@getzep/zep-cloud 3.0.3 → 3.0.5

package/examples/langgraph/README.md

@@ -1,165 +0,0 @@
-# LangGraph CLI Agent
-
-This is a simple CLI agent built with LangGraph.js that can answer questions and use tools to search for information. It also integrates with Zep for memory persistence.
-
-## Prerequisites
-
-- Node.js 18+ installed
-- OpenAI API key
-- Tavily API key (for search functionality)
-- Zep API key (for memory persistence)
-
-## Setup
-
-1. Install dependencies:
-
-```bash
-npm install
-```
-
-2. Create a `.env` file in this directory with your API keys:
-
-```
-OPENAI_API_KEY=your_openai_api_key_here
-TAVILY_API_KEY=your_tavily_api_key_here
-ZEP_API_KEY=your_zep_api_key_here
-```
-
-## Running the Agent
-
-To start the CLI agent, run:
-
-```bash
-npm start
-```
-
-### Command-Line Options
-
-The agent supports various command-line options through Commander.js:
-
-```bash
-# Get help
-npm start -- --help
-
-# Basic usage with user ID
-npm start -- --userId john_doe
-
-# Using kebab-case format
-npm start -- --user-id john_doe --session-id vacation_planning
-
-# Customize the system message
-npm start -- --system-message "You are a travel assistant helping with vacation planning."
-
-# Enable debug mode
-npm start -- --debug
-```
-
-Available options:
-
-| Option                       | Description                                |
-| ---------------------------- | ------------------------------------------ |
-| `--userId <id>`              | User ID to associate with the conversation |
-| `--user-id <id>`             | Alternative format for user ID             |
-| `--sessionId <id>`           | Session ID for the conversation            |
-| `--session-id <id>`          | Alternative format for session ID          |
-| `--system-message <message>` | Custom system message to use               |
-| `--debug`                    | Enable debug mode with additional logging  |
-| `--help`                     | Display help information                   |
-| `--version`                  | Display version information                |
-
-### Memory Features
-
-Using the user ID and session ID options allows you to:
-
-- **Maintain separate conversation histories** for different users
-- **Continue specific conversations** by using the same session ID
-- **Group related conversations** under the same user ID
-- **Switch between different contexts** by changing the session ID
-
-The agent will automatically create users and sessions in Zep if they don't exist, and will load previous conversation history when available.
-
-## Features
-
-This will launch an interactive CLI where you can chat with the agent. Type your questions and the agent will respond. The agent can:
-
-- Answer general knowledge questions
-- Search the web for information using Tavily
-- Maintain context throughout the conversation
-- Persist conversation history using Zep memory
-- Retrieve past conversations when restarted
-- Associate conversations with specific users
-- Continue conversations across different sessions
-- Use custom system messages
-
-Type `exit` to quit the application.
-
-## How It Works
-
-This agent is built using LangGraph.js, which allows for creating stateful, multi-step AI workflows as graphs. The agent:
-
-1. Takes user input
-2. Processes it through an LLM (OpenAI)
-3. Decides whether to use tools or respond directly
-4. If tools are needed, executes them and returns to the LLM
-5. Returns the final response to the user
-6. Persists all messages to Zep memory (if configured)
-
-The graph maintains state between interactions, allowing for conversational context to be preserved. Additionally, the Zep memory integration provides long-term persistence of conversation history.
-
-## Memory Integration
-
-The agent uses Zep for memory persistence. When a Zep API key is provided, the agent will:
-
-1. Create a new session in Zep (or use an existing one)
-2. Store all user and AI messages in Zep memory
-3. Maintain conversation history across restarts
-4. Load previous messages when restarted
-5. Provide context to the agent about the previous conversation
-
-### User and Session Management
-
-The agent supports specifying a user ID and session ID via command-line arguments:
-
-- **User ID**: Associates conversations with a specific user in Zep
-- **Session ID**: Uses a specific session ID for the conversation
-
-This allows for powerful use cases such as:
-
-- **Multi-user support**: Different users can have their own conversation histories
-- **Topic-based sessions**: Create different sessions for different topics (e.g., "travel_planning", "tech_support")
-- **Conversation continuity**: Resume specific conversations by using the same session ID
-- **Organizational structure**: Group related conversations under the same user ID
-- **A/B testing**: Compare different conversation flows by using different session IDs
-
-### Advanced Memory Features
-
-The Zep memory integration includes several advanced features:
-
-- **User Management**: Creates and manages users associated with conversations
-- **Session Management**: Creates and retrieves sessions for persistent conversations
-- **Message Retrieval**: Loads previous messages when the agent is restarted
-- **Memory Search**: Can search for relevant messages based on semantic similarity
-- **Context Retrieval**: Can retrieve context information about the conversation
-
-This allows the agent to maintain context over long periods and provide more personalized responses based on past interactions.
