11-19 更新 实时
首页/ AI 大模型/ 正文

PuppyGraph 融资1000万美元助AI智能体查询

AI 大模型 来源:Analytics India Magazine 2026-11-11 13:23 0 阅读 机器翻译
PuppyGraph 融资1000万美元助AI智能体查询

面向企业人工智能(AI)的图查询引擎初创公司 PuppyGraph 已在由 Particle Future 领投、defy.vc 参投的 A 轮融资中筹集 1000 万美元。

“企业数据的洞见很快将由智能体生成,”PuppyGraph 联合创始人兼 CEO Weimo Liu 在一份声明中表示。他补充说,智能体需要能够穿梭于数千张表之间、遵循关系路径,并跨部门和系统检索证据。

该公司表示,这轮融资使其累计融资额达到 1500 万美元,资金将用于工程研发以及企业市场的拓展。

该公司表示,此次融资消息紧随 PuppyGraph 1.0 的正式发布。该版本允许 AI 智能体以统一的图形式,跨最多 3,000 张企业表查询数据,而无需将底层数据迁移到单独的图数据库中。

据 PuppyGraph 称,最初只覆盖 30–50 张表的 AI 智能体项目,当生产负载需要跨越数千张表和多个系统建立连接时,可能会难以扩展。该平台旨在帮助企业在数据库、数据仓库和数据湖之间执行查询。

PuppyGraph 1.0 包含一个 AI 辅助的 schema 构建器,可从现有企业表生成图 schema,供数据团队审阅和批准。随后,智能体可以通过该平台内置的 AI 智能体,或通过 Model Context Protocol (MCP) 连接的可兼容工具,用自然语言探索相互关联的数据。

该引擎还会拒绝未定义的关系引用,并返回可供智能体用于修正查询的反馈。该公司表示,其目标是防止智能体对企业数据之间的连接方式做出无依据的假设。

PuppyGraph 表示,其分布式查询引擎支持 PB 级工作负载,并可在受支持的工作负载上实现亚秒级图查询。在另一项基准测试中,它使用一个四机集群,在 2.26 秒内完成了跨越数十亿条边的 10 跳查询。

该平台支持的数据源包括 Apache Iceberg、Databricks、Snowflake、ClickHouse、PostgreSQL 和 Microsoft Fabric。它还会在图遍历过程中强制执行所配置的行级权限,并提供基于角色的访问控制与单点登录。

该公司的客户包括 AMD、Palo Alto Networks、Datadog、Netskope、Coinbase、Trend Micro、Baker Hughes 和 Sola Security。过去一年其年度经常性收入增长了七倍,不过公司并未披露具体收入数字。

“对于面向用户的应用而言,查询延迟和数据新鲜度是绝不能妥协的。PuppyGraph 让我们能够直接在现有的数据平台上执行多跳图遍历,无需迁移或复制任何一行数据。”Palo Alto Networks 工程总监 Ran Xia 表示。

该公司表示,开发者可以通过免费的 Docker 部署方式使用 PuppyGraph,无需提交支付方式或填写注册表单。

译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。

查看英文原文(PuppyGraph Raises $10 Mn to Help AI Agents Query Enterprise Tables)

PuppyGraph, a graph query engine startup for enterprise artificial intelligence (AI), has raised $10 million in a Series A round led by Particle Future, with participation from defy.vc.

“The insights of enterprise data will soon be generated by agents,” Weimo Liu, Co-founder and CEO of PuppyGraph, said in a statement. He added that agents would need to navigate thousands of tables, follow relationships, and retrieve evidence across departments and systems.

The funding takes the US-based company’s total capital raised to $15 million and will support engineering and enterprise go-to-market expansion, the company said.

The announcement follows the general availability of PuppyGraph 1.0, which allows AI agents to query data across up to 3,000 enterprise tables as a unified graph without transferring the underlying data into a separate graph database, the company said.

According to PuppyGraph, AI agent projects that initially cover 30–50 tables can become difficult to scale when production workloads require connections across thousands of tables and multiple systems. The platform is designed to help enterprises run queries across databases, data warehouses, and data lakes. 

PuppyGraph 1.0 includes an AI-assisted schema builder that generates graph schemas from existing enterprise tables for data teams to review and approve. Agents can then explore connected data using natural language through the platform's built-in AI agent or compatible tools connected through the Model Context Protocol (MCP).

The engine also rejects undefined relationship references and returns feedback that agents can use to correct their queries. It aims to prevent agents from making unsupported assumptions about how enterprise data is connected, the company said.

PuppyGraph said its distributed query engine supports petabyte-scale workloads and subsecond graph queries for supported workloads. In a separate benchmark, it completed a 10-hop query across billions of edges in 2.26 seconds using a four-machine cluster.

The platform supports data sources including Apache Iceberg, Databricks, Snowflake, ClickHouse, PostgreSQL, and Microsoft Fabric. It also enforces configured row-level permissions during graph traversals and offers role-based access controls and single sign-on.

The company counts AMD, Palo Alto Networks, Datadog, Netskope, Coinbase, Trend Micro, Baker Hughes, and Sola Security among its customers. Its annual recurring revenue grew sevenfold over the past year, although it did not disclose the revenue figure.

“For a user-facing application, query latency and data freshness are non-negotiable. PuppyGraph allows us to execute multi-hop graph traversals directly on our existing data platform without moving or duplicating a single row,” said Ran Xia, Engineering Director at Palo Alto Networks.

The company said that developers can access PuppyGraph through a free Docker deployment, without submitting a payment method or registration form. 


正文保留英文原文,标题与摘要由机器翻译。

本文由采集器自动同步

原文发布于 Analytics India Magazine(PuppyGraph Raises $10 Mn to Help AI Agents Query Enterprise Tables),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。

阅读原文 ↗