A curated collection of Astron workflow examples and user cases for AI-powered applications
Have a question about Astron Agent or Astron RPA — deployment, model access, troubleshooting, or how a workflow is built? Two ways to get an answer grounded in this project's knowledge:
| Where | What it answers from | |
|---|---|---|
| 🟢 Live AI Chat | awesome-astron-workflow.dev/chat | Answers only from the curated knowledge-base/ (FAQ + anonymized customer cases + cleaned community Q&A) plus official iFLYTEK/Astron docs. |
| 🤖 Ask DeepWiki | deepwiki.com/FenjuFu/Awesome-Astron-Workflow | Indexes the whole repository, including knowledge-base/, and answers across all of it. |
📚 Knowledge base source:
knowledge-base/is the desensitized corpus behind both. See knowledge-base/README.md for layout and how it is compiled into the retrieval index. DeepWiki indexes this folder automatically as part of the repository — no extra configuration is needed for it to answer from these documents.
Warning
Zeabur will shut down completely by 2026-12-31 (announcement). Dedicated server renewals stop on 2026-12-01, and servers shut down on their individual expiration dates. Our community demos may go offline earlier when their server expires. Use the official SaaS, RepoCloud / Railway or the official Docker Compose files instead.
Use iFLYTEK's official SaaS for a stable environment, or deploy your own copy of the three iFLYTEK Astron open-source stacks. Existing Zeabur demo links remain at awesome-astron-workflow.dev/demo while the demos are available; availability through the platform's final shutdown date is not guaranteed.
| Project | Official SaaS | RepoCloud | Railway | Docker Compose | Zeabur |
|---|---|---|---|---|---|
| Astron Agent | agent.xfyun.cn | Marketplace listing | — | docker/astronAgent | Template |
| Astron RPA (server stack) | iflyrpa.com → Cowork | AI deploy | — | docker | Template |
| SkillHub | skill.xfyun.cn | AI deploy | skillhub-registry (v0.2.22) · skillhub (v0.1.0) | compose.release.yml | Template |
- Official SaaS: iFLYTEK's hosted service. Sign in with your own iFLYTEK account. Accounts and data are separate from the open-source demos, and features may differ from the open-source version. The RPA SaaS (Xingchen RPA) is built around a desktop client and has been upgraded to Xingchen Cowork.
- RepoCloud: an AI agent deploys the GitHub repo to a VPS in about 5–15 minutes. Astron Agent has a marketplace listing. Astron RPA and SkillHub go through RepoCloud's generic AI deploy.
- Railway: community SkillHub templates. The v0.2.22 one includes the security scanner and needs the Hobby plan or higher.
- Docker Compose: the Compose files in each official repository. Run them on any server you control. Astron Agent needs about 8 GB RAM minimum.
- Zeabur (shutting down): legacy template links are retained for reference and may already be unavailable. If you deployed with them, back up your data and migrate before your server expires. Zeabur also disclosed a security incident on 2026-10-14, so rotate the passwords and API keys used in that deployment.
- The RepoCloud, Railway and Zeabur entries are third-party or community deploy entries, not official iFLYTEK releases. Change every default password and secret before real use. The Astron RPA stacks deploy the server only. Install the desktop client separately.
- Live Demos & One-Click Deploy
- Overview
- Use Cases & Workflows
- Contributing
- License
Astron is a powerful workflow engine designed for building AI-driven applications. This repository collects real-world user cases and their corresponding workflow implementations to help developers understand how to build various AI applications with Astron.
Each case includes:
- A user case document (WeChat article)
- A complete workflow YAML file
- A summary of key features
Description: An "AI for Everyone" comprehensive assistant workflow. It integrates a multi-intent recognition system to intelligently distinguish user needs (such as consultation, order queries, or casual chat), combines online search for systematic knowledge organization, and uses a specialized knowledge base to provide precise business Q&A services.
- User Case: Video Showcase
- Workflow: bks-AI.yml
Key Features:
- Smart Intent Routing: Accurately identifies consultation, order queries, and casual chat scenarios
- Deep Online Search: Real-time retrieval of network information with structured summarization
- Exclusive Knowledge Base: Provides professional business answers related to Bks-AI
- Multi-source Recall Fusion: Combines search results, knowledge base, and LLM generation for responses
Description: A smart resume optimization tool that helps job seekers create professional resumes with AI-driven suggestions and formatting.
- User Case: WeChat article
- Workflow: AI-Resume-Assistant.yml
Key Features:
- Resume content analysis and optimization
- AI-powered writing suggestions
- Professional formatting templates
- Industry-specific keyword recommendations
- Multi-format export support
Description: An advanced translation tool that provides accurate, context-aware translations, supporting multiple languages and domains.
