| Session Sheet | Recording | Slides | Repo | Homework | Feedback |
|---|---|---|---|---|---|
| Multi-Agent Applications | Recording! passcode: Ka!v5.@# |
Session 5 Slides | You are here! | Session 5 Assignment: Multi Agents | Feedback 1/27 |
BREAKOUT ROOM #1:
- Task 1: Dependencies & Environment Setup
- Task 2: Understanding Multi-Agent Systems
- Task 3: Building a Supervisor Agent Pattern
- Task 4: Adding Tavily Search for Web Research
- Activity #1: Add a Custom Specialist Agent
BREAKOUT ROOM #2:
- Task 5: Agent Handoffs Pattern
- Task 6: Building a Wellness Agent Team
- Task 7: Context Engineering & Optimization
- Task 8: Visualizing and Debugging with LangSmith
- Activity #2: Multi-Agent Consultation Pattern
You'll need API keys for:
- OpenAI - For GPT-5.2 (supervisor) and GPT-4o-mini (specialist agents)
- Tavily - For web search capabilities (free tier available at tavily.com)
- LangSmith (optional) - For tracing and debugging
- Install UV, which you can do through this resource
- Run the command
uv sync - Open your Jupyter notebook and select
.venvfor your kernel.
Run the notebook!
- Customize your multi-agent system with one of the following enhancements:
- Add a new specialist agent to the team (e.g., a Fitness Coach, Nutrition Expert)
- Implement a different multi-agent pattern (Network/Swarm, Debate)
- Add human-in-the-loop approval for certain agent decisions
- Implement context summarization to manage long conversations
- Create a diagram showing your multi-agent architecture
- Record a Loom video walking through the notebook, the questions, and your enhancements!
- Show your multi-agent architecture diagram in a Loom video and explain the agent interactions
- Make a social media post about your multi-agent wellness system and tag @AIMakerspace
- Share 3 lessons learned
- Share 3 lessons not learned
Here's a template to get your post started!
Built my first Multi-Agent AI System using LangGraph!
What I learned about multi-agent patterns:
1. Supervisor pattern = orchestrator that routes to specialist agents
2. Handoffs = agents can transfer control to each other based on expertise
3. Context engineering is critical - garbage in, garbage out
Key insight: Don't build multi-agents unless you really need them!
Start simple, add complexity only when necessary.
A huge shoutout to @AI Makerspace for making this possible.
#AI #Agents #LangGraph #MultiAgent #BuildInPublic
Note: Completing an Advanced Build earns full credit in place of doing the base assignment notebook questions/activities.
Build a Personal Wellness Planner - a multi-agent system that can create, save, and manage personalized wellness plans using file system tools.
1. Multi-Agent Architecture:
- A Planner Supervisor (GPT-5.2) that coordinates the planning process
- Specialist agents for each wellness domain (Exercise, Nutrition, Sleep, Stress)
- A File Manager Agent that handles reading/writing wellness plans
2. File System Tools: Create tools that allow agents to:
save_wellness_plan(filename, content)- Save a plan to a markdown fileload_wellness_plan(filename)- Load an existing planlist_saved_plans()- List all saved wellness plansappend_to_plan(filename, section, content)- Add a section to an existing plan
3. Workflow:
User: "Create a wellness plan for someone who wants to lose weight and reduce stress"
│
▼
┌─────────────────┐
│ Planner Super- │
│ visor (GPT-5.2) │
└────────┬────────┘
│ coordinates
┌────────────────┼────────────────┐
│ │ │
▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐
│ Exercise │ │ Nutrition│ │ Stress │
│ Agent │ │ Agent │ │ Agent │
└────┬─────┘ └────┬─────┘ └────┬─────┘
│ │ │
└───────────────┼───────────────┘
▼
┌─────────────────┐
│ File Manager │
│ Agent │
└────────┬────────┘
│
▼
📄 wellness_plan_2024.md
4. Example Output File (plans/wellness_plan_weight_loss.md):
# Personal Wellness Plan: Weight Loss & Stress Reduction
Generated: 2024-01-26
## Goals
- Lose 15 pounds over 3 months
- Reduce daily stress levels
## Exercise Plan
- Monday/Wednesday/Friday: 30-min cardio
- Tuesday/Thursday: Strength training
- Weekend: Active recovery (walking, yoga)
## Nutrition Plan
- Daily calorie target: 1,800 cal
- Meal prep on Sundays
- Increase protein intake to 100g/day
## Stress Management Plan
- Morning: 10-min meditation
- Evening: Digital detox after 8pm
- Weekly: One relaxing activity (bath, nature walk)
## Weekly Check-in Template
- [ ] Completed workouts: _/5
- [ ] Stayed within calorie target: _/7 days
- [ ] Meditation sessions: _/7- Add a Progress Tracker Agent that can update plans with completed items
- Implement plan versioning (save revisions with timestamps)
- Add web search to include latest wellness research in plans
- Complete all steps of the Main Assignment above
- Include your multi-agent implementation with file I/O tools
- Include at least 2 example generated wellness plans in a
plans/folder - Document your architecture with a diagram
- Add, commit and push your modifications to your repository
When submitting, provide:
- Your Loom video link demonstrating the planner creating and saving a wellness plan
- The GitHub URL to your completed notebook with the Advanced Build
- Screenshots of generated plan files
Follow these steps to prepare and submit your homework:
- Pull the latest updates from upstream into the main branch of your AIE9 repo:
- (You should have completed this process already.) For your initial repo setup, see Initial_Setup
- To get the latest updates from AI Makerspace into your own AIE9 repo, run the following commands:
git checkout main git pull upstream main git push origin main - IMPORTANT: Start Cursor from the
05_Multi_Agent_with_LangGraphfolder (you can also use the File -> Open Folder menu option of an existing Cursor window) - Answer Questions 1 - 4 using the
##### Answer:markdown cell below them. - Complete Activity #1 and Activity #2 in the notebook.
- Add, commit and push your modified
Multi_Agent_Applications_Assignment.ipynbto your GitHub repository.
When submitting your homework, provide:
- Your Loom video link
- The GitHub URL to your completed notebook