivos em executores ativos dentro de fluxos de trabalho.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Aprender com Dados Passados: Adaptando-se ao Longo do Tempo<\/b><b><br>\n<\/b> Ao analisar dados hist\u00f3ricos e comportamento, os agentes melhoram com o tempo \u2014 antecipando as necessidades do usu\u00e1rio, refinando respostas e otimizando fluxos de trabalho com base nos padr\u00f5es de uso.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">O que \u00e9 um sistema multi-agente (MAS)?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Arquitetura Multiagente \u00e9 um design de sistema onde m\u00faltiplos agentes independentes trabalham juntos para resolver problemas ou executar tarefas. Cada agente tem seu pr\u00f3prio papel, como coletar dados, analisar informa\u00e7\u00f5es ou tomar decis\u00f5es. Esses agentes se comunicam e colaboram para alcan\u00e7ar um objetivo comum, tornando o sistema mais organizado; isso \u00e9 exatamente como uma equipe onde cada membro faz um trabalho espec\u00edfico, mas todos trabalham em dire\u00e7\u00e3o ao mesmo resultado! N\u00f3s fizemos uso da Arquitetura Multiagente para o Polaris.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Por que a mudan\u00e7a na arquitetura de agente \u00fanico?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Um \u00fanico agente de IA opera de forma independente, lidando com tarefas espec\u00edficas autonomamente. Isso funciona bem para aplica\u00e7\u00f5es diretas, como um sistema de Gera\u00e7\u00e3o Aumentada por Recupera\u00e7\u00e3o (RAG), onde um agente responde a consultas de usu\u00e1rios com base em um LLM e em uma base de conhecimento. No entanto, em aplica\u00e7\u00f5es pr\u00e1ticas, as intera\u00e7\u00f5es dos usu\u00e1rios raramente s\u00e3o simples. Elas frequentemente envolvem l\u00f3gica complexa, racioc\u00ednio em v\u00e1rias etapas e a necessidade de trabalhar em modelos de dados din\u00e2micos e requisitos de neg\u00f3cios em evolu\u00e7\u00e3o. Nesse ponto, os sistemas de agente \u00fanico come\u00e7am a atingir limites de desempenho e escalabilidade. Eles podem falhar ao encadear m\u00faltiplas opera\u00e7\u00f5es, adaptar-se a mudan\u00e7as de esquema ou coordenar fluxos de trabalho sutis.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Por que as Arquiteturas Multi-Agente (MAS) funcionam?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">A separa\u00e7\u00e3o de conceitos inerente ao projeto de sistemas multi-agente (MAS) resulta em sistemas mais robustos e f\u00e1ceis de manter. Cada agente foca em sua tarefa espec\u00edfica, reduzindo a complexidade e facilitando a identifica\u00e7\u00e3o e resolu\u00e7\u00e3o de problemas. Essa abordagem brilha em cen\u00e1rios como o controle de ve\u00edculos aut\u00f4nomos, onde agentes separados lidam com a navega\u00e7\u00e3o, detec\u00e7\u00e3o de obst\u00e1culos e din\u00e2mica do ve\u00edculo, permitindo um desenvolvimento e solu\u00e7\u00e3o de problemas focados em cada \u00e1rea.<\/p>\n\n\n\n<h4 class=\"wp-block-heading\">Supervisor vs Sistemas Multiagentes em Rede<\/h4>\n\n\n\n<center><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-17345 aligncenter\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image1-2-1024x533-1.png\" alt=\"Why Does Multi-Agent Architecture (MAS) work?\" width=\"600\" height=\"312\"><\/center>\n\n\n\n<p class=\"wp-block-paragraph\">Um supervisor central controla a atribui\u00e7\u00e3o de tarefas e monitora subagentes, que possuem autonomia limitada. Isso permite um gerenciamento mais f\u00e1cil, decis\u00f5es otimizadas e resolu\u00e7\u00e3o de conflitos. No entanto, \u00e9 propenso a falhas de ponto \u00fanico e tem dificuldades com a escalabilidade. A flexibilidade \u00e9 limitada em ambientes din\u00e2micos ou em r\u00e1pida mudan\u00e7a.\n<\/p>\n\n\n\n<h4 class=\"wp-block-heading\">Arquitetura Descentralizada (Ponto a Ponto) de M\u00faltiplos Agentes<\/h4>\n\n\n\n<center><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-17346 aligncenter\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image3-2-1024x860-1.png\" alt=\"Decentralized (Peer-to-Peer) Multi-Agent Architecture\" width=\"600\" height=\"504\"><\/center>\n\n\n\n<p class=\"wp-block-paragraph\">Os agentes agem de forma independente, e todos os agentes t\u00eam a capacidade de interagir entre si. Ele escala bem e \u00e9 resiliente a falhas, mas carece de supervis\u00e3o centralizada. Isso leva a uma coordena\u00e7\u00e3o complexa, maior overhead de comunica\u00e7\u00e3o e desafios na resolu\u00e7\u00e3o de conflitos, sendo tamb\u00e9m mais dif\u00edcil garantir a coer\u00eancia global.