Topic 4: Exploring Tools

Learning Goals

• Now how to use ToolNode and create_react_agent to further abstract agent code

• Know how to invoke tools that are on different servers

• Become familiar with a wide range of tools for LLMS, both those built into LangGraph and popular 3rd party tools

Tasks

1. Let's show off our work from last week! Please connect to the Zoom and be ready to share your desktop. Note that this is not a remote class, we are simply using Zoom for screen sharing.

   https://virginia.zoom.us/j/91761114965?pwd=u9upwQ1eaEqCKZkxOOQsjv7TykeePX.1

   Meeting ID: 917 6111 4965 Passcode: 297850

2. New partners! Find your new partner here and go sit together and introduce yourselves.

3. Study the code of the toolnode_example.py and the react_agent_example.py. Run them and and look at the Mermaid graphs they produce. Answer the following questions in your portfolio:

   1. What features of Python does ToolNode use to dispatch tools in parallel? What kinds of tools would most benefit from parallel dispatch?

   2. How do the two programs handle special inputs such as "verbose" and "exit"?

   3. Compare the graph diagrams of the two programs. How do they differ if at all?

   4. What is an example of a case where the structure imposed by the LangChain react agent is too restrictive and you'd want to pursue the toolnode approach?

4. Skim through this agent_tools_reference.

5. Work with your partner to implement one of these 2-Hour Agent Projects. Start with the bare minimum functionality and add features later as time allows. Aim to get your project completed by the end of class Thursday. It is okay to use a coding agent as long as you understand and can explain the code it writes.

Resources

toolnode_example.py

react_agent_example.py

agent_tools_reference

2-Hour Agent Projects

MoltReader, a text-to-speech reader for MoltBook, A Social Network for AI Agents

Portfolio

Create a subdirectory in your GitHub portfolio named Topic4Exploring and save your programs, each modified version named to indicate its task number and purpose. Create appropriately named text files saving the outputs from your terminal sessions running the programs. Create README.md with a table of contents of the directory.