LangGraph Retrieval Agent¶

We can implement Retrieval Agents in LangGraph.

Setup¶

Load env vars¶

Add a .env variable in the root of the ./examples folder with your variables.

In [1]:

Copied!

// import dotenv from 'dotenv';

// dotenv.config();
// import dotenv from 'dotenv';

// dotenv.config();

Install dependencies¶

npm install cheerio zod zod-to-json-schema langchain @langchain/openai @langchain/core @langchain/community @langchain/textsplitters

Retriever¶

In [2]:

Copied!





import { CheerioWebBaseLoader } from "@langchain/community/document_loaders/web/cheerio";
import { RecursiveCharacterTextSplitter } from "@langchain/textsplitters";
import { MemoryVectorStore } from "langchain/vectorstores/memory";
import { OpenAIEmbeddings } from "@langchain/openai";

const urls = [
  "https://lilianweng.github.io/posts/2023-06-23-agent/",
  "https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/",
  "https://lilianweng.github.io/posts/2023-10-25-adv-attack-llm/",
];

const docs = await Promise.all(
  urls.map((url) => new CheerioWebBaseLoader(url).load()),
);
const docsList = docs.flat();

const textSplitter = new RecursiveCharacterTextSplitter({
  chunkSize: 500,
  chunkOverlap: 50,
});
const docSplits = await textSplitter.splitDocuments(docsList);

// Add to vectorDB
const vectorStore = await MemoryVectorStore.fromDocuments(
  docSplits,
  new OpenAIEmbeddings(),
);

const retriever = vectorStore.asRetriever();
import { CheerioWebBaseLoader } from "@langchain/community/document_loaders/web/cheerio";
import { RecursiveCharacterTextSplitter } from "@langchain/textsplitters";
import { MemoryVectorStore } from "langchain/vectorstores/memory";
import { OpenAIEmbeddings } from "@langchain/openai";

const urls = [
  "https://lilianweng.github.io/posts/2023-06-23-agent/",
  "https://lilianweng.github.io/posts/2023-03-15-prompt-engineering/",
  "https://lilianweng.github.io/posts/2023-10-25-adv-attack-llm/",
];

const docs = await Promise.all(
  urls.map((url) => new CheerioWebBaseLoader(url).load()),
);
const docsList = docs.flat();

const textSplitter = new RecursiveCharacterTextSplitter({
  chunkSize: 500,
  chunkOverlap: 50,
});
const docSplits = await textSplitter.splitDocuments(docsList);

// Add to vectorDB
const vectorStore = await MemoryVectorStore.fromDocuments(
  docSplits,
  new OpenAIEmbeddings(),
);

const retriever = vectorStore.asRetriever();

Agent state¶

We will define a graph.

You may pass a custom state object to the graph, or use a simple list of messages.

Our state will be a list of messages.

Each node in our graph will append to it.

In [3]:

Copied!





import { Annotation } from "@langchain/langgraph";
import { BaseMessage } from "@langchain/core/messages";

const GraphState = Annotation.Root({
  messages: Annotation<BaseMessage[]>({
    reducer: (x, y) => x.concat(y),
    default: () => [],
  })
})
import { Annotation } from "@langchain/langgraph";
import { BaseMessage } from "@langchain/core/messages";

const GraphState = Annotation.Root({
  messages: Annotation({
    reducer: (x, y) => x.concat(y),
    default: () => [],
  })
})

In [4]:

Copied!





import { createRetrieverTool } from "langchain/tools/retriever";
import { ToolNode } from "@langchain/langgraph/prebuilt";

const tool = createRetrieverTool(
  retriever,
  {
    name: "retrieve_blog_posts",
    description:
      "Search and return information about Lilian Weng blog posts on LLM agents, prompt engineering, and adversarial attacks on LLMs.",
  },
);
const tools = [tool];

const toolNode = new ToolNode<typeof GraphState.State>(tools);
import { createRetrieverTool } from "langchain/tools/retriever";
import { ToolNode } from "@langchain/langgraph/prebuilt";

const tool = createRetrieverTool(
  retriever,
  {
    name: "retrieve_blog_posts",
    description:
      "Search and return information about Lilian Weng blog posts on LLM agents, prompt engineering, and adversarial attacks on LLMs.",
  },
);
const tools = [tool];

const toolNode = new ToolNode(tools);

Nodes and Edges¶

Each node will -

1/ Either be a function or a runnable.