-
-## Customization
-
-You can modify `agent.ts` to:
-
-- Change the LLM model
-- Add additional tools
-- Adjust the graph structure
-- Customize the CLI interface
-- Configure memory settings
-
-The `zep-memory.ts` file contains the Zep memory integration and can be customized to:
-
-- Change how messages are stored and retrieved
-- Modify user and session management
-- Add additional memory features
-- Customize message conversion between LangChain and Zep formats
-
-## License
-
-MIT

package/examples/langgraph/agent.ts

@@ -1,261 +0,0 @@
-// agent.ts
-import dotenv from "dotenv";
-
-// Load environment variables from .env file
-dotenv.config();
-
-import { TavilySearchResults } from "@langchain/community/tools/tavily_search";
-import { ChatOpenAI } from "@langchain/openai";
-import { HumanMessage, AIMessage, BaseMessage, SystemMessage } from "@langchain/core/messages";
-import { ToolNode } from "@langchain/langgraph/prebuilt";
-import { StateGraph, MessagesAnnotation } from "@langchain/langgraph";
-import { fileURLToPath } from "url";
-import path from "path";
-import { Command } from "commander";
-import { ZepMemory } from "./zep-memory";
-
-// Define the tools for the agent to use
-const tools = [new TavilySearchResults({ maxResults: 3 })];
-const toolNode = new ToolNode(tools);
-
-// Create a model and give it access to the tools
-const model = new ChatOpenAI({
-  model: "gpt-4o-mini",
-  temperature: 0,
-}).bindTools(tools);
-
-// Define the function that determines whether to continue or not
-function shouldContinue({ messages }: typeof MessagesAnnotation.State) {
-  const lastMessage = messages[messages.length - 1] as AIMessage;
-
-  // If the LLM makes a tool call, then we route to the "tools" node
-  if (lastMessage.tool_calls?.length) {
-    return "tools";
-  }
-  // Otherwise, we stop (reply to the user) using the special "__end__" node
-  return "__end__";
-}
-
-// Define the function that calls the model
-async function callModel(state: typeof MessagesAnnotation.State) {
-  const response = await model.invoke(state.messages);
-  return { messages: [...state.messages, response] };
-}
-
-// Define the workflow as a graph
-const workflow = new StateGraph(MessagesAnnotation)
-  .addNode("agent", callModel)
-  .addNode("tools", toolNode)
-  .addEdge("__start__", "agent")
-  .addConditionalEdges("agent", shouldContinue, ["tools", "__end__"])
-  .addEdge("tools", "agent");
-
-// Compile the graph
-export const graph = workflow.compile();
-
-// Parse command line arguments using Commander
-function parseCommandLineArgs() {
-  const program = new Command();
-  
-  program
-    .name("langgraph-agent")
-    .description("LangGraph CLI Agent with Zep memory integration")
-    .version("1.0.0")
-    .option("--userId <id>", "User ID to associate with the conversation")
-    .option("--user-id <id>", "User ID to associate with the conversation (alternative format)")
-    .option("--sessionId <id>", "Session ID for the conversation")
-    .option("--session-id <id>", "Session ID for the conversation (alternative format)")
-    .option("--system-message <message>", "Custom system message to use")
-    .option("--debug", "Enable debug mode with additional logging");
-  
-  program.parse();
-  
-  const options = program.opts();
-  
-  // Handle alternative formats and naming
-  return {
-    userId: options.userId,
-    sessionId: options.sessionId,
-    systemMessage: options.systemMessage || "You are a helpful assistant. Answer the user's questions to the best of your ability.",
-    debug: !!options.debug
-  };
-}
-
-// Check if this file is being run directly
-const isMainModule = () => {
-  if (typeof require !== 'undefined' && require.main === module) {
-    return true;
-  }
-  
-  if (import.meta.url) {
-    try {
-      const currentFilePath = fileURLToPath(import.meta.url);
-      const currentFileName = path.basename(currentFilePath);
-      // Check if this script was run directly with node/tsx
-      return process.argv[1] && process.argv[1].endsWith(currentFileName);
-    } catch (e) {
-      return false;
-    }
-  }
-  
-  return false;
-};
-
-// Parse command line arguments once and store the result
-const args = isMainModule() ? parseCommandLineArgs() : {
-  userId: undefined,
-  sessionId: undefined,
-  systemMessage: "You are a helpful assistant. Answer the user's questions to the best of your ability.",
-  debug: false
-};
-
-// Initialize Zep memory if API key is available
-let zepMemory: ZepMemory | undefined;
-if (process.env.ZEP_API_KEY) {
-  zepMemory = new ZepMemory(process.env.ZEP_API_KEY, args.sessionId, args.userId);
-  
-  if (args.debug) {
-    console.log("Zep memory initialized with session ID:", zepMemory.getSessionId());
-    if (args.userId) {
-      console.log("Using user ID:", args.userId);
-    }
-  }
-}
-
-// CLI interface
-if (isMainModule()) {
