Skip to content

Dooders/AgentMind

Repository files navigation

AgentMind

Status

AgentMind is a modular cognitive framework for building interpretable, adaptive artificial agents. Each cognitive function—like attention, memory, inhibition, and reasoning—is modeled as an independent module. This design allows researchers and developers to simulate humanlike behavior, evaluate decision-making processes, and study modular learning dynamics in artificial agents.


Future Features

  • 🧠 Modular Cognition: Plug-and-play modules for perception, memory, decision-making, and self-regulation.
  • 📈 Learning by Design: Support for reinforcement, supervised, or rule-based learning inside each module.
  • 🔍 Interpretability: Track internal states and outputs of each module step-by-step.
  • 🧪 Diagnostics Ready: Built-in support for logging, introspection, and post-simulation assessment.
  • 🔄 Evolvable Architecture: Modules can be tuned, replaced, or evolved for different simulation needs.

Example Modules

Module Description
Attention Control Filters inputs based on salience and relevance
Working Memory Temporarily holds active concepts and data
Impulse Inhibition Blocks hasty or low-confidence actions
Abstract Reasoning Performs symbolic or conceptual inference
Metacognition Self-monitors and tunes thresholds dynamically
Long-Term Memory Stores and retrieves persistent experience
Goal Maintenance Manages long-term objectives across time

Planned Code Structure

agentmind/
├── __init__.py
├── core/
│   ├── __init__.py
│   ├── agent.py              # AgentMindCore and runtime loop
│   ├── context.py            # Context/state object shared between modules
│   ├── memory.py             # WorkingMemory, LongTermMemory
│   └── feedback.py           # Outcome tracking, delayed reward manager
├── modules/
│   ├── __init__.py
│   ├── base.py               # MentalModule base class
│   ├── attention.py
│   ├── working_memory.py
│   ├── inhibition.py
│   ├── reasoning.py
│   ├── flexibility.py
│   ├── emotion.py
│   ├── pattern_recognition.py
│   ├── long_term_memory.py
│   ├── metacognition.py
│   └── willpower.py
├── learning/
│   ├── __init__.py
│   ├── rl.py                 # Reinforcement learning support
│   ├── heuristics.py         # Simple rule/threshold updaters
│   └── evaluation.py         # Success metrics, logs
├── envs/
│   ├── __init__.py
│   ├── toy_env.py            # Minimal delayed-reward environment
│   └── gym_adapter.py        # Optional wrapper for OpenAI Gym
├── utils/
│   ├── __init__.py
│   ├── logger.py             # Logging, visualization, plots
│   └── persistence.py        # Save/load agent state, memory, config
└── demo/
    ├── run_demo.py           # Runs a scripted scenario
    └── config.json           # Demo parameters

About

Proposal for an intelligent computational agent

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published