2/ Modify the state.

The edges choose which node to call next.

We can lay out an agentic RAG graph like this:

Edges¶

In [5]:

Copied!





import { END } from "@langchain/langgraph";
import { pull } from "langchain/hub";
import { z } from "zod";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { ChatOpenAI } from "@langchain/openai";
import { AIMessage, BaseMessage } from "@langchain/core/messages";

/**
 * Decides whether the agent should retrieve more information or end the process.
 * This function checks the last message in the state for a function call. If a tool call is
 * present, the process continues to retrieve information. Otherwise, it ends the process.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {string} - A decision to either "continue" the retrieval process or "end" it.
 */
function shouldRetrieve(state: typeof GraphState.State): string {
  const { messages } = state;
  console.log("---DECIDE TO RETRIEVE---");
  const lastMessage = messages[messages.length - 1];

  if ("tool_calls" in lastMessage && Array.isArray(lastMessage.tool_calls) && lastMessage.tool_calls.length) {
    console.log("---DECISION: RETRIEVE---");
    return "retrieve";
  }
  // If there are no tool calls then we finish.
  return END;
}

/**
 * Determines whether the Agent should continue based on the relevance of retrieved documents.
 * This function checks if the last message in the conversation is of type FunctionMessage, indicating
 * that document retrieval has been performed. It then evaluates the relevance of these documents to the user's
 * initial question using a predefined model and output parser. If the documents are relevant, the conversation
 * is considered complete. Otherwise, the retrieval process is continued.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise<Partial<typeof GraphState.State>>} - The updated state with the new message added to the list of messages.
 */
async function gradeDocuments(state: typeof GraphState.State): Promise<Partial<typeof GraphState.State>> {
  console.log("---GET RELEVANCE---");

  const { messages } = state;
  const tool = {
    name: "give_relevance_score",
    description: "Give a relevance score to the retrieved documents.",
    schema: z.object({
      binaryScore: z.string().describe("Relevance score 'yes' or 'no'"),
    })
  }

  const prompt = ChatPromptTemplate.fromTemplate(
    `You are a grader assessing relevance of retrieved docs to a user question.
  Here are the retrieved docs:
  \n ------- \n
  {context} 
  \n ------- \n
  Here is the user question: {question}
  If the content of the docs are relevant to the users question, score them as relevant.
  Give a binary score 'yes' or 'no' score to indicate whether the docs are relevant to the question.
  Yes: The docs are relevant to the question.
  No: The docs are not relevant to the question.`,
  );

  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
  }).bindTools([tool], {
    tool_choice: tool.name,
  });

  const chain = prompt.pipe(model);

  const lastMessage = messages[messages.length - 1];

  const score = await chain.invoke({
    question: messages[0].content as string,
    context: lastMessage.content as string,
  });

  return {
    messages: [score]
  };
}

/**
 * Check the relevance of the previous LLM tool call.
 *
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {string} - A directive to either "yes" or "no" based on the relevance of the documents.
 */
function checkRelevance(state: typeof GraphState.State): string {
  console.log("---CHECK RELEVANCE---");

  const { messages } = state;
  const lastMessage = messages[messages.length - 1];
  if (!("tool_calls" in lastMessage)) {
    throw new Error("The 'checkRelevance' node requires the most recent message to contain tool calls.")
  }
  const toolCalls = (lastMessage as AIMessage).tool_calls;
  if (!toolCalls || !toolCalls.length) {
    throw new Error("Last message was not a function message");
  }

  if (toolCalls[0].args.binaryScore === "yes") {
    console.log("---DECISION: DOCS RELEVANT---");
    return "yes";
  }
  console.log("---DECISION: DOCS NOT RELEVANT---");
  return "no";
}

// Nodes

/**
 * Invokes the agent model to generate a response based on the current state.
 * This function calls the agent model to generate a response to the current conversation state.
 * The response is added to the state's messages.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise<Partial<typeof GraphState.State>>} - The updated state with the new message added to the list of messages.
 */
async function agent(state: typeof GraphState.State): Promise<Partial<typeof GraphState.State>> {
  console.log("---CALL AGENT---");
  
  const { messages } = state;
  // Find the AIMessage which contains the `give_relevance_score` tool call,
  // and remove it if it exists. This is because the agent does not need to know
  // the relevance score.
  const filteredMessages = messages.filter((message) => {
    if ("tool_calls" in message && Array.isArray(message.tool_calls) && message.tool_calls.length > 0) {
      return message.tool_calls[0].name !== "give_relevance_score";
    }
    return true;
  });