\n<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Escolhemos: Arquitetura Baseada em Supervisor<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Nosso aplicativo faz uso do <a href=\"https:\/\/langchain-ai.github.io\/langgraph\/tutorials\/multi_agent\/agent_supervisor\/\" target=\"_blank\" rel=\"noopener\"><b>Agente Supervisor do LangGraph<\/b><\/a>:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>Racioc\u00ednio centralizado, consist\u00eancia e coer\u00eancia\n<ul>\n<li aria-level=\"2\">O racioc\u00ednio complexo se beneficia de ter uma vis\u00e3o global dos dados, da inten\u00e7\u00e3o do usu\u00e1rio e do contexto. O supervisor pode manter uma l\u00f3gica coerente em v\u00e1rias etapas.<\/li>\n<li aria-level=\"2\">Um \u00fanico ponto de tomada de decis\u00e3o garante que os resultados estejam alinhados (por exemplo, o gr\u00e1fico corresponde \u00e0 explica\u00e7\u00e3o, o resumo reflete a an\u00e1lise).<\/li>\n<li aria-level=\"2\">O controle central permite a aloca\u00e7\u00e3o din\u00e2mica de tarefas para subagentes especializados (por exemplo, gerador de gr\u00e1ficos, agente de consulta). Evita aduplica\u00e7\u00e3o de esfor\u00e7o e otimiza o uso de recursos.<\/li>\n<\/ul>\n<\/li>\n\n\n<li>Tratamento de erros, recupera\u00e7\u00e3o e escalabilidade mais f\u00e1ceis\n<ul>\n<li aria-level=\"2\">Erros podem ser detectados e gerenciados de forma centralizada. O supervisor pode tentar novamente as tarefas, reatribuir fun\u00e7\u00f5es ou gerar respostas de conting\u00eancia.<\/li>\n<li aria-level=\"2\">\u00a0O controle central permite a aloca\u00e7\u00e3o din\u00e2mica de tarefas para subagentes especializados (por exemplo, gerador de gr\u00e1ficos, agente de consulta). Evita aduplica\u00e7\u00e3o de esfor\u00e7o e otimiza o uso de recursos.<\/li>\n<li aria-level=\"2\">Mais f\u00e1cil adicionar, substituir ou atualizar subagentes sem redesenhar o sistema inteiro.<\/li>\n<\/ul>\n<\/li>\n\n<\/ol>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">N\u00facleo Polaris<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Em sua ess\u00eancia, o Polaris faz uso de uma rede de agentes de IA especializados, cada um otimizado para diferentes aspectos do ciclo de vida da intera\u00e7\u00e3o de dados. Agora, vamos entender quais s\u00e3o os componentes e a arquitetura geral de m\u00faltiplos agentes de alto n\u00edvel do Polaris com a ajuda de um exemplo:<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter size-large wp-image-17347\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image7-1024x499-1.png\" alt=\"polaris core\" width=\"900\" height=\"439\"><\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Compreens\u00e3o e orquestra\u00e7\u00e3o: Agente Supervisor<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">A <a href=\"https:\/\/github.com\/langchain-ai\/langgraph-supervisor-py\" target=\"_blank\" rel=\"noopener\">Agente Supervisor<\/a> atua como o controlador central e o orquestrador inteligente do sistema multi-agente.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>Fun\u00e7\u00f5es:<\/b><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><b>An\u00e1lise de Inten\u00e7\u00e3o:<\/b> Analisar a entrada do usu\u00e1rio e extrair inten\u00e7\u00f5es e par\u00e2metros relacionados a tarefas.\n<ul>\n<li aria-level=\"2\">Exemplo: O usu\u00e1rio pergunta: \u201cPor que as vendas totais de eletr\u00f4nicos no 1\u00ba trimestre de 2024 ca\u00edram na regi\u00e3o da APAC.\u201d O Agente Supervisor analisa isso para identificar Inten\u00e7\u00e3o: \u201cRaz\u00e3o - an\u00e1lise causal para queda nas vendas\u201d, Categoria de Produto: \u201cEletr\u00f4nicos\u201d, Per\u00edodo de Tempo: \u201c1\u00ba trimestre de 2024\u201d, Regi\u00e3o: \u201cAPAC\u201d.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>L\u00f3gica de Roteamento de Agentes:<\/b> Implementa um mecanismo de decis\u00e3o ou camada de orquestra\u00e7\u00e3o baseada em regras para direcionar tarefas aos agentes apropriados.\n<ul>\n<li aria-level=\"2\">Com base na inten\u00e7\u00e3o analisada \u201can\u00e1lise causal para queda nas vendas\u201d, o Agente Supervisor decide primeiro encaminhar a tarefa para o Especialista em Consultas para buscar dados de vendas, depois para o Especialista em Gr\u00e1ficos para visualiza\u00e7\u00e3o, em seguida para o agente de racioc\u00ednio para identifica\u00e7\u00e3o causal e, finalmente, para o Especialista em Relat\u00f3rios para resumo.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Gerenciamento de Contexto:<\/b> Mant\u00e9m o contexto e o estado da conversa global.<\/li>\n\n\n<li><b>Tratamento de Erros e Recupera\u00e7\u00e3o:<\/b> Monitora o sucesso\/falha de tarefas e pode reatribuir ou reformular subtarefas com base no feedback do agente.