Workstream AI

Inspiration

Instead of manually adding new features or workflows, what if an app could take natural language requests and update its own UI and backend automatically? That became the seed for WorkStream-AI — a self-evolving workflow automator.

What it does

WorkStream-AI lets users submit feature requests and workflow automations in plain language. These requests update the project spec (.kiro/spec.yaml) and trigger regeneration so new components or workflows appear inside the app. For example, asking for a "Notes Widget" automatically creates and mounts it on the dashboard. Workflows like "Generate daily sales report" can also be scheduled and executed.

How we built it

I used Kiro initially to scaffold the project: it generated the first React + Tailwind frontend, created the .kiro spec, and guided the structure of agent hooks like evolve_ui and automate_workflow.
When I hit Kiro’s request limit and regional subscription issue, I implemented a local fallback pipeline that mirrors the Kiro spec-to-code process. This ensured I could complete the demo: the backend appends requests to spec.yaml, runs regeneration, and the frontend dynamically mounts generated components.

Challenges we ran into

• Limited Kiro quota in my region forced me to design a local regeneration fallback.
  
• Ensuring new components auto-mounted into the dashboard without manual imports required dynamic loading logic.
  
• Debugging cache and hot reload issues took time before the self-evolving loop worked smoothly.
  

Accomplishments that we're proud of

• Built a working prototype that evolves its own UI from natural language requests.
  
• Integrated workflow automation with uploaded business data.
  
• Delivered a hackathon-ready demo despite hitting tool limitations, by being resourceful and transparent.
  

What we learned

• Spec-driven development with Kiro can dramatically speed up prototyping.
  
• Hooks and specs provide a clear way to connect natural language → code.
  
• Fallback planning is essential when tools have limits; designing a self-sufficient pipeline improved reliability.

What's next

• Turn WorkStream-AI into a developer-facing service where anyone can evolve their apps via API.
  
• Expand workflows to integrate with real APIs (Slack, email, etc.).
  
• Refine the UI/UX and improve persistence (store workflows and features in a database).
  

Built With

• express.js
• hooks
• node.js
• react
• tailwind