You'll build your workflow on the Coze Coding platform. First, you need an account and a project.
What to do
- Go to code.coze.cn
- Sign up with your CUHK email address
- Click "New Project" in the top right corner
- Name your project:
URSP3600 - [Your Name]
- Choose the Python template
- Click "Create"
The course RAG contains all the knowledge from the 13-week syllabus — concepts, reading lists, Hong Kong data sources, and project guidelines. Your AI nodes will use this to answer questions accurately.
What to do
- Go to the Resources page and download "Course RAG Knowledge Base"
- In your Coze Coding project, click the "Skills" tab on the left sidebar
- Click "Add Skill" → "Knowledge Base"
- Upload the downloaded RAG file
- Name it:
ursp3600_course_knowledge
- Wait for indexing to complete (usually 1-2 minutes)
✓
Knowledge base imported
Each node is an AI agent that handles one part of your smart city project. Drag them from the node library onto the canvas.
What to do
- Go to the Resources page and download "6 Methodology Node Templates"
- In your Coze Coding project, open the workflow canvas (the main view)
- Click the "+" button to add a new node
- For each of the 6 nodes below, create a new LLM Agent node:
- Problem Framer — transforms urban observations into structured problem statements
- Data Stack Designer — recommends Hong Kong government data sources for your domain
- System Architect — designs 4-layer system architecture with BIM/digital twin connection
- AI Composer — configures AI capabilities (RAG, prediction, NLP) for your system
- Value Canvas — evaluates public/institutional/commercial value and feasibility
- Skill Generator — packages your complete project into a reusable Skill document
- For each node, paste the corresponding System Prompt from the downloaded template file
- Set the model to
doubao-seed-2-0-lite-260215 for all nodes
✓
System prompts configured
Connect the nodes in sequence. The output of each node becomes the context for the next one. Choose one of the three preset templates.
Choose your template
- From Scratch (recommended for beginners):
Problem Framer → Data Stack → System Architect → Skill Generator
- AI First (if your project is AI-heavy):
Problem Framer → AI Composer → Value Canvas → Skill Generator
- Data Driven (if your project is data-heavy):
Data Stack → System Architect → Value Canvas → Skill Generator
How to connect
- On the canvas, hover over a node until you see the output dot (right side)
- Click and drag from the output dot to the input dot (left side) of the next node
- Repeat until all nodes are connected in your chosen sequence
- Connect the Start node to your first node, and your last node to the End node
Now it's time to run your workflow. Input your project description and let the AI nodes process it step by step.
What to do
- Click the "Run" button (top right of the canvas)
- In the input panel, fill in:
project_description: Describe your smart city project in 1-3 sentences
Example: "CUHK campus students spend too much time finding parking spots during peak hours."project_domain: Choose your domain — transport, environment, energy, building, or public_service
- Click "Execute"
- Watch each node process in sequence — you'll see the output appear step by step
- Review the final output from the Skill Generator node
The Skill Generator node produces a standardized JSON document that packages your entire project. You can install this Skill onto your own AI Agent.
What to do
- After running the workflow, find the output from the Skill Generator node
- Copy the JSON output
- Save it as a file:
my_smart_city_skill.json
- This file contains all 7 project elements:
- Problem definition
- System logic
- Data model
- BIM/digital twin connection
- AI workflow logic
- Interface design
- Implementation value
- Go to the Resources page to see the Skill document format specification
Now go beyond the course templates. Use AI to create your own knowledge base and Skill document based on your specific project domain.
What to do
- Build your own RAG:
- Collect domain-specific documents (research papers, government reports, case studies)
- In Coze Coding, create a new Knowledge Base
- Upload your documents — the platform will automatically index and vectorize them
- Name it:
[your_project]_knowledge
- Create your own Skill:
- Use the Skill Generator node with your custom RAG as input
- Or manually write a Skill document following the format specification on the Resources page
- Include your project's unique data sources, system architecture, and AI workflow
- Test it:
- Install your Skill onto a new AI Agent in Coze Coding
- Ask the Agent questions about your project domain
- Verify it can answer using your custom knowledge base
For your final project, build a complete custom workflow that combines multiple methodology nodes with your own RAG and Skills. This is your capstone deliverable.
What to do
- Design your workflow:
- Choose which methodology nodes you need (you don't have to use all 6)
- Decide the order based on your project logic
- Add any additional nodes (data processing, visualization, etc.)
- Integrate your custom RAG:
- Connect your custom knowledge base to the relevant nodes
- Make sure each node can access the domain-specific information it needs
- Run and iterate:
- Run your workflow with your actual project data
- Review the outputs and refine your node configurations
- Run multiple times until you're satisfied with the results
- Submit for Week 13:
- Export your final workflow configuration
- Include your Skill document
- Prepare a 5-minute presentation showing your workflow in action
✓
Final results generated