Welcome to the iFLYTEK Open Source Community 👋

[FenjuFu]

Welcome to the iFLYTEK Open Source Community 👋

Welcome to the official GitHub Discussions space for the iFLYTEK open-source ecosystem.

This space is intended for:

• Technical discussions
• Community collaboration
• Contributor onboarding
• Feature ideas and roadmap discussions
• Questions related to Astron Agent, Astron RPA, SkillHub, and related projects

Open-source projects

Astron Agent

Enterprise-grade, commercial-friendly agentic workflow platform for building next-generation SuperAgents.

Astron RPA

AI-native RPA platform designed for automation workflows and intelligent digital workers.

SkillHub

Open-source agent skill registry and governance platform for enterprise environments.

Community participation

We welcome:

• Issues
• Pull Requests
• Feature discussions
• Documentation improvements
• Ecosystem integrations

Community guidelines

Please:

• Be respectful and collaborative
• Keep discussions technical and constructive
• Use Discussions for questions and ideas
• Use Issues for bug reports and feature tracking

We look forward to building the open-source AI ecosystem together 🚀

[lostlight530]

Start Here ｜ 一眼看懂讯飞 Astron 开源生态

只回答四个问题

一、讯飞开源在做什么
不是开源几个工具类库，是把智能体从"能聊"推进到"能干活"

Astron Agent 负责智能体的工作流编排与多智能体协作
Astron RPA 把智能体的决策转化为真实系统的自动化执行
SkillHub + iFly Skills 提供可复用、可治理的技能市场
MemFlywheel 让技能随使用持续演化，形成长期记忆闭环
其余项目覆盖终端接入、执行引擎、教程体系与垂直模型验证
核心命题：智能体从理解意图，到规划执行，到技能复用，到记忆演化，全链路打通

二、核心主线
九个项目不是并列关系，是一条流水线

思考层：Astron Agent — 规划 · 编排 · 决策
技能层：SkillHub / iFly Skills — 技能复用 · 治理 · 分发
执行层：Astron RPA — 自动化 · 跨系统操作
记忆层：MemFlywheel — 长期记忆 · 技能演化 · 闭环
终端层：各终端项目 — 接入 · 呈现 · 交互

四个问题一条线回答：

智能体怎么想 → Astron Agent（规划与编排）

能做什么 → SkillHub + iFly Skills（技能池）

怎么做到 → Astron RPA（执行引擎）

怎么越做越好 → MemFlywheel（记忆与演化）

三、项目之间如何连接
官方全景图已经给出了底层关系，只是还没被讲成故事——

Astron Agent 是大脑：接收任务，拆解步骤，调度技能，生成工作流
SkillHub 是工具箱：Agent 从这里拉取技能，也可以把新技能注册回来
iFly Skills 是官方技能基座 + 治理规则：定义什么是合格技能，怎么审核，怎么版本管理
Astron RPA 是手：Agent 决定"需要打开浏览器填表"，RPA 就去执行
MemFlywheel 是记忆：这次填表踩了坑，下次自动规避，技能参数自动调优
其他项目是外设与验证：终端把结果呈现给人，教程降低入门门槛，垂直模型验证特定场景效果

不是一个项目调用另一个的简单关系，是数据从左流到右、反馈从右回到左的闭环

四、开发者从哪里开始
第一天：跑通 Astron Agent Quickstart → 理解"智能体怎么想"
第三天：浏览 SkillHub → 看看社区有什么现成技能可以复用
第一周：发布第一个 iFly Skill → 理解技能注册与治理流程
第二周：接 Astron RPA → 让智能体的决策落地执行
第一个月：接 MemFlywheel → 让技能随使用自动进化

不需要九个仓库全看完才动手，只需要抓住主线：Agent 想 → SkillHub 取技能 → RPA 执行 → MemFlywheel 记住并改进

结尾
讯飞 Astron 不是九个仓库的集合，是一套智能体操作系统——每个项目各管一层，连起来就是完整闭环

建议在社区 Discussion 中用这四个问题作为入口，把官方全景图从"目录"变成"故事"

[lostlight530]

English Translation
Start Here | A One-Minute Overview of iFLYTEK Astron Open Source Ecosystem

Four questions, four answers.

1. What is iFLYTEK Astron doing?

It's not about open-sourcing a few utility libraries. It's about moving agents from "just chat" to "real action."

Astron Agent: Workflow orchestration and multi-agent collaboration

Astron RPA: Translating agent decisions into automated execution across real systems

SkillHub + iFly Skills: A reusable, governable skill marketplace

MemFlywheel: Continuous skill evolution through usage, forming a closed-loop long-term memory

Other projects cover terminal access, execution engines, tutorial systems, and vertical model validation.

Core proposition: A complete end-to-end pipeline — from intent understanding, to planning and execution, to skill reuse, to memory evolution.

2. What is the main thread?

The nine projects are not a flat list. They are a pipeline.

Thinking Layer: Astron Agent — Planning · Orchestration · Decision-making

Skill Layer: SkillHub / iFly Skills — Reuse · Governance · Distribution

Execution Layer: Astron RPA — Automation · Cross-system operations

Memory Layer: MemFlywheel — Long-term memory · Skill evolution · Closed loop

Terminal Layer: Terminal projects — Access · Presentation · Interaction

Four questions, one continuous thread:

How does the agent think? → Astron Agent (planning & orchestration)

What can it do? → SkillHub + iFly Skills (skill pool)

How does it do it? → Astron RPA (execution engine)

How does it improve? → MemFlywheel (memory & evolution)

3. How do the projects connect?

The official landscape already shows the underlying relationships — it just hasn't been told as a story yet.

Astron Agent is the brain: Receives tasks, breaks them down, schedules skills, generates workflows

SkillHub is the toolbox: Agents pull skills from here, and can also register new skills back

iFly Skills is the official skill foundation + governance rules: Defines what qualifies as a skill, how it's reviewed, how versions are managed

Astron RPA is the hands: When the agent decides "I need to open a browser and fill out a form," RPA executes it

MemFlywheel is the memory: If something went wrong this time, it's automatically avoided next time — skill parameters self-tune

Other projects are peripherals and validators: terminals present results to users, tutorials lower the entry barrier, vertical models validate performance in specific scenarios.

It's not a simple "Project A calls Project B" relationship. It's a closed loop where data flows left to right, and feedback flows right to left.

4. Where should a developer start?

Day 1: Run Astron Agent Quickstart → Understand "how the agent thinks"

Day 3: Browse SkillHub → See what skills the community already has

Week 1: Publish your first iFly Skill → Understand the skill registration and governance process

Week 2: Connect Astron RPA → Turn agent decisions into real execution

Month 1: Connect MemFlywheel → Let skills evolve automatically through usage

You don't need to read all nine repositories before you start. Just follow the main thread: Agent thinks → pulls skills from SkillHub → RPA executes → MemFlywheel remembers and improves.

Closing

iFLYTEK Astron is not a collection of nine repositories. It's an agent operating system — each project owns one layer, and together they form a complete closed loop.

Recommendation: Use these four questions as the entry point in community Discussions. Turn the official landscape from a "table of contents" into a "story."