\n<ul>\n<li aria-level=\"2\">Exemplo<b>:<\/b> Se o Especialista em Consultas relatar que a coluna solicitada, <em>Tipo_de_Produto,<\/em>\u00a0n\u00e3o existe no esquema, o Agente Supervisor pode redirecionar a solicita\u00e7\u00e3o para o Especialista em Racioc\u00ednio para sugerir colunas relevantes alternativas ou informar o usu\u00e1rio sobre os dados ausentes.<\/li>\n<\/ul>\n<\/li>\n\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Extra\u00e7\u00e3o de dados relevantes: Especialista em Consulta<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">O Especialista de Consulta traduz perguntas em linguagem natural para SQL++, buscando assim os dados necess\u00e1rios.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>\u00a0Fun\u00e7\u00f5es:<\/b><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><b>Infer\u00eancia de Esquema e Anota\u00e7\u00f5es<\/b><strong>:<\/strong> Infere o esquema de dados usando o comando SQL++ INFER, que busca os nomes das colunas, o tipo de dado das colunas e documentos de amostra; isso, junto com a ajuda de anota\u00e7\u00f5es, ajuda a entender os dados, relacionamentos de tabelas, tipos de dados e restri\u00e7\u00f5es.\n<ul>\n<li aria-level=\"2\">Quando o SQL++ INFER \u00e9 executado em uma cole\u00e7\u00e3o, ele pode identificar um campo simplesmente como <em>\u201camount\u201d: N\u00daMERO<\/em>. Sem mais contexto, o Especialista em Consultas n\u00e3o saberia se isso se refere a <em>valor_da_venda<\/em>, <em>valor_do_desconto<\/em>, ou <em>quantidade<\/em>. No entanto, por meio de anota\u00e7\u00f5es, \u00e9 explicitamente dito ao Polaris: <em>\u201cquantidade\u201d<\/em> campo em \u2018<em>vendas corporativas<\/em>\u2018 a cole\u00e7\u00e3o representa \u2018<em>valor total de vendas<\/em>\u2018 para uma transa\u00e7\u00e3o. Esta anota\u00e7\u00e3o \u00e9 crucial porque quando o usu\u00e1rio pergunta \u201c<em>vendas totais<\/em>\u201c, o Query Expert agora mapeia com seguran\u00e7a <em>vendas<\/em> para o <em>quantidade<\/em>\u00a0campo, gerando corretamente <em>SOMA(amount)<\/em>.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Canonicaliza\u00e7\u00e3o de Entrada<\/b><strong>:<\/strong> Transforma a entrada de linguagem natural original do usu\u00e1rio em uma forma mais detalhada, inequ\u00edvoca e estruturada. Isso ajuda a ferramenta IQ a compreender melhor a tarefa.\n<ul>\n<li aria-level=\"2\">Exemplo: Entrada do usu\u00e1rio: \u201c<em>vendas no m\u00eas passado<\/em>.\u201d Entrada canonizada: \u201c<em>Recuperar o valor total de vendas para a categoria \u201cEletr\u00f4nicos\u201d nos \u00faltimos 30 dias a partir da data atual.<\/em>\u201c<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Tradu\u00e7\u00e3o de NL para SQL++:<\/b> Chamada para a ferramenta IQ para converter NL em SQL++<\/li>\n\n\n<li><b>Verifica\u00e7\u00f5es de Qualidade de Dados e Recupera\u00e7\u00e3o de Erros: <\/b>o agente inspeciona em busca de valores nulos e outros problemas de integridade de dados que possam afetar a interpreta\u00e7\u00e3o. Se a qualidade dos dados for baixa (por exemplo, todos nulos em uma coluna), o agente reformula a consulta ou retorna um aviso para interven\u00e7\u00e3o do usu\u00e1rio. Com base no diagn\u00f3stico de erros, o agente ajusta automaticamente a consulta (por exemplo, corrige nomes de colunas ou limita o tamanho dos resultados) e tenta a execu\u00e7\u00e3o novamente de forma inteligente.\n<ul>\n<li aria-level=\"2\">Exemplo: Se o <em>valor_da_venda<\/em> column might contain nulls, the Query Expert automatically adds: <em>AND sale_amount IS NOT NULL<\/em> to the generated query to ensure accurate sum calculations.<b>\u00a0<\/b><\/li>\n<\/ul>\n<\/li>\n\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Insight generation: Charting Expert<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\"><b><br>\n<\/b> Responsible for converting structured query results into meaningful visual representations, tailored to the nature of the data and user query.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>\u00a0Fun\u00e7\u00f5es:<\/b><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><b>Chart Selection Logic:<\/b> Uses rule-based heuristics\u00a0 to select appropriate chart types based on data characteristics (e.g., dimensions, metrics, time series).\n<ul>\n<li aria-level=\"2\">Example: Based on the rules given in the prompt and the type of data, the expert will choose an appropriate chart, for example if it is sales and time series data where we need to identify some trend, it will select a\u00a0 line chart.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Dynamic Visualization Generation:<\/b> Constructs visualizations using libraries like Plotly and Seaborn.