LangGraph & DSPy: Build Controllable AI Agents with Tools

This lesson 2 captures the core architecture of LangGraph and by the end, learners should:

1. Understand the concept of stateful architecture in LangGraph

2. Understand how state evolves, how nodes write to the state, and how edges shape the graph flow

3. Write and execute their first LangGraph workflow

4. Restructure their flows by adding nodes and edges to shape their graphs

This lesson 3 goes deeper into state evolution and impact of memory on LangGraph workflows and shows:

1. How each node writes to the state as the execution flows

2. How memory allows persistence of interactions in a graph flow

3. How to trace LangGraph workflows with LangSmith

4. How to visualize a simple graph structure using ASCII

This Lesson 4 gives the agent (LLM with automony to use tools) three tools ad learners will:

1. Learn how to build a Google Knowledge Graph and Organic Search tools using SerpAPI

2. Augment an LLM in a LangGraph flow with tools and give it autonomy through nodes and edges

3. Trace and observe LLM tool calling and execution via the LangSmith platform

4. See how LLMs generate tool calling arguments and queries

This lesson 5 introduces the DSPy framework and illustrate how to use DSPy optimizers. Learners will:

1. Understand the place of DSPy in the larger prompt and context engineering in LLM apps

2. Build a DSPy flow that uses BootstrapFewShot optimizer to train a module

3. Observe how the trained module yields better prompts automatically

In this Lesson 6, we integrate LangGraph with DSPy and learners will:

1. See how LLMs in LangGraph construct their tool arguments and queries

2. Build agentic workflows that integrate DSPy into the query building layer of LangGraph flows

3. Adjust DSPy training with examples and metrics to give the effective query optimizations they desire