AI News AI资讯 4d ago Updated 3d ago 更新于 3天前 48

Nous Research Ships Bot Mode for Hermes Agent, Turning Agent Profiles Into a Roster of Named Bots Nous Research 为 Hermes Agent 推出 Bot Mode,将智能体配置文件转变为命名机器人名单

Nous Research shipped Bot Mode for Hermes Agent v0.20.3, transforming individual agent profiles into named bots with isolated chat, memory, skills, and pinned models Each bot is fundamentally a Hermes profile stored at ~/.hermes/profiles/<name>/, with no new storage layer introduced Bot-to-bot communication works through real CLI handoffs into a persistent Agent Inbox, using @mention syntax for work handoffs Group chats support 2-6 bots with up to three serial rounds of member turns, with a mult Nous Research推出Hermes Agent的Bot Mode,将单一agent会话升级为多bot协作系统,每个bot拥有独立profile、memory、skills和pinned model Bot间通过持久Agent Inbox和CLI handoffs实现@mention消息传递与工作交接,群组聊天支持2-6个bot、最多3轮串行协调 从一天公开beta快速迭代为Hermes Agent v0.20.3的捆绑默认功能,采用MIT许可,无新增存储层 适用场景覆盖软件工程、AI研究、技术内容运营、量化研究和IT运维,明确定位为桌面工作站工具而非企业级基础设施

72
Hot 热度
68
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • Nous Research shipped Bot Mode for Hermes Agent v0.20.3, transforming individual agent profiles into named bots with isolated chat, memory, skills, and pinned models
  • Each bot is fundamentally a Hermes profile stored at ~/.hermes/profiles//, with no new storage layer introduced
  • Bot-to-bot communication works through real CLI handoffs into a persistent Agent Inbox, using @mention syntax for work handoffs
  • Group chats support 2-6 bots with up to three serial rounds of member turns, with a multi-source roster pulling from all connected profiles
  • The feature shipped from a one-day public beta plugin to a bundled, default-on Desktop plugin, with the standalone repository now archived

Why It Matters

Bot Mode represents a practical step toward multi-agent orchestration without requiring complex infrastructure, making it immediately accessible to solo builders and small teams. The MIT licensing and desktop-first approach lower the barrier to experimenting with agent collaboration patterns that were previously only available in enterprise-grade platforms.

Technical Details

  • Architecture: Bots are Hermes profiles at ~/.hermes/profiles// with isolated config, memory, skills, credentials, and chat history—no new storage layer
  • Communication: Bot-to-bot messaging uses real CLI handoffs (hermes -p <bot> chat -c "Agent Inbox" -q "...") with attribution and SOUL.md-based reply protocols
  • Group coordination: Group chats support 2-6 bots with up to three serial rounds; mentioned bots respond immediately, others respond when unmentioned, with brief replies or pass options
  • Configuration: Per-bot settings include model pinning, custom SOUL.md, per-skill/toolset enablement, and profile cloning via profiles.*gateway RPCs (list, create, describe, configure)
  • Avatars: Geometric faces in 7 shapes/10 colors, image upload, portrait generation, or pixel pets with animated eyes during bot activity

Industry Insight

  • Solo developers and small engineering teams can deploy multi-agent workflows immediately without enterprise infrastructure, but regulated organizations should treat this as a workstation tool rather than managed infrastructure due to lack of SSO, audit logs, and policy controls
  • The rapid beta-to-default rollout (one-day public beta to bundled feature) signals Nous Research's commitment to community-driven development and suggests similar open-source agent frameworks will see accelerated iteration cycles
  • The @mention handoff pattern and persistent Agent Inbox provide a lightweight alternative to complex agent orchestration frameworks, potentially influencing how smaller teams approach multi-agent system design

TL;DR

  • Nous Research推出Hermes Agent的Bot Mode,将单一agent会话升级为多bot协作系统,每个bot拥有独立profile、memory、skills和pinned model
  • Bot间通过持久Agent Inbox和CLI handoffs实现@mention消息传递与工作交接,群组聊天支持2-6个bot、最多3轮串行协调
  • 从一天公开beta快速迭代为Hermes Agent v0.20.3的捆绑默认功能,采用MIT许可,无新增存储层
  • 适用场景覆盖软件工程、AI研究、技术内容运营、量化研究和IT运维,明确定位为桌面工作站工具而非企业级基础设施

为什么值得看

本文展示了开源agent架构从单点工具向多智能体协作系统的演进路径,为开发者提供了可复用的bot间通信与协调模式。其"bot即profile"的设计哲学降低了多agent系统的实现复杂度,同时MIT许可确保了企业级应用的灵活性。

技术解析

  • 架构设计:Bot Mode采用"bot即profile"的核心设计理念,每个bot存储在~/.hermes/profiles//目录下,共享Hermes现有存储层而非引入新系统,通过profiles.*gateway RPCs(list/create/describe/configure)管理配置,保持系统表面积最小化
  • 通信机制:Bot间通过持久Agent Inbox进行消息传递,实际实现为CLI命令hermes -p chat -c "Agent Inbox" -q "...",配合SOUL.md定义回复协议,@mention触发工作交接与报告返回
  • 群组协调:支持2-6个bot的群组聊天,最多3轮串行成员回合,提及的bot响应,无人被提及时所有bot响应,每bot简短回复或传递;多源roster可从所有Connections拉取bot
  • 功能演进:从beta到正式版增加了群组组织、跨机器同步、多源roster等能力,开发从独立插件仓库转入Hermes Desktop代码库apps/desktop/src/plugins/hermes-bots/
  • 配置管理:每个bot可独立配置provider、model、SOUL.md、skills和tools,支持profile克隆,avatar支持几何图形、上传图像、生成肖像或pixel pet,运行时眼睛会扫描

行业启示

  • 多agent协作范式:Bot Mode展示了轻量级多agent系统的可行路径——通过现有基础设施扩展而非重建,降低了企业采用多agent架构的门槛,为" specialist bot + handoff chain"工作流提供了参考实现
  • 开源agent工具链成熟:从一天beta到默认功能的快速迭代,反映了开源AI工具生态的敏捷开发模式,同时明确区分个人工作站工具与企业级基础设施的适用边界(无SSO/审计日志/策略层),提醒监管行业用户需自行补充控制层
  • 桌面端agent的定位策略:Nous Research明确将Hermes Agent定位为solo builder、startup和中小工程团队的工作站工具,而非企业管理平台,这种精准定位避免了与云原生agent平台的正面竞争,开辟了差异化市场空间

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

Open Source 开源 Agent Agent LLM 大模型 Product Launch 产品发布