<\/li>\n\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Reporting and summarization: Report Expert<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Compiles insights, visualizations, and context into structured reports.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>\u00a0Fun\u00e7\u00f5es:<\/b><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><b>Content Aggregation:<\/b> Automatically summarizes query results, embeds visualizations, the methodology and includes metadata (e.g., data sources, query parameters).<\/li>\n\n\n<li><b>Versioning &amp; Audit Logs:<\/b> Optionally integrates version control and logging for compliance and traceability of generated reports.<\/li>\n\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Explanation and reasoning: Reasoning Expert<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Provides causal reasoning, trend analysis, and hypothesis generation by interpreting data insights through the lens of domain knowledge and logical inference.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><b>\u00a0Fun\u00e7\u00f5es:<\/b><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><b>LLM-based Reasoning:<\/b> Leverages LLMs to reason over data results, uncover latent patterns, and generate explanatory narratives.<\/li>\n\n\n<li><b>Contextual Augmentation:<\/b> Utilizes domain-specific knowledge extracted from the user\u2019s database\u00a0 to provide grounded explanations.<\/li>\n\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">Workflow<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">The Polaris platform is designed to turn natural language questions into intelligent, multi-modal insights by orchestrating a team of specialized agents. Here\u2019s how the workflow unfolds:<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><b><b>Polaris System Initialization<br>\n<\/b><\/b><img loading=\"lazy\" decoding=\"async\" class=\"aligncenter wp-image-17348 size-large\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image2-2-1024x631-1.png\" alt=\"Polaris platform for natural language\" width=\"900\" height=\"555\">The user begins by selecting the relevant <i>bucket, scope, collection, and metadata collection<\/i>. Based on this context, Polaris initializes specialized agents and uses the schema, metadata, and sample data to prompt an LLM, which generates example questions to guide user exploration.\n<div class=\"mceTemp\"><\/div>\n<\/li>\n\n\n<li><b><b>Natural Language Interaction<br>\n<\/b><\/b><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-17349 size-large aligncenter\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image5-1024x563-1.png\" alt=\"Polaris System Initialization\" width=\"900\" height=\"495\">Users interact with Polaris through a simple chat interface, posing questions in natural language. This removes the need for manual query writing or schema exploration.\n<div class=\"mceTemp\"><\/div>\n<\/li>\n\n\n<li><b><b>Intelligent Query Processing<br>\n<\/b><\/b><center>\n<figure id=\"attachment_17350\" aria-describedby=\"caption-attachment-17350\" style=\"width: 900px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-17350 size-large\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image4-1024x532-1.jpg\" alt=\"Natural Language Interaction\" width=\"900\" height=\"468\"><figcaption id=\"caption-attachment-17350\" class=\"wp-caption-text\">High-Level Design Diagram<\/figcaption><\/figure>\n<p>\u00a0<\/p>\n<p>\u00a0<\/p>\n<p>\u00a0<\/p>\n<p><\/p><\/center>A <i>Agente Supervisor<\/i> receives the user&#8217;s query and assigns responsibilities to specialized agents:\n<ul>\n<li aria-level=\"1\">A <b>Query Expert<\/b> handles core data access tasks: inferring the schema, translating the natural language query into SQL++ using a generator tool, and executing the query.<\/li>\n<li aria-level=\"1\">Tools supporting the Query Expert include the <i>Schema Inference Tool<\/i>, <i>SQL++ Generator Tool<\/i>, e <i>Query Execution Tool<\/i>.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Multi-Faceted Response Generation<\/b><b><br>\n<\/b> Based on the results of the initial query, the Supervisor coordinates:\n<ul>\n<li aria-level=\"1\">A <b>Charting Expert<\/b>, which creates data visualizations via a <i>Chart Generator Tool<\/i>.<\/li>\n<li aria-level=\"1\">A <b>Report Expert<\/b>, responsible for generating textual summaries using a <i>Report Generator Tool<\/i>.<\/li>\n<li aria-level=\"1\">A <b>Reasoning Expert<\/b>, which adds context, rationale, or further explanations to enrich the response.