@ai-agent/storage-memory

In-memory storage adapter for the AI Agent framework. This package provides simple, fast storage for conversation sessions and messages, suitable for development, testing, and simple applications.

Installation

npm install @ai-agent/storage-memory

Usage

Session Management

import { Agent } from '@ai-agent/core';
import { SessionManager } from '@ai-agent/storage-memory';

// Create agent
const agent = new Agent(config);

// Create session manager
const sessionManager = new SessionManager();

// Create a new session
const sessionId = sessionManager.createSession({
  metadata: { userId: '123' }
});

// Process a message
const userMessage = 'Hello, agent!';
sessionManager.addUserMessage(sessionId, userMessage);

const response = await agent.chat(userMessage, {
  sessionId,
  history: sessionManager.getHistory(sessionId)
});

// Store the response
sessionManager.addAssistantMessage(sessionId, response);

Message Storage

import { MessageStore } from '@ai-agent/storage-memory';

// Create message store
const messageStore = new MessageStore();

// Store messages
messageStore.storeUserMessage(sessionId, 'Hello!');
const response = await agent.chat('Hello!', { sessionId });
messageStore.storeAssistantMessage(sessionId, response);

// Query messages
const history = messageStore.getSessionHistory(sessionId);
const recentMessages = messageStore.queryMessages({
  sessionId,
  limit: 10,
  order: 'desc'
});

// Query tool calls
const toolCalls = messageStore.queryToolCalls({
  sessionId,
  toolName: 'search',
  success: true
});

Features

• Fast: All data stored in memory for instant access
• Simple: No database setup required
• Flexible: Rich query API for messages and tool calls
• Type-safe: Full TypeScript support

Limitations

• Not persistent: Data is lost when the process exits
• Memory-bound: Not suitable for large-scale applications
• Single-process: Cannot share data across multiple processes

For production applications, consider:

• @ai-agent/storage-prisma - SQL database support
• @ai-agent/storage-mongodb - MongoDB support
• @ai-agent/storage-redis - Redis support

API Reference

SessionManager

Manages conversation sessions with message history.

Methods

• createSession(options?) - Create a new session
• getSession(sessionId) - Get session by ID
• getOrCreateSession(sessionId, options?) - Get or create session
• hasSession(sessionId) - Check if session exists
• querySessions(options?) - Query sessions with filters
• getHistory(sessionId) - Get conversation history
• addUserMessage(sessionId, content, metadata?) - Add user message
• addAssistantMessage(sessionId, response, metadata?) - Add assistant message
• addSystemMessage(sessionId, content) - Add system message
• closeSession(sessionId) - Mark session as inactive
• deleteSession(sessionId) - Delete session
• clearSession(sessionId) - Clear all messages in session
• updateMetadata(sessionId, metadata) - Update session metadata

MessageStore

Provides persistent storage for messages with rich query capabilities.

Methods

• storeUserMessage(sessionId, content, metadata?) - Store user message
• storeAssistantMessage(sessionId, response, metadata?) - Store assistant message
• storeSystemMessage(sessionId, content) - Store system message
• getMessage(messageId) - Get message by ID
• queryMessages(options?) - Query messages with filters
• getSessionHistory(sessionId) - Get conversation history
• queryToolCalls(options?) - Query tool calls with filters
• getToolCallsForMessage(messageId) - Get tool calls for a message
• deleteMessage(messageId) - Delete a message
• deleteSessionMessages(sessionId) - Delete all messages in session
• getMessageCount(sessionId) - Get message count for session
• getTotalMessageCount() - Get total message count
• getTotalToolCallCount() - Get total tool call count
• clear() - Clear all data

License

MIT