-  // This will run when the script is executed directly
-  const runCLI = async () => {
-    const systemMessage = args.systemMessage;
-    
-    console.log("🦜🔗 LangGraph Agent CLI");
-    if (args.userId) {
-      console.log(`User ID: ${args.userId}`);
-    }
-    if (args.sessionId) {
-      console.log(`Session ID: ${args.sessionId}`);
-    }
-    console.log("Type 'exit' to quit the application");
-    console.log("------------------------------");
-    
-    // Create a readline interface
-    const readline = await import("readline");
-    const rl = readline.createInterface({
-      input: process.stdin,
-      output: process.stdout,
-    });
-    
-    // Create a new chat session
-    const config = { configurable: { sessionId: args.sessionId || "cli-session" } };
-    let state = { messages: [] as BaseMessage[] };
-    
-    // Initialize Zep memory if available
-    if (zepMemory) {
-      try {
-        await zepMemory.initialize(args.userId);
-        console.log("Connected to Zep memory service");
-        
-        // Try to load previous messages from Zep memory
-        try {
-          const previousMessages = await zepMemory.getMessages(10);
-          if (previousMessages.length > 0) {
-            state.messages = previousMessages;
-            console.log(`Loaded ${previousMessages.length} messages from previous conversation`);
-          }
-        } catch (error) {
-          console.error("Failed to load previous messages:", error);
-          console.log("Starting a new conversation");
-        }
-      } catch (error) {
-        console.error("Failed to initialize Zep memory:", error);
-        console.log("Continuing without memory persistence");
-        zepMemory = undefined;
-      }
-    }
-    
-    const askQuestion = () => {
-      rl.question("\nYou: ", async (input) => {
-        if (input.toLowerCase() === "exit") {
-          console.log("Goodbye!");
-          rl.close();
-          return;
-        }
-        
-        // Add the user's message to the state
-        const userMessage = new HumanMessage(input);
-        let systemMessageWithContext = systemMessage;
-        
-        // Persist user message to Zep memory if available
-        let context: string | undefined;
-        if (zepMemory) {
-          try {
-            // Add the user message to Zep memory and get the context, if available
-            context = await zepMemory.addMessage(userMessage, true);
-          } catch (error) {
-            console.error("Failed to persist user message to Zep memory:", error);
-          }
-          
-          try {
-            if (args.debug) {
-              console.log(context);
-            }
-            if (context) {
-              systemMessageWithContext = systemMessage + "\n" + context;
-            }
-          } catch (error) {
-            console.error("Failed to get memory with context:", error);
-          }
-        }
-        
-        // Get the last 5 messages from the current state (excluding any previous system messages)
-        const previousMessages = state.messages
-          .filter(msg => !(msg instanceof SystemMessage))
-          .slice(-5);
-        
-        // Create new state with system message at the beginning followed by the history messages
-        state.messages = [
-          new SystemMessage(systemMessageWithContext), 
-          ...previousMessages, 
-          userMessage
-        ];
-        
-        try {
-          // Process the input through the graph
-          const result = await graph.invoke(state, config);
-          
-          // Extract the AI's response using instanceof instead of _getType()
-          const aiMessages = result.messages.filter((msg) => msg instanceof AIMessage);
-          const lastAIMessage = aiMessages[aiMessages.length - 1] as AIMessage;
-          
-          console.log(`\nAI: ${lastAIMessage.content}`);
-          
-          // Persist AI message to Zep memory if available
-          if (zepMemory) {
-            try {
-              await zepMemory.addMessage(lastAIMessage);
-            } catch (error) {
-              console.error("Failed to persist AI message to Zep memory:", error);
-            }
-          }
-          
-          // Update the state for the next interaction
-          // We don't want to completely replace the state, just update the messages
-          // to maintain our structure with the system message at the beginning
-          const resultMessages = result.messages.filter(msg => !(msg instanceof SystemMessage));
-          state = { 
-            messages: resultMessages 
-          };
-        } catch (error) {
-          console.error("Error:", error);
-        }
-        
-        // Ask for the next input
-        askQuestion();
-      });
-    };
-    
-    askQuestion();
-  };
-  
-  runCLI().catch(console.error);
-}