<\/li>\n<\/ul>\n<\/li>\n\n\n<li><b>Comprehensive Insight Delivery<\/b><b><br>\n<\/b> Polaris synthesizes the structured query results, visual outputs, and narrative explanations into a cohesive, user-friendly response. This multi-modal insight is delivered back through the chat interface, combining clarity, depth, and interactivity.<\/li>\n\n\n<li><b>Iterative Exploration<\/b><b><br>\n<\/b> Users are encouraged to ask follow-up questions. Since the system retains context and state across the session, the agent network can build on previous interactions to support deep, iterative data exploration.<\/li>\n\n<\/ol>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Usage of ReAct agents<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">What is a ReAct agent?<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">&#8220;A ReAct agent is an AI agent that uses the \u201creasoning and acting\u201d (ReAct) framework to combine chain of thought (CoT) reasoning with external tool use. The ReAct framework improves the ability of a large language model (LLM) to handle complex tasks and decision-making in agentic workflows.&#8221;\u2014<a href=\"https:\/\/www.ibm.com\/think\/topics\/react-agent\" target=\"_blank\" rel=\"noopener\">Dave Bergmann, IBM<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><figure id=\"attachment_17351\" aria-describedby=\"caption-attachment-17351\" style=\"width: 600px\" class=\"wp-caption aligncenter\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-17351\" src=\"https:\/\/www.couchbase.com\/wp-content\/uploads\/sites\/5\/2026\/05\/image6-1024x576-1.png\" alt=\"Working of a ReAct Agent\" width=\"600\" height=\"338\"><figcaption id=\"caption-attachment-17351\" class=\"wp-caption-text\">Working of a ReAct Agent<\/figcaption><\/figure><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Unlike traditional Artificial Intelligence (AI) systems, ReAct agents don\u2019t separate decision-making from task execution. This framework inherently creates a feedback loop in which the model problem-solves by iteratively repeating this interleaved <i>thought-action-observation<\/i> process. We use the <a href=\"https:\/\/langchain-ai.github.io\/langgraph\/reference\/agents\/#langgraph.prebuilt.chat_agent_executor.create_react_agent\" target=\"_blank\" rel=\"noopener\">inbuilt LangGraph ReAct<\/a>\u00a0framework in our application, and each of the expert is modeled as a ReAct agent.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The unseen architects: the power of efficient prompts in AI-driven data analysis<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">In the realm of data analysis, the spotlight often shines on algorithms, statistical models, and visualization techniques. However, behind every insightful chart, every well-structured report, and every data-driven conclusion lies a crucial , unseen aspect: the prompt.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>To-Do List Prompting<\/strong><b><br>\n<\/b> To-do list prompting gives the model a persistent, structured task list that it refers to at every step. Instead of relying on memory or previous messages, the full plan is injected into each prompt. Therefore the agent has a clear understanding of all tasks it has to check off . This prevents drifting, repetition, or skipping steps.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Identity Prompting<\/strong><b><br>\n<\/b> Identity prompting tells the model <i>what it is<\/i>, not just <i>what it should do<\/i>. This establishes a consistent role or persona that influences how the model behaves and responds. <i>Prompts like &#8220;You are very proficient in data visualization tasks.&#8221; <\/i>can instantly trigger domain-specific behavior\u2014clear, confident, and focused responses.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Self-Reflection Prompting<\/strong><b><br>\n<\/b> Self-reflection prompting instructs the model to evaluate its own output after completing a task. This allows the model to introspect and verify whether it has met the user&#8217;s goal, and make corrections if needed. In our application, we\u2019ve implemented self-reflection prompting within the <b>Query Expert agent<\/b>. After the SQL query is generated and executed the agent checks if all required data points are present.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Prompting is more of an art than a science\u2014there\u2019s no one-size-fits-all formula. However, by applying proven heuristics and clear task framing, we can guide models toward more accurate, useful, and context-aware outputs. The key is experimentation, iteration, and learning what works best in your specific application.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Demo of Polaris, the multi-agent conversational interface<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><iframe loading=\"lazy\" title=\"Polaris - Demo\" width=\"500\" height=\"281\" src=\"https:\/\/www.youtube.com\/embed\/E5ThhiwUASg?