Ai-memory-booster NPM

AI Memory Booster

AI Memory Booster is an AI-powered memory enhancement module that enables long-term AI memory and retrieval of conversations using embeddings and vector databases.

🚀 Features

Memory Storage: Store and retrieve long-term conversation history.
Embeddings Support: Utilize FAISS and ChromaDB for efficient AI memory.
Flexible Deployment: Use AI Memory Booster as an npm module or run it as an API service to make RESTful API calls.
Built-in UI: Manage configurations easily through the integrated UI component.
Extensible & Open-Source: Built for developers to integrate into AI chatbots and automation tools.

📦 Installation

Pre-requisites

Before running AI Memory Booster, ensure that you have Ollama and ChromaDB installed.

Install Ollama

Follow the official installation guide at: https://ollama.ai/

Install ChromaDB

Ensure ChromaDB is installed and running:

pip install chromadb
chroma run --path ./chroma_db

Install AI Memory Booster via npm:

npm install ai-memory-booster

🎯 Usage

AI Memory Booster provides the following methods:

Store Memory

Injects a memory entry into AI Memory Booster.

await AI_Memory.storeMemory("Today is Wednesday and it is raining", "Today is Wednesday, how's the weather?", "It is raining");

Retrieve Memory

Fetches relevant stored memories based on a user's message.

const memory = await AI_Memory.retrieveMemory("How's the weather today?");
console.log(memory);

Response Example:

{
  "id": "99c326e8-e234-4049-8661-3d9427944071",
  "distance": 0.1, // The distance between the query and result (smaller is better)
  "summary": "Today is Wednesday and it is raining",
  "userMessage": "Today is Wednesday, how's the weather?",
  "aiMessage": "It is raining",
  "timestamp": 1740999657717
}

Forget Memory

Deletes a memory entry by ID.

const success = await AI_Memory.forget("99c326e8-e234-4049-8661-3d9427944071");
console.log(success); // true if deleted, false if ID is null

Forget All Memories

Deletes all stored memory entries.

const success = await AI_Memory.forgetAll();
console.log(success); // true if deleted, false if no memories exist

Get LLM Specification

Retrieves details about the current LLM module.

const llmSpec = await AI_Memory.getLlmSpec();
console.log(llmSpec);

Chat with AI Memory Booster

Interacts with the AI while using stored memory (For chat).

const response = await AI_Memory.chat("How's the weather today?");
console.log(response); // "It is raining today!"

Generate content with AI Memory Booster

Interacts with the AI while using stored memory (For generating text content).

const response = await AI_Memory.generate("How's the weather today?");
console.log(response); // "It is raining."

🛠 Configuration

AI Memory Booster allows full customization through config.json. Users can modify AI model settings, server configurations, memory management, and ChromaDB parameters. The configuration can be updated in two ways:

Directly modifying config.json
Using AI Memory Booster UI for easy configuration updates

Example `config.json`:

{
  "aiModel": "llama3.2", // The primary AI model (must be downloaded first)
  "learnFromChat": true, // Determines if AI should learn from conversations
  "host": "localhost", // The host address when running AI Memory Booster as an API service
  "port": 4000, // The port for the AI Memory Booster API service
  "baseKeepAlive": 3000, // Duration (ms) the LLM module stays active after each call
  "extendedKeepAlive": 10000, // Extended duration (ms) the LLM remains active if no other requests arrive
  "similarityResultCount": 3, // Number of similar records retrieved from the database
  "categorySureThreshold": 49, // Threshold for AI to confidently classify a response
  "summaryCharacterLimit": 256, // Maximum character length for conversation summaries
  "dimension": 768, // Dimensionality of vector embeddings
  "similarityThreshold": 0.7, // Threshold for similarity-based searches
  "consolidateConversationThreshold": 256, // Threshold for summarizing conversations
  "chromaDBHost": "http://localhost:8000", // ChromaDB service URL (host and port)
  "tenant": "default_tenant", // ChromaDB tenant name
  "collection": "ai_memory_booster", // ChromaDB collection name used by AI Memory Booster
  "rolePrompt": "You are a personal assistant. Below is the conversation history to understand the context. The conversation history is enclosed between 'Conversation History Start:' and 'Conversation History End.' 'AI' represents you, and 'User' represents the person currently talking to you. When the user says 'I', 'mine', or 'my', it refers to the user, not you ('AI'). Do not fabricate responses.", // Prompt for how AI should respond based on past memory
  "debug": false, // If turn on the debug message
  "archive": false, // If turn on the memory archive mode
}

ConfigManager Usage:

Users can dynamically get and set configurations via configManager:

const port = AI_Memory.configManager.getPort();
console.log(`Running on port: ${port}`);

AI_Memory.configManager.setPort(8080);
console.log(`Updated port: ${AI_Memory.configManager.getPort()}`);

🔧 Development

Clone and install dependencies:

git clone https://github.com/aotol/ai-memory-booster.git
cd ai-memory-booster
npm install

To start AI Memory Booster as a standalone API service:

npx ai-memory-booster start

To launch AI Memory Booster UI, go to:

cd ai-memory-booster/web/ai-memory-ui/
npm run dev

Make sure you have set the environment variable: AI_MEMORY_BOOSTER_API_URL to identify where AI Memory Booster API service is running. e.g.:

export AI_MEMORY_BOOSTER_API_URL="http://localhost:4000"

Alternatively, create a .env.local file under ai-memory-ui directory and set the content to:

AI_MEMORY_BOOSTER_API_URL=http://localhost:4000

📜 License

MIT License - Free for personal and research use. AI Memory Booster is open-source under the MIT License. If you require enterprise support or commercial licensing, please contact us.