Class ZepMemory

Class used to manage the memory of a chat session, including loading and saving the chat history, and clearing the memory when needed. It uses the ZepClient to interact with the Zep service for managing the chat session's memory.

The class provides options for handling different LLM requirements:

• Use separateMessages=true (default) for models that fully support system messages
• Use separateMessages=false for models like Claude that have limitations with system messages

Example

const sessionId = randomUUID();
const zepURL = "http://your-zep-url";

// Initialize ZepMemory with session ID, base URL, and API key
const memory = new ZepMemory({
  sessionId,
  baseURL: zepURL,
  apiKey: "change_this_key",
  // Set to false for models like Claude that have limitations with system messages
  // Defaults to true for backward compatibility
  separateMessages: false,
});

// Create a ChatOpenAI model instance with specific parameters
const model = new ChatOpenAI({
  model: "gpt-3.5-turbo",
  temperature: 0,
});

// Create a ConversationChain with the model and memory
const chain = new ConversationChain({ llm: model, memory });

// Example of calling the chain with an input
const res1 = await chain.call({ input: "Hi! I'm Jim." });
console.log({ res1 });

// Follow-up call to the chain to demonstrate memory usage
const res2 = await chain.call({ input: "What did I just say my name was?" });
console.log({ res2 });

// Output the session ID and the current state of memory
console.log("Session ID: ", sessionId);
console.log("Memory: ", await memory.loadMemoryVariables({}));

Copy