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe><\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Challenges and future work<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">Polaris represents a paradigm shift in how organizations can harness their data assets, especially through natural language interactions &#8211; enabling intuitive data discovery and significantly accelerating decision-making. A major advancement has been our development of a dynamic multi-agent\u00a0 architecture that adapts its approach based on the context and can work with diverse datasets.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">However, several challenges remain. One key area has been managing data annotations. Ensuring consistent and meaningful annotations across varied columns is critical to maintaining the quality of insights generated by AI agents. We could explore integrating with a global data catalog to make this easier.\u00a0 Another significant challenge is data cleanliness, while we mitigate some of these issues at the query level through conditional clauses and basic data cleaning\u2014there is still room for improvement in upstream data validation and preprocessing.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Additionally, handling large-scale data retrieval has been a technical hurdle. In real-world scenarios, retrieved datasets often exceed the context window limits of current large language models. To address this, we perform aggregation operations and generate visual summaries such as charts to provide high-level insights without overwhelming the model.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Looking ahead, future work will focus on enhancing annotation pipelines, improving data quality management, and exploring more efficient methods of summarization and multi-turn agent collaboration to scale Polaris even further.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Conclusion: ushering in a new era of data interaction<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Polaris is more than just a new tool,by combining the power of a multi-agent AI system with the simplicity of natural language conversation, Polaris democratizes data access, empowers business users, and accelerates the journey from data to decision. We believe Polaris will unlock significant value for our customers, fostering a more agile, data-informed, and competitive enterprise.<\/p>","protected":false},"excerpt":{"rendered":"<p>In today\u2019s fast-paced environment, the ability to swiftly access, understand, and act upon data is no longer a luxury\u200a, \u200ait\u2019s a necessity. However, many organizations find that while they are rich in data, deriving timely, actionable insights remains a significant challenge, particularly for non-technical business users. Also, technical users need to understand their data to [&hellip;]<\/p>\n","protected":false},"author":85655,"featured_media":4852,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"inline_featured_image":false,"_acf":"","footnotes":""},"categories":[920,598,775],"tags":[],"ppma_author":[992,993],"class_list":["post-4856","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-agentic-ai-apps","category-artificial-intelligence-ai","category-engineering"],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v28.3 (Yoast SEO v28.3) - https:\/\/yoast.com\/product\/yoast-seo-premium-wordpress\/ -->\n<title>Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture<\/title>\n<meta name=\"description\" content=\"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.couchbase.com\/blog\/pt\/polaris-multi-agent-conversational-ai\/\" \/>\n<meta property=\"og:locale\" content=\"pt_BR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture\" \/>\n<meta property=\"og:description\" content=\"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.couchbase.com\/blog\/pt\/polaris-multi-agent-conversational-ai\/\" \/>\n<meta property=\"og:site_name\" content=\"The Couchbase Blog\" \/>\n<meta property=\"article:published_time\" content=\"2025-07-23T17:37:43+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png\" \/>\n\t<meta property=\"og:image:width\" content=\"2400\" \/>\n\t<meta property=\"og:image:height\" content=\"1256\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Varuni H K - Software Engineer, Soham Sarkar - Software Engineer\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Varuni H K - Software Engineer\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"14 