learning_agent

Learning Agent Example

This example demonstrates AgentGo's Learning System - agents that learn and improve over time.

Features Demonstrated

• User Profiles: Persistent user information across sessions
• User Memories: Extracted facts, preferences, and context from conversations
• Knowledge Transfer: Learned knowledge that can be shared across users
• SQLite Storage: Lightweight local storage for learning data

How It Works

1. Enable Learning: Set Learning: true and provide a LearningMachine
2. Automatic Extraction: After each conversation, the agent extracts:
   • User preferences ("I prefer short answers")
   • Factual information ("My name is John")
   • Contextual knowledge
3. Memory Retrieval: On subsequent interactions, the agent recalls learned information
4. Knowledge Sharing: Learned insights can benefit other users

Run the Example

export OPENAI_API_KEY=your-api-key
go run cmd/examples/learning_agent/main.go

Expected Output

=== First Interaction ===
Agent: Hello John! I'll keep my responses concise as you prefer.

=== Second Interaction ===
Agent: Your name is John.

=== Learned Memories ===
User Profile: {...}

User Memories (3):
  - [fact] My name is John
  - [preference] I prefer short, concise answers
  - [context] Hi! My name is John and I prefer short, concise answers.

Code Walkthrough

// Create learning storage
learningStorage, _ := sqlite.New("./learning.db")

// Create learning machine
learningMachine, _ := learning.NewMachine(learningStorage)

// Create agent with learning
ag, _ := agent.New(agent.Config{
    Name:            "Learning Assistant",
    Model:           model,
    UserID:          "user-123",      // Required for learning
    Learning:        true,             // Enable learning
    LearningMachine: learningMachine,
})

Storage Options

AgentGo supports multiple storage backends for learning:

SQLite (local, lightweight):

storage, _ := sqlite.New("./learning.db")

PostgreSQL (production, scalable):

import "github.com/jholhewres/agent-go/pkg/agentgo/learning/postgres"

db, _ := sql.Open("postgres", "postgresql://localhost/agentgo")
storage, _ := postgres.New(db, "public")

GDPR Compliance

Delete all user data:

err := learningMachine.DeleteUserData(ctx, "user-123")

Next Steps

• Try different conversation patterns
• Experiment with multiple users
• Integrate with PostgreSQL for production
• Build agents that transfer knowledge between users