/good-agentic-coach-talk

Primary LanguageTypeScriptMIT LicenseMIT

Intro

This directory contains a simple LangGraph. This directory contains a single graph, located inside the index.ts file.

Setup

To set up the intro project, install the dependencies:

yarn install

Environment variables

The intro project requires Tavily and OpenAI API keys to run. Sign up here:

Once you have your API keys, create a .env file in this directory and add the following:

TAVILY_API_KEY=YOUR_API_KEY
OPENAI_API_KEY=YOUR_API_KEY

LangGraph Config

The LangGraph configuration file for the intro project is located inside langgraph.json. This file defines the single graph implemented in the project: simple_agent.

Tavily

Follow these steps to obtain access to the Tavily API:

  1. Sign Up

  2. Access the Dashboard

    • Once logged in, navigate to the Dashboard.
    • Here, you can manage your API keys, monitor usage, and configure settings.

  3. Generate Your API Key

    • Go to the API Keys section in the dashboard.
    • Click on Generate API Key to create a new API key.
    • Copy the API key, as you will need it for all API requests.

Editing the project

To edit the project, open the index.ts file and make changes to the graph.

Code Regions

  1. Imports

    This section imports necessary modules and components from @langchain/langgraph, @langchain/core/messages, and model.js.

    import { ToolNode } from "@langchain/langgraph/prebuilt";
import {
  END,
  MessagesAnnotation,
  START,
  StateGraph,
} from "@langchain/langgraph";
import { AIMessage, BaseMessage, HumanMessage } from "@langchain/core/messages";
import { TavilySearchResults } from "@langchain/community/tools/tavily_search";
import { model } from "model.js";

    Look at model.js, it contains the model that will be used in the graph. We are using the Groq API.

  2. Tools

    const webSearchTool = new TavilySearchResults({
  maxResults: 4,
});
const tools = [webSearchTool];
const toolNode = new ToolNode(tools as any);
    • Initializes a web search tool with a maximum of 4 results.
    • Creates a ToolNode with the initialized tools.

  3. Model

    const callModel = async (state: typeof MessagesAnnotation.State) => {
  const { messages } = state;
  const llmWithTools = model.bindTools(tools);
  const result = await llmWithTools.invoke(messages);
  return { messages: [result] };
};
    • Defines an asynchronous function callModel that binds tools to the model and invokes it with the current messages.

  4. Conditionals

    const shouldContinue = (state: typeof MessagesAnnotation.State) => {
  const { messages } = state;
  const lastMessage = messages[messages.length - 1];
  if (
    lastMessage._getType() !== "ai" ||
    !(lastMessage as AIMessage).tool_calls?.length
  ) {
    return END;
  }
  return "tools";
};
    • Defines a function shouldContinue to determine whether the workflow should continue or end based on the last message.

  5. Graph

    const workflow = new StateGraph(MessagesAnnotation)
  .addNode("agent", callModel)
  .addEdge(START, "agent")
  .addNode("tools", toolNode)
  .addEdge("tools", "agent")
  .addConditionalEdges("agent", shouldContinue, ["tools", END]);

export const graph = workflow.compile({
  // Uncomment if running locally
  // checkpointer: new MemorySaver(),
});
graph.name = "graph";
    • Creates a StateGraph with nodes and edges, defining the workflow of the graph.

  6. Draw Graph

    import { saveGraphAsImage } from "drawGraph.js";
await saveGraphAsImage(graph);
    • Imports a function to save the graph as an image and calls it.

  7. Usage

    const agentFinalState = await graph.invoke(
  { messages: [new HumanMessage("what is the current weather in sf")] },
  { configurable: { thread_id: "42" } },
);

console.log(
  agentFinalState.messages[agentFinalState.messages.length - 1].content,
);

const agentNextState = await graph.invoke(
  { messages: [new HumanMessage("what about ny")] },
  { configurable: { thread_id: "42" } },
);

console.log(
  agentNextState.messages[agentNextState.messages.length - 1].content,
);
    • Demonstrates how to invoke the graph with initial messages and log the final state messages.

Running the project

To run the intro project, use the following commands:

yarn install
yarn run build
yarn run start