Building Live Voice Agents: Deploying Real-Time Multimodal AI Systems to Production
暫譯: 打造即時語音代理:將即時多模態 AI 系統部署至正式環境
Hotz, Heiko, Kartakis, Sokratis
- 出版商: O'Reilly
- 出版日期: 2026-12-15
- 售價: $2,820
- 貴賓價: 9.5 折 $2,679
- 語言: 英文
- 頁數: 561
- 裝訂: Quality Paper - also called trade paper
- ISBN: 9798341661127
- ISBN-13: 9798341661127
-
相關分類:
Large language model、語音辨識 Speech-recognition
海外代購書籍(需單獨結帳)
相關主題
商品描述
As users increasingly expect AI to interact as naturally as humans, engineers must move beyond static prompts to building systems that perceive, reason, and act instantly, mimicking human interaction more closely. Multimodal, Real-Time AI Agent Systems takes you from agent fundamentals to architecting advanced, bidirectional, multimodal, multi-agent systems, focusing on the difficult leap from proof of concept to production. You'll start by building agents directly with foundation models to understand core components and then master modern agent frameworks that simplify and scale implementation.
Written by industry practitioners, this book connects agentic concepts with cutting-edge standards and protocols of interoperability. You'll learn to build enterprise-grade agent platforms that enforce scalable AgentOps, rigorous evaluation, and extreme security measures required for live streaming interactions.
Through practical multimodal agent examples, you'll understand how to:
- Design scalable, low-latency architectures for single and multi-agent systems
- Engineer the complete lifecycle of live multimodal streaming, using backend-centric architectures to minimize frontend complexity
- Implement unified protocols like the Model Context Protocol and Agent-to-Agent for tool use and agent discovery
- Apply operational best practices for security, testing, and scaling to support thousands of concurrent live agents
隨著使用者日益期待 AI 能像人類一樣自然地互動,工程師必須超越靜態提示,轉而建構能即時感知、推理與行動的系統,更貼近地模擬人類互動方式。《Multimodal, Real-Time AI Agent Systems》將帶領您從 AI agent 基礎一路學習,進階至雙向、多模態、多 agent 系統的架構設計,並聚焦於從概念驗證(proof of concept)邁向正式上線環境時所面臨的艱難挑戰。您將先直接使用 foundation models 建構 agent,以理解其核心元件;接著掌握能簡化實作並支援擴展的現代 agent frameworks。
本書由業界實務工作者撰寫,將 agentic 概念與尖端的互通性標準及協定串聯起來。您將學習如何建構企業級 agent 平台,落實可擴展的 AgentOps、嚴謹的評估機制,以及支援即時串流互動所需的高強度安全措施。
透過實用的多模態 agent 範例,您將了解如何:
• 設計適用於單一 agent 與多 agent 系統的可擴展、低延遲架構
• 運用以後端為核心的架構,簡化前端複雜度,打造完整的即時多模態串流生命週期
• 實作 Model Context Protocol 與 Agent-to-Agent 等統一協定,以支援工具使用與 agent 探索
• 應用安全性、測試與擴展的營運最佳實務,支援數千個同時運作的即時 agent
以上中文內容由 AI 翻譯