对话式 AI AgentConversational AI Agent
通过平板端语音或文字交互完成校准、采图、训练、调参、查询和诊断，降低产线人员学习成本。Use voice or text on the tablet to calibrate, capture images, train models, tune parameters, query data, and diagnose issues.
AI Agent + 嵌入式 AI 工业现场智能服务 AI Agent + Embedded AI Industrial Services
跟它说，它就会。 Tell It. It Does It.
StarAgent 把视觉检测、预测性维护、嵌入式 AI 和现场执行连接起来，让操作工用自然语言完成巡检、诊断与异常处置。 StarAgent connects visual inspection, predictive maintenance, embedded AI, and field execution so operators can inspect, diagnose, and handle exceptions through natural language.
StarAgent
AI Agent
视觉检测Visual Inspection
预测性维护Predictive Maintenance
嵌入式 AIEmbedded AI
现场执行Field Execution
Capabilities
会看、会听、会预警的工业现场 AgentAn Industrial Agent That Sees, Listens, and Warns
产品矩阵围绕工业现场的两类高频需求设计：视觉检测负责看见缺陷、状态和安全风险；预测性维护负责理解设备振动、温度、电流和运行趋势，提前发现停机风险。 The matrix is built around two common industrial needs: visual inspection sees defects, states, and safety risks; predictive maintenance understands vibration, temperature, current, and operating trends before downtime happens.
视觉检测Visual Inspection
嵌入式 AI 网关运行轻量视觉模型，实时识别外观缺陷、位置偏差、状态异常、异物和安全风险。The embedded AI gateway runs lightweight vision models to detect defects, position deviations, state anomalies, foreign objects, and safety risks.
预测性维护Predictive Maintenance
接入振动、温度、电流、压力等设备数据，识别异常趋势、健康评分和剩余寿命风险，辅助安排维护窗口。Connect vibration, temperature, current, pressure, and other equipment data to identify anomaly trends, health scores, and remaining-life risks.
Products
产品矩阵：视觉检测 + 预测性维护Product Matrix: Visual Inspection + Predictive Maintenance
以 StarAgent 为统一交互入口，向下连接工业相机、传感器和设备控制，向上连接训练平台、知识库和运维流程。首批产品方向覆盖视觉检测与设备预测性维护。 StarAgent is the shared interaction layer, connecting cameras, sensors, and controls below, while linking training, knowledge, and maintenance workflows above. The first product directions cover visual inspection and predictive maintenance.
查看产品方案View product conceptVisual Inspection
极智·视觉魔盒SmartVision Magic Box Pro
面向工业产线、设备区域和关键工位的视觉检测设备，支持少样本训练、实时检测、异常记录、报警联动和现场执行。A visual inspection device for production lines, equipment zones, and critical workstations, supporting few-shot training, real-time detection, anomaly records, alert linkage, and field execution.
视觉检测Visual Inspection
覆盖缺陷、异物、人员行为、设备状态和安全风险识别。Covers defect, foreign-object, behavior, equipment-state, and safety-risk detection.
极智·维保魔盒SmartMaintenance Magic Box Pro
接入设备传感数据，输出健康评分、异常趋势、维护建议和停机风险预警。Connects equipment sensor data to produce health scores, anomaly trends, maintenance suggestions, and downtime-risk alerts.
嵌入式 AI 边缘推理Embedded AI Edge Inference
现场侧完成视觉推理、时序分析、缓存和规则执行，弱网环境可独立运行。Runs vision inference, time-series analysis, caching, and rule execution on site, even with weak connectivity.
对话式运维 AgentConversational Ops Agent
通过自然语言查询异常、生成巡检记录、解释根因并配置报警策略。Uses natural language to query anomalies, generate inspection logs, explain causes, and configure alert policies.
Product Loop
从现场感知到智能执行，全链路 AI 原生From Field Sensing to Intelligent Execution
Agent 运行在平板端负责理解意图、规划步骤、调用工具和维护记忆；设备侧负责图像采集、传感数据接入、边缘推理、报警联动和记录留存。 The tablet agent understands intent, plans steps, calls tools, and manages memory. The device side handles image capture, sensor data, edge inference, alert linkage, and records.