  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  }).bindTools(tools);

  const response = await model.invoke(filteredMessages);
  return {
    messages: [response],
  };
}

/**
 * Transform the query to produce a better question.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise<Partial<typeof GraphState.State>>} - The updated state with the new message added to the list of messages.
 */
async function rewrite(state: typeof GraphState.State): Promise<Partial<typeof GraphState.State>> {
  console.log("---TRANSFORM QUERY---");

  const { messages } = state;
  const question = messages[0].content as string;
  const prompt = ChatPromptTemplate.fromTemplate(
    `Look at the input and try to reason about the underlying semantic intent / meaning. \n 
Here is the initial question:
\n ------- \n
{question} 
\n ------- \n
Formulate an improved question:`,
  );

  // Grader
  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  });
  const response = await prompt.pipe(model).invoke({ question });
  return {
    messages: [response],
  };
}

/**
 * Generate answer
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise<Partial<typeof GraphState.State>>} - The updated state with the new message added to the list of messages.
 */
async function generate(state: typeof GraphState.State): Promise<Partial<typeof GraphState.State>> {
  console.log("---GENERATE---");
  
  const { messages } = state;
  const question = messages[0].content as string;
  // Extract the most recent ToolMessage
  const lastToolMessage = messages.slice().reverse().find((msg) => msg._getType() === "tool");
  if (!lastToolMessage) {
    throw new Error("No tool message found in the conversation history");
  }

  const docs = lastToolMessage.content as string;

  const prompt = await pull<ChatPromptTemplate>("rlm/rag-prompt");

  const llm = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  });

  const ragChain = prompt.pipe(llm);

  const response = await ragChain.invoke({
    context: docs,
    question,
  });

  return {
    messages: [response],
  };
}
import { END } from "@langchain/langgraph";
import { pull } from "langchain/hub";
import { z } from "zod";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { ChatOpenAI } from "@langchain/openai";
import { AIMessage, BaseMessage } from "@langchain/core/messages";

/**
 * Decides whether the agent should retrieve more information or end the process.
 * This function checks the last message in the state for a function call. If a tool call is
 * present, the process continues to retrieve information. Otherwise, it ends the process.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {string} - A decision to either "continue" the retrieval process or "end" it.
 */
function shouldRetrieve(state: typeof GraphState.State): string {
  const { messages } = state;
  console.log("---DECIDE TO RETRIEVE---");
  const lastMessage = messages[messages.length - 1];

  if ("tool_calls" in lastMessage && Array.isArray(lastMessage.tool_calls) && lastMessage.tool_calls.length) {
    console.log("---DECISION: RETRIEVE---");
    return "retrieve";
  }
  // If there are no tool calls then we finish.
  return END;
}

/**
 * Determines whether the Agent should continue based on the relevance of retrieved documents.
 * This function checks if the last message in the conversation is of type FunctionMessage, indicating
 * that document retrieval has been performed. It then evaluates the relevance of these documents to the user's
 * initial question using a predefined model and output parser. If the documents are relevant, the conversation
 * is considered complete. Otherwise, the retrieval process is continued.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise>} - The updated state with the new message added to the list of messages.
 */
async function gradeDocuments(state: typeof GraphState.State): Promise> {
  console.log("---GET RELEVANCE---");

  const { messages } = state;
  const tool = {
    name: "give_relevance_score",
    description: "Give a relevance score to the retrieved documents.",
    schema: z.object({
      binaryScore: z.string().describe("Relevance score 'yes' or 'no'"),
    })
  }

  const prompt = ChatPromptTemplate.fromTemplate(
    `You are a grader assessing relevance of retrieved docs to a user question.
  Here are the retrieved docs:
  \n ------- \n
  {context} 
  \n ------- \n
  Here is the user question: {question}
  If the content of the docs are relevant to the users question, score them as relevant.
  Give a binary score 'yes' or 'no' score to indicate whether the docs are relevant to the question.
  Yes: The docs are relevant to the question.
  No: The docs are not relevant to the question.`,
  );

