Stelline
Enable developers to generate cross-platform robot skills using CrewAI and Copilot Kit accessing robot system profiles to produce working code.
YouTube Video
Project Description
Stelline — Prompt-to-Motion Robotics Pipeline
What it does (running local code)
Natural-language prompt → motion
Example: “Drive a one-meter square.”
CrewAI agent crew (Planner → Coder → QA) generates a fully-tested Python script that uses the Yahboom Transbot library.
Auto-deploy via SSH to the Jetson-Nano-powered Transbot; the robot performs the maneuver live.
A Next.js 14 dashboard streams every stage—prompt, code diff, unit-test results, on-device logs—so judges watch the entire flow in real time.
How it meets the judging criteria
Criterion Stelline Response
Running code End-to-end pipeline executes live on physical hardware during the demo.
Innovation & creativity Bridges LLM-agent workflows with real-time robotic deployment, turning hours of firmware work into a single prompt.
Real-world impact Slashes prototyping time for educators, hobbyists, and industrial integrators; lowers the barrier for non-programmers.
Theme alignment Demonstrates practical AI agents orchestrating planning, coding, testing, and deployment—exactly the hackathon focus.
Tech stack & tools
Front end: Next.js 14 App Router, React, TypeScript, TailwindCSS
Agent orchestration: CrewAI framework using OpenAI GPT-4o-mini
Code generation & tests: Python 3.10, auto-generated pytest, smolagent patterns
Hardware & libs: Yahboom Transbot SDK on Jetson Nano (Ubuntu 22.04)
Ops: SSH deployment, Git version control, GitHub Actions CI
By uniting a visual workflow UI with an autonomous agent crew and real robot hardware, Stelline shows how AI agents can step out of the chat box and power tangible, real-world systems—instantly.
Prior Work
I had the robot setup the night before.