对话规划Conversational Planning
用户说出目标，Agent 拆解成采图、训练、调参、测试等步骤。The user states a goal; the agent breaks it into capture, train, tune, and test steps.
边缘分析Edge Analysis
工业相机和设备传感器采集现场数据，嵌入式 AI 网关完成视觉识别和健康趋势分析。Cameras and equipment sensors collect field data, while the embedded AI gateway runs visual recognition and health-trend analysis.
执行闭环Execution Loop
异常结果触发报警、停线建议、工单或现场控制，Agent 可继续解释原因和处置步骤。Anomalies trigger alerts, stop-line suggestions, work orders, or field control, while the agent explains causes and actions.
Solutions
工业现场优先场景Priority Industrial-Site Scenarios
把现场异常处理拆成五个动作：先感知风险，再在边缘侧判断优先级，交给人机协同确认，随后联动设备处置，最后沉淀为样本和规则。 Field exception handling is organized into five actions: sense risks, judge priority at the edge, confirm with human-agent collaboration, coordinate field action, and feed samples and rules back into the system.
现场感知Field Sensing
视觉、传感器和设备状态捕捉异常信号。Vision, sensors, and equipment states capture anomaly signals.
边缘判断Edge Judgment
本地分析缺陷、健康趋势、风险等级和处置窗口。Local analysis scores defects, health trends, risk levels, and action windows.
人机协同Human-Agent Collaboration
操作工通过对话复核、追问、补充现场信息并确认方案。Operators verify, ask follow-ups, add context, and confirm actions through dialogue.
现场处置Field Action
联动报警、工单、停线建议、剔除机构或控制设备。Coordinate alerts, work orders, stop-line advice, rejectors, or equipment controls.
复盘学习Review & Learning
把处置结果回流为样本、规则和知识，优化下次判断。Feed outcomes back as samples, rules, and knowledge to improve the next judgment.
Scenarios
典型应用场景Application Scenarios
视觉检测与安全识别Visual Inspection and Safety Detection
实时识别缺陷、异物、人员闯入、设备状态异常和关键区域安全风险。Detect defects, foreign objects, personnel intrusion, equipment-state anomalies, and safety risks in real time.
设备健康趋势预警Equipment Health Trend Alerts
融合振动、温度、电流等数据，提前发现轴承、泵、风机、传送机构等设备异常趋势。Fuse vibration, temperature, and current data to detect early anomaly trends in bearings, pumps, fans, and conveyors.
Agent 巡检与工单联动Agent Inspection and Work Orders
对话查询异常、生成巡检记录、解释根因，并联动报警、工单和现场控制策略。Query anomalies, generate inspection logs, explain causes, and link alerts, work orders, and field-control policies.
Why StarC-Hub
为什么它是 AI Agent 原生工业现场设备Why It Is AI-Agent-Native Industrial Equipment
自然语言操作Natural-Language Operation
一线人员通过对话完成巡检、训练、查询、诊断和报警配置。Operators inspect, train, query, diagnose, and configure alerts through conversation.
视觉检测Visual Inspection
识别缺陷、状态、安全风险和现场异常，支持边缘实时处理。Detect defects, states, safety risks, and field anomalies with edge real-time processing.
预测性维护Predictive Maintenance
基于设备时序数据判断健康趋势，提前提示维护窗口和停机风险。Use equipment time-series data to evaluate health trends and warn about maintenance windows and downtime risks.
本地边缘执行Local Edge Execution
弱网不影响本地推理、报警、缓存和现场联动，云端用于训练和知识增强。Local inference, alerts, caching, and field linkage continue under weak connectivity; cloud supports training and knowledge.
从一个现场问题开始，扩展成智能服务矩阵Start from One Site Problem, Expand into a Smart-Service Matrix
可以先从视觉检测或设备预测性维护切入，再逐步扩展到巡检 Agent、报警工单、知识库和更多现场执行能力。Start with visual inspection or predictive maintenance, then expand into inspection agents, alert work orders, knowledge bases, and more field-execution capabilities.