  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
  }).bindTools([tool], {
    tool_choice: tool.name,
  });

  const chain = prompt.pipe(model);

  const lastMessage = messages[messages.length - 1];

  const score = await chain.invoke({
    question: messages[0].content as string,
    context: lastMessage.content as string,
  });

  return {
    messages: [score]
  };
}

/**
 * Check the relevance of the previous LLM tool call.
 *
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {string} - A directive to either "yes" or "no" based on the relevance of the documents.
 */
function checkRelevance(state: typeof GraphState.State): string {
  console.log("---CHECK RELEVANCE---");

  const { messages } = state;
  const lastMessage = messages[messages.length - 1];
  if (!("tool_calls" in lastMessage)) {
    throw new Error("The 'checkRelevance' node requires the most recent message to contain tool calls.")
  }
  const toolCalls = (lastMessage as AIMessage).tool_calls;
  if (!toolCalls || !toolCalls.length) {
    throw new Error("Last message was not a function message");
  }

  if (toolCalls[0].args.binaryScore === "yes") {
    console.log("---DECISION: DOCS RELEVANT---");
    return "yes";
  }
  console.log("---DECISION: DOCS NOT RELEVANT---");
  return "no";
}

// Nodes

/**
 * Invokes the agent model to generate a response based on the current state.
 * This function calls the agent model to generate a response to the current conversation state.
 * The response is added to the state's messages.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise>} - The updated state with the new message added to the list of messages.
 */
async function agent(state: typeof GraphState.State): Promise> {
  console.log("---CALL AGENT---");
  
  const { messages } = state;
  // Find the AIMessage which contains the `give_relevance_score` tool call,
  // and remove it if it exists. This is because the agent does not need to know
  // the relevance score.
  const filteredMessages = messages.filter((message) => {
    if ("tool_calls" in message && Array.isArray(message.tool_calls) && message.tool_calls.length > 0) {
      return message.tool_calls[0].name !== "give_relevance_score";
    }
    return true;
  });

  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  }).bindTools(tools);

  const response = await model.invoke(filteredMessages);
  return {
    messages: [response],
  };
}

/**
 * Transform the query to produce a better question.
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise>} - The updated state with the new message added to the list of messages.
 */
async function rewrite(state: typeof GraphState.State): Promise> {
  console.log("---TRANSFORM QUERY---");

  const { messages } = state;
  const question = messages[0].content as string;
  const prompt = ChatPromptTemplate.fromTemplate(
    `Look at the input and try to reason about the underlying semantic intent / meaning. \n 
Here is the initial question:
\n ------- \n
{question} 
\n ------- \n
Formulate an improved question:`,
  );

  // Grader
  const model = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  });
  const response = await prompt.pipe(model).invoke({ question });
  return {
    messages: [response],
  };
}

/**
 * Generate answer
 * @param {typeof GraphState.State} state - The current state of the agent, including all messages.
 * @returns {Promise>} - The updated state with the new message added to the list of messages.
 */
async function generate(state: typeof GraphState.State): Promise> {
  console.log("---GENERATE---");
  
  const { messages } = state;
  const question = messages[0].content as string;
  // Extract the most recent ToolMessage
  const lastToolMessage = messages.slice().reverse().find((msg) => msg._getType() === "tool");
  if (!lastToolMessage) {
    throw new Error("No tool message found in the conversation history");
  }

  const docs = lastToolMessage.content as string;

  const prompt = await pull("rlm/rag-prompt");

  const llm = new ChatOpenAI({
    model: "gpt-4o",
    temperature: 0,
    streaming: true,
  });

  const ragChain = prompt.pipe(llm);

  const response = await ragChain.invoke({
    context: docs,
    question,
  });

  return {
    messages: [response],
  };
}

Graph¶

Start with an agent, callModel
Agent make a decision to call a function
If so, then action to call tool (retriever)
Then call agent with the tool output added to messages (state)

In [6]:

Copied!





import { StateGraph } from "@langchain/langgraph";