minutos\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/\"},\"author\":{\"name\":\"Varuni H K - Software Engineer\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#\\\/schema\\\/person\\\/a3d7b24c5c349f1b159d2b75f58f0981\"},\"headline\":\"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture\",\"datePublished\":\"2025-07-23T17:37:43+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/\"},\"wordCount\":2822,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/05\\\/blog-polaris-core.png\",\"articleSection\":[\"Agentic AI Applications\",\"Artificial Intelligence (AI)\",\"Engineering\"],\"inLanguage\":\"pt-BR\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/\",\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/\",\"name\":\"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/05\\\/blog-polaris-core.png\",\"datePublished\":\"2025-07-23T17:37:43+00:00\",\"description\":\"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#breadcrumb\"},\"inLanguage\":\"pt-BR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-BR\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#primaryimage\",\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/05\\\/blog-polaris-core.png\",\"contentUrl\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/05\\\/blog-polaris-core.png\",\"width\":2400,\"height\":1256},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/polaris-multi-agent-conversational-ai\\\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/\",\"name\":\"The Couchbase Blog\",\"description\":\"Couchbase, the NoSQL Database\",\"publisher\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"pt-BR\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#organization\",\"name\":\"The Couchbase Blog\",\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-BR\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/06\\\/logo.svg\",\"contentUrl\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/wp-content\\\/uploads\\\/sites\\\/5\\\/2026\\\/06\\\/logo.svg\",\"width\":\"1024\",\"height\":\"1024\",\"caption\":\"The Couchbase Blog\"},\"image\":{\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#\\\/schema\\\/logo\\\/image\\\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/#\\\/schema\\\/person\\\/a3d7b24c5c349f1b159d2b75f58f0981\",\"name\":\"Varuni H K - Software Engineer\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-BR\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g843434b5ffa6bdd2518608db8ee59378\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g\",\"caption\":\"Varuni H K - Software Engineer\"},\"url\":\"https:\\\/\\\/www.couchbase.com\\\/blog\\\/pt\\\/author\\\/varunihk\\\/\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Polaris: Intelig\u00eancia de Dados Conversacional Impulsionada por IA para Empresas Atrav\u00e9s de uma Arquitetura Multi-Agente","description":"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.couchbase.com\/blog\/pt\/polaris-multi-agent-conversational-ai\/","og_locale":"pt_BR","og_type":"article","og_title":"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture","og_description":"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.","og_url":"https:\/\/www.couchbase.com\/blog\/pt\/polaris-multi-agent-conversational-ai\/","og_site_name":"The Couchbase Blog","article_published_time":"2025-07-23T17:37:43+00:00","og_image":[{"width":2400,"height":1256,"url":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png","type":"image\/png"}],"author":"Varuni H K - Software Engineer, Soham Sarkar - Software Engineer","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Varuni H K - Software Engineer","Est. reading time":"14 minutos"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#article","isPartOf":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/"},"author":{"name":"Varuni H K - Software Engineer","@id":"https:\/\/www.couchbase.com\/blog\/#\/schema\/person\/a3d7b24c5c349f1b159d2b75f58f0981"},"headline":"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture","datePublished":"2025-07-23T17:37:43+00:00","mainEntityOfPage":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/"},"wordCount":2822,"commentCount":0,"publisher":{"@id":"https:\/\/www.couchbase.com\/blog\/#organization"},"image":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#primaryimage"},"thumbnailUrl":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png","articleSection":["Agentic AI Applications","Artificial Intelligence (AI)","Engineering"],"inLanguage":"pt-BR","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/","url":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/","name":"Polaris: Intelig\u00eancia de Dados Conversacional Impulsionada por IA para Empresas Atrav\u00e9s de uma Arquitetura