@aigne/example-afs-memory

Memory Example

This example demonstrates how to create a chatbot with long-term memory using the AIGNE Framework. The chatbot can remember user information across conversations and recall details shared in previous interactions.

The example uses two powerful AFS (Agentic File System) modules:

• AFSHistory: Automatically records all conversations for context
• UserProfileMemory: Intelligently extracts and stores user information (name, interests, location, etc.)

Agentic File System (AFS) is a virtual file system abstraction that provides AI agents with unified access to various storage backends. For comprehensive documentation, see AFS Documentation.

Prerequisites

• Node.js (>=20.0) and npm installed on your machine
• An OpenAI API key for interacting with OpenAI's services
• Optional dependencies (if running the example from source code):
  • Pnpm for package management
  • Bun for running unit tests & examples

Quick Start (No Installation Required)

export OPENAI_API_KEY=YOUR_OPENAI_API_KEY # Set your OpenAI API key

# First conversation - introduce yourself
npx -y @aigne/example-afs-memory --input "I'm Bob, and I like blue color"
# Response: Nice to meet you, Bob — I've saved that your favorite color is blue...

# Second conversation - the bot remembers you!
npx -y @aigne/example-afs-memory --input "Tell me all info about me you known"
# Response: Here's what I currently have stored about you:
#   * Name: Bob
#   * Interests / favorite color: blue

# Run in interactive chat mode
npx -y @aigne/example-afs-memory --chat

Installation

Clone the Repository

git clone https://github.com/AIGNE-io/aigne-framework

Install Dependencies

cd aigne-framework/examples/afs-memory

pnpm install

Setup Environment Variables

Setup your OpenAI API key in the .env.local file:

OPENAI_API_KEY="" # Set your OpenAI API key here

Using Different Models

You can use different AI models by setting the MODEL environment variable along with the corresponding API key. The framework supports multiple providers:

• OpenAI: MODEL="openai:gpt-4.1" with OPENAI_API_KEY
• Anthropic: MODEL="anthropic:claude-3-7-sonnet-latest" with ANTHROPIC_API_KEY
• Google Gemini: MODEL="gemini:gemini-2.0-flash" with GEMINI_API_KEY
• AWS Bedrock: MODEL="bedrock:us.amazon.nova-premier-v1:0" with AWS credentials
• DeepSeek: MODEL="deepseek:deepseek-chat" with DEEPSEEK_API_KEY
• OpenRouter: MODEL="openrouter:openai/gpt-4o" with OPEN_ROUTER_API_KEY
• xAI: MODEL="xai:grok-2-latest" with XAI_API_KEY
• Ollama: MODEL="ollama:llama3.2" with OLLAMA_DEFAULT_BASE_URL

For detailed configuration examples, please refer to the .env.local.example file in this directory.

Run the Example

# Run in interactive mode
pnpm start

# Run with a single message
pnpm start --input "I'm Bob, and I like blue color"

Example Conversation Flow

# First conversation - introduce yourself
$ pnpm start --input "I'm Bob, and I like blue color"
Response: Nice to meet you, Bob — I've saved that your favorite color is blue.
If you'd like, I can remember more about you (location, hobbies, projects, etc.)...

# Second conversation - the bot remembers you!
$ pnpm start --input "Tell me all info about me you known"
Response: Here's what I currently have stored about you:
  * Name: Bob
  * Interests / favorite color: blue

Would you like to add or update anything (location, hobbies, projects, family, etc.)?

How Memory Works

This example uses two complementary memory modules that work together to provide intelligent, personalized conversations:

1. AFSHistory Module - Conversation Context

Purpose: Records every conversation turn to provide recent context.

How it works:

• Automatically saves each user message and AI response pair
• Stores conversations with timestamps and unique IDs
• Enables the AI to reference recent conversations

Example:

# First conversation
$ pnpm start --input "I'm Bob, and I like blue color"
Response: Nice to meet you, Bob — I've saved that your favorite color is blue...

# The conversation is automatically saved to history.sqlite3

2. UserProfileMemory Module - User Information Extraction

Purpose: Intelligently extracts and stores structured user information from conversations.

How it works:

• Listens to conversation history events
• Analyzes each conversation using AI to identify user information
• Extracts relevant details (name, interests, location, family, projects, etc.)
• Stores in a structured JSON profile in user_profile.sqlite3
• Updates incrementally using JSON Patch operations

What it remembers:

• Name and personal details
• Location (country, city, address)
• Interests and hobbies
• Family members and relationships
• Projects and work
• Languages spoken
• Birthday and other personal info

Example:

# After Bob introduces himself, the profile is automatically created:
{
  "name": [{ "name": "Bob" }],
  "interests": [{ "content": "blue color" }]
}

# In the next conversation, the bot can recall this information:
$ pnpm start --input "Tell me all info about me you known"
Response: Here's what I currently have stored about you:
  * Name: Bob
  * Interests / favorite color: blue

3. Memory Injection - How the AI Uses Memory

When you send a message, the system automatically:

Step 1: Inject User Profile into System Prompt

You are a friendly chatbot

<related-memories>
User Profile Memory: This contains structured information about the user...

- name:
    - name: Bob
  interests:
    - content: blue color
</related-memories>

Step 2: Add Recent Conversation History

[
  {
    "role": "system",
    "content": "You are a friendly chatbot... [profile injected here]"
  },
  {
    "role": "user",
    "content": "I'm Bob and I like blue color"
  },
  {
    "role": "assistant",
    "content": "Nice to meet you, Bob..."
  },
  {
    "role": "user",
    "content": "Tell me all info about me you known"
  }
]

Step 3: AI Generates Personalized Response

The AI can now:

• Address the user by name
• Reference their interests
• Provide personalized recommendations
• Maintain context across sessions

Key Design Benefits

• Automatic: No manual profile management needed
• Intelligent: AI determines what's important to remember
• Incremental: Profile updates gradually over time
• Persistent: Memory survives across multiple conversations
• Structured: Information is organized in a consistent format
• Privacy-Aware: All data stored locally in SQLite

Related Examples

• AFS LocalFS Example - File system access with AI agents
• AFS MCP Server Example - Mount MCP servers as AFS modules

Related Packages

• @aigne/afs - AFS core package
• @aigne/afs-user-profile-memory - User profile memory module

TypeScript Support

This package includes full TypeScript type definitions.

License

MIT