SenteLabsAI 推出开源虚拟高管系统 Open Executive

SenteLabsAI 推出了名为 Open Executive 的开源项目，这是一个由 8 个专家 AI Agent 组成的虚拟高管团队，通过统一高管声音为企业提供咨询 1。该团队涵盖战略、财务、人力、法务、运营、市场、产品与董事会沟通等职能，具体由 CSO、CFO、CHRO、General Counsel、COO、CMO、CPO 以及 Board Communications Director 等 8 个智能体分别担任 1。

在技术架构上，Open Executive 基于 RAG 架构，结合 ChromaDB 向量库与 Claude 模型构建 1。系统的 Executive Orchestrator 使用 claude-sonnet-4-6 模型，情景记忆提取使用 claude-haiku-4-5 模型 1。该项目支持公司文档上传、情景记忆、内置调度器及 Discord 和 Slack 等集成，并可切换至本地模型运行 1。项目内置的评估套件 evals/ 包含 29 个场景，由 claude-opus-4-7 作为 LLM-as-judge 打分，CI 门槛为平均 ≥ 3.5/5 1。

Open Executive 项目采用 Apache 2.0 许可证，仓库地址为 https://github.com/SenteLabsAI/OpenExecutive.git 1。系统首次运行需要 Python 3.11+ 和 Node 22+，初始启动会下载约 90 MB 的嵌入模型 1。在调度机制方面，调度器通过 UPDATE … RETURNING 防止重复触发，API 必须以单实例运行（max_machines_running = 1） 1。

SenteLabsAI has introduced Open Executive, an open-source project featuring a virtual executive team of eight expert AI agents designed to provide corporate consulting through a unified voice 1. This AI suite encompasses a comprehensive range of leadership functions, specifically including roles for a Chief Strategy Officer, Chief Financial Officer, Chief Human Resources Officer, General Counsel, Chief Operating Officer, Chief Marketing Officer, Chief Product Officer, and a Board Communications Director 1.

Built on a Retrieval-Augmented Generation architecture with ChromaDB vector storage, the system supports company document uploads, episodic memory, a built-in scheduler, and integrations with Discord and Slack, while also allowing users to switch to local models 1. The Executive Orchestrator operates using claude-sonnet-4-6, and episodic memory extraction is powered by claude-haiku-4-5 1. To maintain quality, the project features an evaluation suite of 29 scenarios where claude-opus-4-7 serves as an LLM-as-judge, enforcing a continuous integration threshold of an average score of at least 3.5 out of 5 1.

The project is distributed under the Apache 2.0 license and can be accessed via its GitHub repository at https://github.com/SenteLabsAI/OpenExecutive.git 1. Running the system for the first time requires Python 3.11 or later and Node 22 or later, and it will download an embedding model of approximately 90 MB during initial startup 1. Furthermore, the scheduler utilizes an UPDATE ... RETURNING command to prevent duplicate triggers, and the API is restricted to single-instance operation with the max_machines_running parameter set to 1 1.