Human-like thinking, interaction, and response
From 2 weeks to just 2 days
Human-like thinking, interaction, and response
From 2 weeks to just 2 days
Designed for Diverse Scenarios
Designed for Diverse Scenarios
Guiding · Reception · Companionship · Elderly Care
· Guiding
· reception
· Companionship
· Elderly care
Beijing Art Center—one of the three major cultural landmarks in Beijing's sub-center, with 280+ performances annually and 190,000 audience members. The 125,000 square meter super-large space has complex circulation, and audiences easily get lost, urgently needing more considerate on-site guidance. Staff need to handle thousands of repetitive inquiries daily, working under high-intensity conditions for long periods, urgently needing more intelligent consultation assistance. Before each performance, multiple staff need to be arranged for circulation guidance and curtain calls, causing waste of manpower, urgently needing more efficient curtain call methods.OrionStar intelligent reception robots can relieve consultation pressure at service desks, meet audience demands promptly, make curtain calls warm, and after going on duty, audience satisfaction increased to 98%
The reception robot now handles guest reminders and send-offs. Our team can finally focus on deeper, more meaningful services like art tours.
Currently, the OrionStar GreetingBot Nova and OrionStar GreetingBot Mini in the voice interaction robot series can upgrade. If ecosystem partners have models they want to upgrade, they can tell OrionStar in the comments section~
OrionStar AgentOS integrates multi-language voice recognition technology, supporting over 10 mainstream languages including Chinese, English, Japanese, Korean, Thai, Spanish, French, Italian, German, etc. For dialects, Cantonese recognition and response are currently supported, but Hakka/Sichuan dialect are not yet supported, and will be gradually improved in the future.
OrionStar AgentOS makes decisions through multimodal visual recognition, sound source distance and angle, and voiceprint recognition algorithms, combined with front-end noise reduction algorithms optimized based on OrionStar robot microphone arrays.
OrionStar AgentOS has built-in underlying large model capabilities. Since it is deeply integrated with the Agent Brain (main Agent) thinking process, it cannot be simply replaced. However, in Agent applications, developers can flexibly integrate third-party capabilities to replace system capabilities.
Due to the wake-up-free mechanism restrictions (we filter non-human interactions), if there are special requirements, you can manually turn off the wake-up-free capability through the Agent secondary development API.
1) Applied AI Agent development paradigm architecture, simplified many development processes for voice interaction scenarios, no longer need complex configuration of voice backends, some basic capability modules are also built into AgentOS, developers can focus more on their own business process logic.
2) Developed AgentOS SDK and deep adaptation with AI programming tools, AI-powered empowerment.
Due to the underlying dependencies of the robot system, basic Android knowledge is needed, but with the assistance of AI programming tools and our AgentOS SDK optimization for AI programming tools, this process will become increasingly simple. For example: In our internal hackathon activities, colleagues who completely don't understand code can complete the development of a new Agent voice interaction application based on AgentOS SDK in a few days with AI assistance.
It can be defined as a global intent, or as an intent within a specific business scenario.
This function can be completed through natural language constraints, such as emphasizing and specifying when defining Action functions and parameter descriptions. If specified very clearly, if someone says ID card, OrionStar AgentOS will judge semantically that it's not a phone number and ask the user to provide a phone number instead of an ID card number.
The following methods can intervene in the Agent's thinking process:
1) You can define and orchestrate through natural language in application development, continuously optimize your Prompt, guide AgentOS on what logic and steps to think with.
2) You can also provide some examples in the application for Agent to reference.
3) Also supports global-level Action strong intervention strategies, such as for a certain sentence, you want it to do a deterministic thing.
This requirement can be implemented in secondary development Agent applications.
Agent applications implement themselves, belongs to basic Android capabilities.
1) OrionStar AgentOS can be deeply integrated with robot business application interaction interface logic. Coze/Dify cannot achieve this.
2) AgentOS is the running environment for OrionStar service robot Agent applications, supporting Agent operation and scheduling, while Coze and Dify are workflow building platforms.
Application Video 