- User Case: WeChat article
- Workflow: aitranslator.yml
Key Features:
- Multi-language translation support
- Context-aware translation engine
- Domain-specific translation modes
- Batch translation capability
- Translation quality evaluation
Description: An agent that turns scattered learning notes or work records into structured knowledge cards. Users simply input raw notes, and the agent automatically parses, refines key points, and generates readable knowledge cards for easy review and recall.
- User Case: WeChat article
- Workflow:
knowledge-cards.yml
Key Features:
- Auto-parse raw note text
- AI restructuring and key-point extraction
- Generate structured knowledge cards
- Customizable card templates
- Export in multiple formats
Description: A content summarization tool that extracts and condenses articles from WeChat Official Accounts into concise summaries using the DeepSeek model.
- User Case:
- Workflow: default.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Extract content from URLs
- AI-powered summarization (DeepSeek-V3.2)
- Concise output (~350 words)
- Markdown format support
Description: An advanced resume analysis and job search assistance system supporting PDF resume OCR parsing, intelligent assessment, and job recommendations.
- User Case: Resume Intelligence add
- Workflow: default.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- PDF resume OCR parsing
- Multi-turn interaction for information completion
- Intelligent resume assessment and job recommendations
- Orchestration of multiple large models
Description: An intelligent Q&A assistant designed to answer questions based on a specific documentation knowledge base. It acts as a dedicated assistant (e.g., for HR inquiries) providing accurate responses solely from the provided documents.
- User Case: Video Showcase
- Workflow: Docs Workfow.yml
Key Features:
- Knowledge Base Integration: Retrieves information from a specific document repository
- Role-Based Q&A: Configured as a dedicated assistant (e.g., HR support)
- Strict Context Adherence: Answers based solely on provided documentation to ensure accuracy
- Concise & Friendly: Delivers clear, brief, and professional responses
Description: An open-source AI Research Assistant workflow for Astron Agent. It automates deep online research by breaking down topics, searching the web, scraping content, and compiling structured research reports with citations.
- User Case: Video Showcase
- Workflow: ai-research-assistant
Key Features:
- Automated research planning and execution
- Deep web search and content scraping
- Structured report generation with citations
- Multi-source information synthesis
Description: A digital employee for hotel management that provides intelligent room booking, inquiry, and customer service support to improve hotel operational efficiency.
- User Case: Video Showcase
- Workflow: 酒店管理数字员工.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Intelligent hotel room inquiry and booking
- Automated handling of common customer questions and service requests
- Enhances hotel front desk and customer service operations
- 24/7 digital employee service support
Description: A smart customer service workflow designed for the used car export business. It answers inquiries regarding car models, quotes, and export processes, providing professional business support.
- User Case: Video Showcase
- Workflow: 二手车出口智能客服.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Intelligent inquiry for used car models and quotes
- Automated consultation on vehicle export processes and policies
- Multi-language interaction to support international customers
- Improves customer response efficiency for used car export business
Description: A teaching plan review tool for instructional design scenarios. It supports submitting teaching plan titles, applicable grades, and content, and returns intelligent review results, improvement suggestions, and historical records.
- User Case: Video Showcase
- Workflow: 教案评审员.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Intelligently parses teaching plan content and auto-generates review results
- Provides targeted improvement suggestions for instructional design
- Automatically records teaching plan review history for traceability
- Multi-dimensional evaluation combining applicable grades and subjects
Description: A comprehensive content creation tool that generates platform-specific content optimized for different social media and publishing channels.
- User Case: WeChat article
- Workflow: Multi-platform content writing.yml
Key Features:
- Platform-specific content generation
- Multi-format content support
- SEO optimization
- Scheduling and publishing
- Analytics and performance tracking
Description: An automated podcast creation tool that converts text into engaging audio experiences with natural speech synthesis and audio editing.
- User Case: WeChat article
- Workflow: AI Podcast.yml
Key Features:
- Text-to-speech conversion
- Natural speech synthesis
- Audio editing and enhancement
- Multi-speaker support
- Podcast distribution integrations
Description: Enter a novel title and automatically generate a complete article with accompanying images. Built with Astron’s block-style workflow composition, you can create an agent in one click, enable web access, image generation, and code capabilities in advanced settings, and publish to the Agent Plaza for reuse.
- User Case: WeChat article
- Workflow: 小说图文生成.yml
Key Features:
- One-input novel title → full image-rich article
- Prompt-based agent creation with auto-completion
- Advanced config: web access / image gen / code tools
- Visual block-based workflow authoring
- Publishable to Agent Plaza for convenient reuse
Description: Transforms poetic verses or keywords into visual art by analyzing literary imagery and generating optimized prompts for AI image generation models.