// Define the graph
const workflow = new StateGraph(GraphState)
  // Define the nodes which we'll cycle between.
  .addNode("agent", agent)
  .addNode("retrieve", toolNode)
  .addNode("gradeDocuments", gradeDocuments)
  .addNode("rewrite", rewrite)
  .addNode("generate", generate);
import { StateGraph } from "@langchain/langgraph";

// Define the graph
const workflow = new StateGraph(GraphState)
  // Define the nodes which we'll cycle between.
  .addNode("agent", agent)
  .addNode("retrieve", toolNode)
  .addNode("gradeDocuments", gradeDocuments)
  .addNode("rewrite", rewrite)
  .addNode("generate", generate);

In [7]:

Copied!





import { START } from "@langchain/langgraph";

// Call agent node to decide to retrieve or not
workflow.addEdge(START, "agent");

// Decide whether to retrieve
workflow.addConditionalEdges(
  "agent",
  // Assess agent decision
  shouldRetrieve,
);

workflow.addEdge("retrieve", "gradeDocuments");

// Edges taken after the `action` node is called.
workflow.addConditionalEdges(
  "gradeDocuments",
  // Assess agent decision
  checkRelevance,
  {
    // Call tool node
    yes: "generate",
    no: "rewrite", // placeholder
  },
);

workflow.addEdge("generate", END);
workflow.addEdge("rewrite", "agent");

// Compile
const app = workflow.compile();
import { START } from "@langchain/langgraph";

// Call agent node to decide to retrieve or not
workflow.addEdge(START, "agent");

// Decide whether to retrieve
workflow.addConditionalEdges(
  "agent",
  // Assess agent decision
  shouldRetrieve,
);

workflow.addEdge("retrieve", "gradeDocuments");

// Edges taken after the `action` node is called.
workflow.addConditionalEdges(
  "gradeDocuments",
  // Assess agent decision
  checkRelevance,
  {
    // Call tool node
    yes: "generate",
    no: "rewrite", // placeholder
  },
);

workflow.addEdge("generate", END);
workflow.addEdge("rewrite", "agent");

// Compile
const app = workflow.compile();

In [8]:

Copied!





import { HumanMessage } from "@langchain/core/messages";

const inputs = {
  messages: [
    new HumanMessage(
      "What are the types of agent memory based on Lilian Weng's blog post?",
    ),
  ],
};
let finalState;
for await (const output of await app.stream(inputs)) {
  for (const [key, value] of Object.entries(output)) {
    const lastMsg = output[key].messages[output[key].messages.length - 1];
    console.log(`Output from node: '${key}'`);
    console.dir({
      type: lastMsg._getType(),
      content: lastMsg.content,
      tool_calls: lastMsg.tool_calls,
    }, { depth: null });
    console.log("---\n");
    finalState = value;
  }
}

console.log(JSON.stringify(finalState, null, 2));
import { HumanMessage } from "@langchain/core/messages";

const inputs = {
  messages: [
    new HumanMessage(
      "What are the types of agent memory based on Lilian Weng's blog post?",
    ),
  ],
};
let finalState;
for await (const output of await app.stream(inputs)) {
  for (const [key, value] of Object.entries(output)) {
    const lastMsg = output[key].messages[output[key].messages.length - 1];
    console.log(`Output from node: '${key}'`);
    console.dir({
      type: lastMsg._getType(),
      content: lastMsg.content,
      tool_calls: lastMsg.tool_calls,
    }, { depth: null });
    console.log("---\n");
    finalState = value;
  }
}

console.log(JSON.stringify(finalState, null, 2));

---CALL AGENT---
---DECIDE TO RETRIEVE---
---DECISION: RETRIEVE---
Output from node: 'agent'
{
  type: 'ai',
  content: '',
  tool_calls: [
    {
      name: 'retrieve_blog_posts',
      args: { query: 'types of agent memory' },
      id: 'call_adLYkV7T2ry1EZFboT0jPuwn',
      type: 'tool_call'
    }
  ]
}
---

