Discussion on Home Knowledge Base and Intelligent Agent

In 2025, AI has made significant strides, especially in user experience. It has moved beyond research and is now accessible to end users. The commercial application of AI in home scenarios is gradually taking shape, bringing convenience and transforming everyday life.

To meet the diverse needs of home scenarios, we have studied the planning of home AI systems. Household data, including documents, images, and videos, along with critical data types like household and user profiles, is a key assent. Building a home knowledge base around this data has become central in the development of home AI. An intelligent agent-based technical solution supports home AI services in both proactive and reactive modes. The home knowledge base and the home intelligent agent work in tandem to enable applications across various areas such as health, learning, entertainment, work, fitness, and smart living (Fig. 1).

Home Knowledge Base: The Core of Home AI Services

The home knowledge base is a knowledge system that integrates various domains, such as childcare, learning, health, work, entertainment, daily life, and fitness to provide scientific guidance and personalized services for family members. The data includes:

• Home document data: Includes documents uploaded by users, such as PPTs, Word files, PDFs, images, and videos, as well as user memos.
• User & home profile data: Gathers user behavior, preferences, health data, and home environmental conditions through sensors and interaction records to construct user and home profiles.
• Home knowledge graph: Uses knowledge extraction and structured representation to build a graph that maps relationships between family members, devices, and the environment.
• Home device and environmental data: Includes sensor data (e.g., temperature, humidity), device logs, environmental change records, and basic information, usage history, and maintenance records of smart devices.
• Other home knowledge bases: Contains general knowledge bases, such as encyclopedic knowledge graphs uploaded by users, and curated data.

Home Intelligent Agent: The Pathway to Home AI Services

The home intelligent agent leverages data from the home knowledge base and employs machine learning, natural language processing, and multimodal large models to understand family members’ needs and deliver personalized services. Its key functions include:

• User profile construction: Builds detailed profiles for each family member, covering habits, preferences, and behavioral patterns, by combining user behavior data from the home knowledge base with family member information and preference settings.
• Scenario awareness and dynamic adjustment: Perceives changes in the home environment and provides scenario-appropriate services by integrating environmental data and device status from the home knowledge base to identify the scenario.
• Multimodal interaction and natural language understanding: Supports user interaction through voice, text, gestures, and other modalities, understanding complex instructions and using contextual and historical data to provide more accurate services.
• Active learning and feedback: Continuously learns and adapts to improve service quality based on family members' needs, collecting feedback to dynamically adjust behavioral strategies.
• Personalized recommendations and proactive services: Predicts family members’ needs based on user profiles and scenario awareness, offering personalized services proactively.
• Privacy protection and security: Ensures compliance with privacy and security requirements for the home knowledge base and agent. Data is stored locally to prevent leakage, encryption protects sensitive information, and a user permission management system is in place.

Scenario-Based Implementation

Home AI services are tailored to specific scenarios to meet diverse needs:

• Health and wellness: Focuses on health monitoring and safety protection, with future developments emphasizing proactive intervention, such as automatically triggering emergency responses upon detecting abnormal behavior in elderly family members.
• Companion learning: Currently includes educational robots and smart learning platforms, with future advancements integrating multimodal interaction technologies for emotional interaction and adaptive teaching.
• Entertainment: Currently involves voice control and multi-device coordination (e.g., smart speakers controlling lights, music, and TV), with future trends in holography, socialization, and multi-user virtual engagement.
• Work: Current smart home features optimize the environment and efficiency, with future developments extending to intelligent collaboration.
• Fitness: Currently relies on the coordination of smart wearables and home fitness equipment, with future innovations integrating biometric recognition and virtual coaching, such as personalized training plans.
• Smart home: Currently achieves device interconnectivity and automated control, with future directions focusing on whole-house intelligence and energy self-sufficiency, such as integrating household energy for optimized allocation.

By leveraging the home knowledge base and intelligent agent, AI is evolving from single-function applications to full-scenario services, shifting household services from passive response to active care and driving improved quality of life.