差异 01Difference 01
传统 AI WorkflowTraditional AI Workflow
Workflow 需要用户预先设计步骤。A workflow requires users to predefine steps.
AIOS Runtime
Runtime 先理解目标，再动态拆解任务。Runtime understands the goal first, then decomposes tasks dynamically.
Runtime BehaviorRuntime Behavior
Runtime 围绕目标组织能力，而不是让用户维护固定流程。Runtime coordinates capabilities around a goal instead of asking users to maintain fixed workflows.
AIOS Runtime 负责理解意图、拆解任务、选择能力、编排 Agent、执行工作流并返回结果。它的价值不是生成一段回复，而是把多个能力组织成一次可完成的行为链。AIOS Runtime understands intent, decomposes tasks, selects capabilities, orchestrates agents, executes workflows, and returns results. Its value is not a response, but a completed behavior chain across multiple capabilities.
Runtime 行为链Runtime Chain
01
用户输入User Input
用户表达目标或进入某个空间。The user expresses a goal or enters a space.
02
分析意图Analyze Intent
Runtime 判断目标、上下文、约束和完成标准。Runtime identifies the goal, context, constraints, and completion criteria.
03
拆解任务Decompose Task
复杂目标被拆成可执行子任务。The complex goal is split into executable subtasks.
04
选择能力Select Capabilities
Runtime 找到研究、搜索、设备、语音、演示等可用能力。Runtime finds available capabilities such as research, search, devices, voice, or presentation.
05
编排 AgentOrchestrate Agents
多个 Agent 围绕同一目标分工协作。Multiple agents divide work and collaborate around the same goal.
06
自动执行Execute Automatically
任务在 Runtime 中连续推进。Tasks continue inside Runtime.
07
聚合结果Aggregate Results
Runtime 整合多方输出并生成最终结果。Runtime combines outputs and produces the final result.
Runtime 与传统 AI Workflow 的差异How Runtime Differs from Traditional AI Workflow
差异 02Difference 02
传统 AI WorkflowTraditional AI Workflow
Workflow 通常绑定固定工具。A workflow is usually bound to fixed tools.
AIOS Runtime
Runtime 根据上下文选择可用能力。Runtime selects available capabilities from context.
差异 03Difference 03
传统 AI WorkflowTraditional AI Workflow
Workflow 执行线性自动化。A workflow executes linear automation.
AIOS Runtime
Runtime 让多个 Agent 协作并聚合结果。Runtime lets multiple agents collaborate and aggregates results.
差异 04Difference 04
传统 AI WorkflowTraditional AI Workflow
Workflow 输出中间材料。A workflow often returns intermediate material.
AIOS Runtime
Runtime 返回用户真正需要的完成结果。Runtime returns the completed outcome the user needs.
三个行为链例子Three Behavior Chain Examples
行为证据Behavior Evidence场景 01Example 01
研究任务Research Task
用户：帮我准备一份 AI 玩具市场分析。User: Help me prepare a market analysis for AI toys.
用户收到一份完成的研究报告。The user receives a completed research report.
步骤 01Step 01
分析请求Analyze request
确定行业、受众、输出格式。Identify industry, audience, and output format.
步骤 02Step 02
搜索来源Search sources
收集市场、产品、竞争和趋势信息。Collect market, product, competition, and trend information.
步骤 03Step 03
调用研究 AgentInvoke research agents
分工处理数据、洞察和结构。Divide work across data, insight, and structure.
步骤 04Step 04
生成分析Generate analysis
形成结构化报告。Produce a structured analysis.
步骤 05Step 05
导出演示Export presentation
把结果交付为可使用材料。Deliver the result as usable material.
场景 02Example 02
智能教室Smart Classroom
用户进入教室。The user enters a classroom.
环境开始主动服务。The environment becomes proactive.
步骤 01Step 01
识别身份Detect identity
确认教师、课程和权限。Confirm teacher, course, and permissions.
步骤 02Step 02
激活教室模式Activate classroom mode
切换空间状态。Switch the space state.
步骤 03Step 03
连接教学设备Connect teaching devices
联动屏幕、终端和实训设备。Coordinate screens, terminals, and training devices.
步骤 04Step 04
加载学习 AgentLoad learning agents
准备课程、练习和反馈能力。Prepare lessons, practice, and feedback capabilities.
步骤 05Step 05
同步学生界面Synchronize student interfaces
让学生端进入同一教学上下文。Bring student interfaces into the same teaching context.
场景 03Example 03
能力协作Capability Collaboration
开发者发布翻译、语音合成和演示能力。A developer publishes translation, voice synthesis, and presentation capabilities.
多个能力像一个系统一样工作。Multiple capabilities work as one system.
步骤 01Step 01
选择能力Select capabilities
Runtime 找到当前目标需要的能力组合。Runtime finds the capability set required by the goal.
步骤 02Step 02
构建执行链Build execution chain
确定调用顺序和依赖关系。Define invocation order and dependencies.
步骤 03Step 03
协调 AgentCoordinate agents
让不同 Agent 处理不同子任务。Let different agents handle different subtasks.
步骤 04Step 04
执行工作流Execute workflow
连续调用能力并处理结果。Invoke capabilities continuously and process results.
步骤 05Step 05
返回最终输出Return final output
把多个能力的结果整合为一个交付物。Merge results from multiple capabilities into one deliverable.
Intent OrchestrationIntent Orchestration
Capability CollaborationCapability Collaboration
Runtime ExecutionRuntime Execution
Proactive EnvironmentsProactive Environments
Intelligent WorkflowsIntelligent Workflows