Output from node: 'retrieve'
{
  type: 'tool',
  content: 'Agent System Overview\n' +
    '                \n' +
    '                    Component One: Planning\n' +
    '                        \n' +
    '                \n' +
    '                    Task Decomposition\n' +
    '                \n' +
    '                    Self-Reflection\n' +
    '                \n' +
    '                \n' +
    '                    Component Two: Memory\n' +
    '                        \n' +
    '                \n' +
    '                    Types of Memory\n' +
    '                \n' +
    '                    Maximum Inner Product Search (MIPS)\n' +
    '\n' +
    'Memory stream: is a long-term memory module (external database) that records a comprehensive list of agents’ experience in natural language.\n' +
    '\n' +
    'Each element is an observation, an event directly provided by the agent.\n' +
    '- Inter-agent communication can trigger new natural language statements.\n' +
    '\n' +
    '\n' +
    'Retrieval model: surfaces the context to inform the agent’s behavior, according to relevance, recency and importance.\n' +
    '\n' +
    'Planning\n' +
    '\n' +
    'Subgoal and decomposition: The agent breaks down large tasks into smaller, manageable subgoals, enabling efficient handling of complex tasks.\n' +
    'Reflection and refinement: The agent can do self-criticism and self-reflection over past actions, learn from mistakes and refine them for future steps, thereby improving the quality of final results.\n' +
    '\n' +
    '\n' +
    'Memory\n' +
    '\n' +
    'The design of generative agents combines LLM with memory, planning and reflection mechanisms to enable agents to behave conditioned on past experience, as well as to interact with other agents.',
  tool_calls: undefined
}
---

---GET RELEVANCE---
---CHECK RELEVANCE---
---DECISION: DOCS NOT RELEVANT---
Output from node: 'gradeDocuments'
{
  type: 'ai',
  content: '',
  tool_calls: [
    {
      name: 'give_relevance_score',
      args: { binaryScore: 'no' },
      type: 'tool_call',
      id: 'call_AGE7gORVFubExfJWcjb0C2nV'
    }
  ]
}
---

---TRANSFORM QUERY---
Output from node: 'rewrite'
{
  type: 'ai',
  content: "What are the different types of agent memory described in Lilian Weng's blog post?",
  tool_calls: []
}
---

---CALL AGENT---
---DECIDE TO RETRIEVE---
Output from node: 'agent'
{
  type: 'ai',
  content: "Lilian Weng's blog post describes the following types of agent memory:\n" +
    '\n' +
    '1. **Memory Stream**:\n' +
    '   - This is a long-term memory module (external database) that records a comprehensive list of agents’ experiences in natural language.\n' +
    '   - Each element in the memory stream is an observation or an event directly provided by the agent.\n' +
    '   - Inter-agent communication can trigger new natural language statements to be added to the memory stream.\n' +
    '\n' +
    '2. **Retrieval Model**:\n' +
    '   - This model surfaces the context to inform the agent’s behavior based on relevance, recency, and importance.\n' +
    '\n' +
    'These memory types are part of a broader design that combines generative agents with memory, planning, and reflection mechanisms to enable agents to behave based on past experiences and interact with other agents.',
  tool_calls: []
}
---

{
  "messages": [
    {
      "lc": 1,
      "type": "constructor",
      "id": [
        "langchain_core",
        "messages",
        "AIMessageChunk"
      ],
      "kwargs": {
        "content": "Lilian Weng's blog post describes the following types of agent memory:\n\n1. **Memory Stream**:\n   - This is a long-term memory module (external database) that records a comprehensive list of agents’ experiences in natural language.\n   - Each element in the memory stream is an observation or an event directly provided by the agent.\n   - Inter-agent communication can trigger new natural language statements to be added to the memory stream.\n\n2. **Retrieval Model**:\n   - This model surfaces the context to inform the agent’s behavior based on relevance, recency, and importance.\n\nThese memory types are part of a broader design that combines generative agents with memory, planning, and reflection mechanisms to enable agents to behave based on past experiences and interact with other agents.",
        "additional_kwargs": {},
        "response_metadata": {
          "estimatedTokenUsage": {
            "promptTokens": 280,
            "completionTokens": 155,
            "totalTokens": 435
          },
          "prompt": 0,
          "completion": 0,
          "finish_reason": "stop",
          "system_fingerprint": "fp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3bfp_3cd8b62c3b"
        },
        "tool_call_chunks": [],
        "id": "chatcmpl-9zAaVQGmTLiCaFvtbxUK60qMFsSmU",
        "usage_metadata": {
          "input_tokens": 363,
          "output_tokens": 156,
          "total_tokens": 519
        },
        "tool_calls": [],
        "invalid_tool_calls": []
      }
    }
  ]
}