- User Case:
- Workflow: default.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Deep analysis of literary artistic conception
- Generation of optimized visual descriptions
- Integration with Keling AI painting model
- From text to visual art realization
Description: An intelligent assistant that converts text or WeChat Official Account article links into lively and interesting radio podcast content, supporting speech synthesis and cover generation.
- User Case:
- Workflow: AI.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Parse URL content (supports WeChat Official Accounts, etc.)
- Generate podcast-style scripts
- Ultra-realistic speech synthesis
- AI text-to-image cover generation
Description: An intelligent course creation platform that generates educational content, lesson plans, and learning materials based on specified topics and learning objectives.
- User Case: WeChat article
- Workflow: Course Generator workflows
Key Features:
- Automated course content generation
- Alignment with learning objectives
- Interactive content creation
- Assessment and quiz generation
- Progress tracking and analytics
Description: An AI-powered academic paper recommendation system that works like TikTok for research papers—users swipe through daily recommendations, like or save papers, and the agent learns preferences to improve suggestions.
- User Case: WeChat article
- Workflow:
spark_workflow_config.yml
Key Features:
- Personalized recommendations based on user profile and interaction history
- Daily batch of curated papers fetched via ArXiv API
- Preference learning with feedback loop (like / save / skip)
- Summaries and recommendation rationale via LLM
- Frontend swipe UI and backend service integrating Astron Agent
Description: An AI-powered English essay grading and feedback assistant. Users submit their essay content, and the agent provides structured evaluation, points out grammar and expression issues, suggests improvements, offers example sentences/paragraphs, and estimates the essay level using IELTS or CEFR standards.
- User Case: WeChat article
- Workflow:
AI-English-Essay-Assistant.yml
Key Features:
- Structured evaluation covering organization, grammar, vocabulary, and logic
- Grammar error detection with specific correction suggestions
- Improved example sentences or revised paragraphs
- Proficiency estimation under IELTS or CEFR standards
- Clear, organized output format tailored for student feedback
Description: An "practice-feedback" closed-loop system for middle-school students to improve English writing. Parents/teachers generate prompts in one click; students write online; AI instantly scores and gives improvement tips, while auto-saving history to help kids progress steadily.
- User Case: WeChat article
- Workflow:
english_writting_helper.yml
Key Features:
- Grade-based random prompts from model essay knowledge base
- AI auto-scoring: grammar, vocabulary, structure, logic
- Instant feedback: error location + revision advice + high-score examples
- History log: all practices & comments saved for review
- One-click deploy: Flask web app + local Astron workflow
Description: The "History Knowledge QA" agent is your private history consultant, specializing in global historical events, cultural evolution, and important figures. Whether you are interested in the mysteries of ancient civilizations or turning points in modern history, it can provide accurate information and answer every question you have.
- User Case:
- Workflow: 2.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Real-time online search for information
- Text compliance detection to ensure safety
- Professional historical role-playing
- Precise history knowledge Q&A
Description: A specialized assistant for KET exam essay writing. It analyzes requirements, generates outlines, writes drafts, polishes language, and checks word count compliance.
- User Case:
- Workflow: 20260107220916.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Essay requirement analysis
- Structured outline generation
- Draft writing & language polishing
- Word count compliance check
Description: An importable Astron DSL v1 workflow that calls the official MemFlywheel SDK through a local HTTP adapter to perform explicit memory saves, prompt-time recall, and filesystem-backed verification in recall-only mode. This is not a native Astron integration.
- User Case: Zhihu technical tutorial
- Workflow: astron-agent-memflywheel.yml
- DSH × Spark Ultra example: configuration, portable runner, and sanitized SKILL
Key Features:
- Astron code node calls a local HTTP adapter
- Real integration with the official MemFlywheel SDK
- Explicit memory save and prompt-time recall
- Filesystem-backed evidence in
MEMORY.mdandworkflow/*.md - Clear
recall-onlyand non-native integration boundaries - Verified goal-only DSH turn in which Spark Ultra selected
skill → pwsh → pwsh
Description: An intelligent stock analysis agent that aggregates market data, performs technical and fundamental analysis, and provides investment insights.
- User Case: Video Showcase
- Workflow: Astron-AI-Agent-StockAnalyzer
Key Features:
- Real-time stock data fetching
- Technical indicator analysis
- Fundamental data assessment
- Investment report generation
Description: An intelligent travel planning assistant that helps users design personalized travel routes, book accommodations, and provides real-time travel advice based on user preferences and budget.
- User Case: CSDN Blog
- Workflow: Travel Companion.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Personalized travel route planning
- Accommodation and transportation booking assistance
- Real-time travel advice and emergency support
- Integration with maps and weather services
Description: Your personal film industry guardian and movie expert. Specialized in identifying movies from images, providing professional recommendations, and ensuring high-quality viewing experiences by filtering out bad movies.
