TERMINUS HALI Agent Upgrade: Eight New Features Introduced to Create a More Attentive Personal Assistant

Table of Content [Hide]

    Terminus releases the newly upgraded HALI intelligent body system. In addition to basic intelligent conversation, deep thinking, and call functions, eight new features including meeting assistant, translation and interpretation, writing assistant, schedule management, navigation services, weather inquiry, surrounding search, and news summary have been added, covering the high-frequency needs of users' work and life. The new version of HALI has completed adaptation with BUTTONS Clip headset. When the headset is connected, users only need to call "HALI HALI" to wake up their considerate personal assistant.


    Eight new features focus on work and life scenarios


    Currently, the wave of large models and generative intelligence is driving intelligent wearable devices to transform from "human external attachments" to "digital selves," enabling them to cover richer life scenarios and achieve deeper life penetration. As an intelligent body system created specifically for intelligent wearable devices, this HALI upgrade is also designed for the core needs in typical user scenarios.


    For example, in the work scenario, the meeting assistant function can realize real-time meeting voice transcription and automatically generate summary minutes after the meeting, sending them to all participants. The translation and simultaneous interpretation function can support dozens of languages, including Chinese, English, Russian, Japanese, French, and Spanish, with a real-time translation accuracy of up to 96%. The writing assistant function can provide users with various document templates and creative inspirations while offering text optimization support. The schedule management function connects multiple smart applications, helping users create schedules with "one sentence" and providing multi-terminal reminders before the schedule starts.


    In daily living scenarios, HALI also provides users with rich intelligent services. When the headset is connected, users only need to voice the destination they want to reach, and HALI can provide detailed navigation services. The surrounding search function can help users find nearby restaurants, parking lots, gyms, and even offline activities, so users don’t miss out on the beauty around them in their busy lives. The news summary feature helps users organize news they are interested in and compiles comprehensive, trustworthy, and reliable summaries through DeepSearch. The weather inquiry function can autonomously inform users about the weather conditions and future trends, offering clothing adjustment suggestions.


    Thanks to its use of aerospace titanium and skin-friendly silicone design, the BUTTONS Clip headset achieves a single-side wearing experience as light as 6.7g, making HALI an invisible yet ubiquitous personal assistant. Meetings, meal appointments, and exhibitions, complex processes that previously required switching between multiple apps, can now be easily completed with "one sentence" based on the HALI intelligent body system.


    Multi-intelligent coordination to create "usable" AI


    According to Terminus technology experts, the eight aforementioned features rely on the scenario-specific vertical intelligent body systems created by the Terminus team. As the central hub that coordinates all intelligent body collaborations, HALI can, based on its intent recognition and command parsing capabilities, break down complex problems into multiple sub-problems upon receiving user commands, and call the corresponding subsystems and third-party tools to complete task execution.


    Unlike traditional intelligent bodies that adopt workflow methods, the HALI intelligent system introduces end-to-end reinforcement learning technology. Through reward functions rather than relying on humans, it guides the model to continuously optimize strategies in interactions with subsystems and external tools. Based on this method, HALI achieves truly autonomous sub-task decomposition, tool invocation, and continuous self-validation without external manual arrangement.


    terminus-group-20250825-1.jpg


    Moreover, the vertical intelligent systems for specific functions adopt efficient data compression technology, allowing the system to extract key effective information from user conversation data, filter out unimportant parts, and increase the information density of data storage. By continuously learning users' preferences and thinking styles, the intelligent body system constructs a knowledge graph tailored to each user, making every HALI possess a different knowledge system, creating "thousand faces" intelligent assistants.


    Furthermore, the newly added news summary feature introduces DeepSearch technology, which gradually approaches complex question answers through multiple rounds of "search-reading-reasoning" cycles, breaking through the limitations of traditional search engines and single retrieval augmented generation (RAG), further enhancing the deep search and reasoning capabilities of intelligent bodies.


    According to Terminus technology experts, creating a "useful" product for users is more important than creating a "toy" that looks fancy but has few application scenarios: "The new version of HALI can help users complete most work and daily tasks with just voice commands, without external intervention, becoming a work and life assistant that ordinary people can also have. This not only meets the "trendy" needs of young people but also reduces the threshold for vulnerable groups such as the elderly and visually impaired to obtain intelligent services, providing them with real support." He said.


    In the future, the HALI intelligent body will also complete adaptation for smart glasses, speakers, service robots, and other smart devices, building a complete intelligent life service ecosystem.

    References
    We use cookies to offer you a better browsing experience, analyze site traffic and personalize content. By using this site, you agree to our use of cookies. Visit our cookie policy to leamn more.