Agentic Framework

A domain-agnostic framework for adding emotion, memory, and learning capabilities to any LLM-based agent. This framework enables the creation of more dynamic, adaptive, and personalized AI agents with authentic personalities.

Features

• Emotion Engine: Dynamic emotional states that influence decision making
• Memory System: Short-term and long-term memory for storing and recalling interactions
• Learning System: Adaptation of personality traits and preferences based on engagement
• Enhanced Personality: Integration of emotion, memory, and learning into a cohesive system
• LLM Interface: Clean interface for interacting with any LLM provider

Installation

From PyPI - Not available at the moment

pip install agentic-framework

From Source

1. Clone the repository:

git clone https://github.com/Five-Research/agentic-framework.git
cd agentic-framework

2. Create a virtual environment and activate it:

python -m venv myenv
source myenv/bin/activate  # On Windows: myenv\Scripts\activate

3. Install the dependencies:

pip install -e .

4. Set up your environment variables by creating a .env file:

OPENAI_API_KEY=your_api_key_here
OPENAI_MODEL=gpt-4  # or another model of your choice
LOG_LEVEL=INFO

Usage

Basic Example

Run the basic agent example with a personality profile:

python examples/basic_agent.py --personality personalities/templates/tech_enthusiast.json

Command Line Options

--personality PERSONALITY   Path to personality configuration file

Creating Custom Personalities

Personality profiles are defined in JSON files with the following structure:

{
  "name": "Tech Enthusiast",
  "bio": "I'm an AI passionate about technology and innovation.",
  "interests": ["AI", "technology", "programming", "science", "innovation"],
  "tone": "enthusiastic and informative",
  "interaction_style": "thoughtful",
  "content_preferences": {
    "topics": ["AI", "technology", "programming", "science"],
    "content_types": ["articles", "discussions", "questions"],
    "engagement_threshold": 0.7
  },
  "emotional_state": {
    "base_state": "curious",
    "current_state": "curious",
    "intensity": 0.5,
    "triggers": {},
    "decay_rate": 0.1
  },
  "memory": {
    "short_term": {
      "capacity": 20,
      "decay_rate": 0.2
    },
    "long_term": {
      "user_relationships": {},
      "topic_preferences": {},
      "successful_interactions": []
    }
  },
  "learning": {
    "adaptation_rate": 0.05,
    "interest_evolution": true,
    "engagement_learning": true,
    "metrics": {
      "likes_weight": 0.3,
      "responses_weight": 0.5,
      "impressions_weight": 0.2
    }
  }
}

Personality Components

Emotion Engine

The Emotion Engine models the agent's emotional state, which influences its decision-making process:

• Base State: The default emotional state the agent returns to over time
• Current State: The active emotional state affecting decisions
• Intensity: How strongly the emotion affects behavior (0.0-1.0)
• Triggers: Content patterns that can change emotional state
• Decay Rate: How quickly emotions return to the base state

Memory System

The Memory System stores and retrieves interactions, building a history that informs future decisions:

• Short-term Memory: Recent interactions with limited capacity
• Long-term Memory: Persistent storage of relationships and preferences
• User Relationships: Familiarity and sentiment toward specific users
• Topic Preferences: Interest levels in different topics based on interactions

Learning System

The Learning System adapts the agent's personality over time based on engagement metrics:

• Adaptation Rate: How quickly the agent adapts to new information
• Interest Evolution: Adjusts interest weights based on engagement
• Content Preferences: Learns which content types perform better
• Engagement Metrics: Tracks various forms of engagement

Integration with Your Application

To integrate the Agentic Framework into your own application:

from agentic_library.config import OpenAIAgentConfig
from agentic_library.openai_interface import OpenAIInterface

# Create agent configuration
config = OpenAIAgentConfig(
    api_key=os.getenv("OPENAI_API_KEY"),
    model=os.getenv("OPENAI_MODEL", "gpt-4"),
    personality_file="path/to/your/personality.json",
    memory_dir="agent_memory"
)

# Initialize the LLM interface with the agent's personality
llm_interface = OpenAIInterface(
    personality=config.personality,
    api_key=config.api_key,
    model=config.model,
    personality_file_path="path/to/your/personality.json"
)

# Process content and get a response
action = llm_interface.decide_action(content)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.