Velu推出多智能体流水线以自动生成产品帮助中心
Start here
2026年11月4日，Velu团队公布了一套七阶段多智能体流水线方案，仅通过账号登录浏览器界面，即可自动生成包含文字、截图与视频的产品帮助中心。为解决单Agent记忆超载与成本指数级增长问题，该系统实行分工协作：探索调研、前端数据探测和截屏等机械任务分配给低成本的Haiku模型；规划、并发按任务写作与编辑审核则由Sonnet模型负责。此外，流水线通过注入一致的模拟数据以及SKILL.md规范，以确保文档的质量与样式统一。
AI-generated from reports · Updated 17 hours ago
Report timeline
Follow the reports to see different sides of the story.
- Hacker News · AISelectedVelu Details Multi-Agent Pipeline for Automatically Generating Product Help Centers
The Velu team has built a seven-stage multi-agent pipeline that automatically generates product help centers—complete with text, screenshots, and video—simply by logging into a browser interface. The system delegates routine exploration, frontend data probing, and screenshot capture to lower-cost Haiku models, while reserving Sonnet for planning, concurrent task-based writing, and editorial review. This architecture eliminates single-agent memory overload and exponential cost growth, while ensuring documentation quality and consistent styling through mock-data injection and SKILL.md guidelines.
Story heat trend
Current heat 6·Comparable peak 10（Nov 4 03:00）·Comparable change, last 24 hours–
Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.