Google Colab Notebooks

Hands-on Coding Resources for AI Agent Development

Getting Started with Colab

If you’re new to Google Colab, here are some tips:

  1. Save a Copy: Always use “File → Save a copy in Drive” before editing
  2. Runtime: Make sure you select the appropriate runtime (usually Python 3)
  3. API Keys: Store sensitive keys in Colab Secrets, never hardcode them
  4. Dependencies: Run the pip install cells at the beginning of each notebook
  5. Execution: Run cells in order from top to bottom for best results

Course Colab Notebooks

All notebooks are hosted on Google Colab for easy access and execution. It’s recommended to save a copy to your own Google Drive before making changes.


Week 1 - Getting Started with AI Agents

Project: Introduction to LangChain and LLM Basics

Description: Get started with AI agents by learning the fundamentals of LangChain and Google’s Gemini API. This notebook walks you through setting up your development environment, securing API keys using Colab Secrets, and building your first simple chat agent. Perfect for beginners with hands-on exercises in prompt engineering and context management.

Topics Covered: - Gemini API setup and authentication - LangChain basics and message types - Prompt engineering techniques - Context engineering principles - Model parameters and customization - Chaining LLM calls

Open in Colab →


Week 2 - Workflows and Agents

Project: Understanding LLM Architecture Patterns

Description: Learn the critical differences between workflows and agents, and when to use each pattern. This notebook provides hands-on experience with LangGraph for building structured workflows and introduces you to agent-based architectures. You’ll diagram real-world use cases and translate them into code.

Topics Covered: - Workflows vs. Agents comparison - LangGraph fundamentals - Structured workflow design - Agent routing patterns - Multi-step task orchestration - Choosing the right architecture pattern

Open in Colab →


Week 3 - MCP Lesson

Project: Building AI Agents with Model Context Protocol (MCP)

Description: Learn how to connect AI agents to external services using the Model Context Protocol (MCP). This hands-on notebook teaches you to set up Canvas MCP connections, build agents that use tools, and understand the complete flow from user input to tool execution and back. You’ll create an agent that can interact with Canvas API and other external services.

Topics Covered: - MCP Server setup and configuration - Creating and using tools within an agent - Canvas MCP integration - Understanding the User → AI → MCP Client → MCP Server → Tool flow

Open in Colab →


Week 4 - Wellness Agent (LangSmith Diagnostics)

Project: Health and Wellness Tracking Agent with Observability

Description: Build a wellness tracking agent that logs food intake, exercise, and sleep patterns while learning production-level observability with LangSmith. This notebook demonstrates how to trace agent execution, identify bottlenecks, and debug complex agent behaviors. Perfect for understanding how to monitor AI systems in production.

Topics Covered: - LangSmith tracing and observability - Building stateful agents with memory - Creating custom tools for wellness tracking - Debugging agent tool calls and responses - Using @traceable decorators for visibility

Open in Colab →


Week 4 - Canvas AI Tutor (LangSmith Diagnostics)

Project: Intelligent Canvas Integration Agent

Description: Create an AI tutor that integrates with Canvas LMS using MCP to help students access assignments, grades, and course materials. This notebook focuses on real-world agent diagnostics, showing how to trace complex MCP interactions and optimize agent performance using LangSmith’s evaluation tools.

Topics Covered: - Canvas API integration via MCP - Multi-step agent workflows - LangSmith evaluation datasets - Trace tree analysis for debugging - Production-ready error handling

Open in Colab →


Week 5 - SQL Assistant with Progressive Disclosure

Project: Building Smart Database Agents

Description: Master the progressive disclosure pattern by building an SQL assistant that loads database schemas on-demand instead of overwhelming the context window. This notebook teaches you to create skill-based tools that inject knowledge only when needed, dramatically reducing token costs while maintaining accuracy. Scale to massive databases with 100+ tables efficiently.

Topics Covered: - Progressive disclosure pattern for context management - On-demand skill loading - Database schema organization - Stateful agents with memory - Constrained tools with prerequisites - Cost optimization strategies (10-50x reduction)

Open in Colab →


Week 6 - Wellness Agent (Production & Scaling - BROKEN)

Project: Production-Ready Multi-User Wellness Agent (Debugging Exercise)

Description: This is an intentionally broken production deployment of the Wellness Agent designed as a hands-on debugging and scaling exercise. Learn to identify and fix common production issues including token explosion, context overflow, memory leaks, and concurrency problems. Experience real-world challenges of scaling AI agents from prototype to production.

Topics Covered: - Context management at scale - Token cost optimization - Multi-user state management - Production hardening techniques - Identifying scaling bottlenecks - Memory management strategies

Open in Colab →


Week 7 - Data Drift and Retraining Simulation

Project: Understanding Model Drift and Weight Updates

Description: Simulate data drift scenarios and learn when and how to retrain machine learning models. This notebook demonstrates how model weights become outdated as data distributions change over time, and teaches you systematic approaches to detecting drift and maintaining model accuracy in production environments.

Topics Covered: - Understanding data drift and concept drift - Simulating distribution changes - Detecting model degradation - Retraining strategies and triggers - Weight update methodologies - Production monitoring techniques

Open in Colab →


Support

If you encounter issues with any notebook: - Check that all dependencies are installed - Verify your API keys are correctly configured in Colab Secrets - Review the error messages carefully - Consult the course modules for additional context - Ask for help in the course discussion forum