- User Case:
- Workflow: Pro.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Search movies by image (Image Understanding)
- Professional movie recommendations & reviews
- Real-time web search for movie info
- Compliance and safety filtering
Description: Find the latest online novels by entering the type of novel you want to find, and provide a brief introduction.
- User Case:
- Workflow: default.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Extract novel type from user input
- Aggregate search for latest novels
- LLM-based recommendation and intro
- Personalized reading suggestions
Description: A professional nutrition consultant for your dogs. Provides customized dietary plans based on age, weight, and specific needs, backed by real-time nutritional knowledge search.
- User Case:
- Workflow: default.yml
- Hashtag Event: Astron Training Camp · Cohort #1
Key Features:
- Customized nutrition plans (Age/Weight/etc.)
- Professional dietary advice via LLM
- Web search for nutritional knowledge
- Visual aids with Text-to-Image generation
Description: A complete Astron workflow that accepts synthetic or de-identified clinical text, calls an OpenRare-compatible local bridge, maps phenotype descriptions to auditable HPO clues, and returns a structured result for professional review.
- User Case: English video demo
- Chinese Workflow: astron-openrare-clue-extraction.zh-CN.yml
- English Workflow: astron-openrare-clue-extraction.en-US.yml
Key Features:
- Complete input, extraction, HPO formatting, and result-return workflow
- OpenRare-compatible extraction with CHPO dictionary-constrained mapping
- Auditable HPO identifiers, source commit, provider model, and run trace
- Configurable local bridge through
OPENRARE_BRIDGE_BASE_URL - Designed for clue organization and professional review, not diagnosis
Description: A complete Astron workflow that runs a real PubMed MCP search and detail lookup, asks Spark Ultra to create a readable evidence card, and verifies that every PMID and source URL came from the same run.
- User Case: English video demo
- Chinese Workflow: astron-pubmed-evidence-card.zh-CN.yml
- English Workflow: astron-pubmed-evidence-card.en-US.yml
Key Features:
- Real PubMed MCP search and article-detail retrieval
- Spark Ultra evidence-card generation
- PMID and source-URL consistency checks
- Bridge credentials read from runtime environment variables
- Research-reading output with a medical-safety boundary
Description: A tested Astron workflow that synchronizes a shared workbench through a clearly labeled local UI adapter, returns five auditable page records, and keeps payment, messaging, and public publishing under human approval.
- User Case: Tested integration notes
- Workflow: astron-open-ai-workbench-hands-on.zh-CN.yml
Key Features:
- Four-node Astron workflow completed in 6.94 seconds with zero model tokens
- Idempotent local-adapter synchronization with a real Astron table
- Stable JSON page contract and five fictional sandbox records
- Real database, debugger, user-preview, and publishing-request UI evidence
- Explicit non-native integration and human-approval boundaries
The /activities/admin page uses an admin password configured in .env:
VITE_ADMIN_PASSWORD=your_secure_password_hereAI Chat also requires server-side environment variables. Before local development, copy .env.example to .env and provide at least:
ASTRON_MAAS_API_KEY=your_astron_maas_api_key
ASTRON_MAAS_MODEL=spark-x2.6-4b
ASTRON_MAAS_BASE_URL=https://maas-api.cn-huabei-1.xf-yun.com/v2/chat/completions
ASTRON_MAAS_MAX_TOKENS=32768
ASTRON_MAAS_TEMPERATURE=0.7If ASTRON_MAAS_API_KEY is missing, the /chat page returns an "AI Chat is temporarily unavailable" message. When deploying to Vercel or another hosting platform, set the same environment variables there as well.
The chat uses Spark-X2.6-4B. Enable access and quota for this model on your MaaS API key before deploying; a Coding Plan key may not have access to the general MaaS endpoint. Update both ASTRON_MAAS_MODEL and ASTRON_MAAS_BASE_URL in existing deployments, then redeploy. Model selection is controlled by the server, so cached clients cannot override it with the old model name.
For a self-hosted OpenAI-compatible Spark-X2.6 service, set ASTRON_MAAS_BASE_URL to its full chat completions endpoint and ASTRON_MAAS_MODEL to its served model ID (for example, spark2.6 in the official SGLang example). The Hugging Face collection URL is a download page, not an inference endpoint. Update the chat consent notice and privacy policy if you change the inference provider.
We welcome contributions to this repository! If you have an Astron workflow or user case to share:
- Fork this repository
- Create a new branch for your contribution
- Add your workflow and user case information
- Submit a pull request with a detailed description
Please make sure your contribution includes:
- A clear description of the use case
- A link to the user case document (if available)
- A complete workflow YAML file
- A summary of key features
This repository is licensed under the MIT License. For details, see the LICENSE file.
Note: This repository is a community-driven collection of Astron workflow examples. All user case links and workflow files are the property of their respective owners; references are provided for educational and informational purposes only.