Multi-Agente","isPartOf":{"@id":"https:\/\/www.couchbase.com\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#primaryimage"},"image":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#primaryimage"},"thumbnailUrl":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png","datePublished":"2025-07-23T17:37:43+00:00","description":"Polaris leverages a multi-agent architecture that enable users to interact with their enterprise data through an intuitive, conversational interface.","breadcrumb":{"@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#breadcrumb"},"inLanguage":"pt-BR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/"]}]},{"@type":"ImageObject","inLanguage":"pt-BR","@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#primaryimage","url":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png","contentUrl":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/blog-polaris-core.png","width":2400,"height":1256},{"@type":"BreadcrumbList","@id":"https:\/\/www.couchbase.com\/blog\/polaris-multi-agent-conversational-ai\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.couchbase.com\/blog\/"},{"@type":"ListItem","position":2,"name":"Polaris: AI-Powered Conversational Data Intelligence for the Enterprise Through a Multi-Agent Architecture"}]},{"@type":"WebSite","@id":"https:\/\/www.couchbase.com\/blog\/#website","url":"https:\/\/www.couchbase.com\/blog\/","name":"O Blog da Couchbase","description":"Couchbase, o Banco de Dados NoSQL","publisher":{"@id":"https:\/\/www.couchbase.com\/blog\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.couchbase.com\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"pt-BR"},{"@type":"Organization","@id":"https:\/\/www.couchbase.com\/blog\/#organization","name":"O Blog da Couchbase","url":"https:\/\/www.couchbase.com\/blog\/","logo":{"@type":"ImageObject","inLanguage":"pt-BR","@id":"https:\/\/www.couchbase.com\/blog\/#\/schema\/logo\/image\/","url":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/06\/logo.svg","contentUrl":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/06\/logo.svg","width":"1024","height":"1024","caption":"The Couchbase Blog"},"image":{"@id":"https:\/\/www.couchbase.com\/blog\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/www.couchbase.com\/blog\/#\/schema\/person\/a3d7b24c5c349f1b159d2b75f58f0981","name":"Varuni H K - Software Engineer","image":{"@type":"ImageObject","inLanguage":"pt-BR","@id":"https:\/\/secure.gravatar.com\/avatar\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g843434b5ffa6bdd2518608db8ee59378","url":"https:\/\/secure.gravatar.com\/avatar\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/44319ba3fee522120cb62bb9f7df629984cc77850ba59c0eeeec760fb80f11b8?s=96&d=mm&r=g","caption":"Varuni H K - Software Engineer"},"url":"https:\/\/www.couchbase.com\/blog\/pt\/author\/varunihk\/"}]}},"acf":[],"authors":[{"term_id":992,"user_id":85655,"is_guest":0,"slug":"varunihk","display_name":"Varuni H K - Software Engineer","avatar_url":{"url":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/Screenshot-2025-07-23-at-11.40.18-AM-e1753292627397-1.png","url2x":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/Screenshot-2025-07-23-at-11.40.18-AM-e1753292627397-1.png"},"author_category":"1","first_name":"Varuni","last_name":"H K - Software Engineer","user_url":"","job_title":"","description":""},{"term_id":993,"user_id":85656,"is_guest":0,"slug":"sohamsarkar","display_name":"Soham Sarkar - Software Engineer","avatar_url":{"url":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/T024FJS4M-U084SMW3ZHV-950821ba6dbf-512-2.jpeg","url2x":"https:\/\/www.couchbase.com\/blog\/wp-content\/uploads\/sites\/5\/2026\/05\/T024FJS4M-U084SMW3ZHV-950821ba6dbf-512-2.jpeg"},"author_category":"1","first_name":"Soham","last_name":"Sarkar - Software Engineer","user_url":"","job_title":"","description":""}],"_links":{"self":[{"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/posts\/4856","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/users\/85655"}],"replies":[{"embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/comments?post=4856"}],"version-history":[{"count":0,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/posts\/4856\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/media\/4852"}],"wp:attachment":[{"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/media?parent=4856"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/categories?post=4856"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/tags?post=4856"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.couchbase.com\/blog\/pt\/wp-json\/wp\/v2\/ppma_author?post=4856"}],"curies":[{"name":"bem-jogado","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}