agentbase 0.2.0

Open Source AI Agent Storage with Learning and Adaptation

AgentBase 🤖

AgentBase is a comprehensive open-source storage solution for AI agents with advanced memory management, experience replay, lifelong learning, and concept drift detection capabilities.

🚀 Features

• 🧠 Memory Management: Fast-access cache with multiple eviction policies (LRU, LFU, TTL, Priority)
• 🔄 Experience Replay: Reinforcement learning buffer with multiple sampling strategies
• 📚 Lifelong Learning: Track model updates, mistakes, and corrections over time
• 📊 Metadata Logging: Comprehensive training and experiment tracking
• 🔍 Concept Drift Detection: Monitor data distribution changes with multiple algorithms
• ⚡ High Performance: Thread-safe operations with efficient data structures
• 🎯 Easy Integration: Simple API that works with any ML framework

📦 Installation

From PyPI (Recommended)

pip install agentbase

With Optional Dependencies

# For machine learning features
pip install agentbase[ml]

# For visualization
pip install agentbase[viz]

# For all features
pip install agentbase[full]

# For development
pip install agentbase[dev]

From Source

git clone https://github.com/bestagents/agentbase.git
cd agentbase
pip install -e .

🎯 Quick Start

from agentbase import AgentBase

# Initialize AgentBase with default configuration
agent = AgentBase()

# Cache some data
agent.cache.set("user_preferences", {"theme": "dark", "lang": "en"})
cached_data = agent.cache.get("user_preferences")

# Store experience for reinforcement learning
experience = {
    "state": [1, 2, 3],
    "action": 0,
    "reward": 1.0,
    "next_state": [2, 3, 4],
    "done": False
}
agent.replay_buffer.add(**experience)

# Sample experiences for training
batch = agent.replay_buffer.sample(batch_size=32)

# Log training metadata
agent.logger.log_hyperparameters({"learning_rate": 0.001, "batch_size": 32})
agent.logger.log_metrics({"loss": 0.5, "accuracy": 0.95})

# Track model updates
agent.lifelong_learning.add_model_update(
    model_id="v1.0",
    performance={"accuracy": 0.95},
    metadata={"optimizer": "adam", "epochs": 100}
)

# Detect concept drift
new_data = [1.2, 2.3, 3.4, 4.5]
drift_detected = agent.drift_detector.detect_drift(new_data)
if drift_detected:
    print("Concept drift detected!")

🧩 Core Components

MemoryCache

Fast-access storage with multiple eviction policies:

from agentbase import MemoryCache

cache = MemoryCache(max_size=1000, eviction_policy="lru")
cache.set("key", "value", ttl=3600)  # TTL in seconds
value = cache.get("key")

ExperienceReplayBuffer

Store and sample experiences for reinforcement learning:

from agentbase import ExperienceReplayBuffer

buffer = ExperienceReplayBuffer(max_size=10000)
buffer.add(state=[1,2,3], action=0, reward=1.0, next_state=[2,3,4], done=False)
batch = buffer.sample(batch_size=32, method="uniform")

LifelongLearningStore

Track model evolution and learning patterns:

from agentbase import LifelongLearningStore

store = LifelongLearningStore()
store.add_model_update("v1.0", performance={"accuracy": 0.95})
store.record_mistake("incorrect_prediction", context={"input": [1,2,3]})
store.record_correction("fixed_logic", improvement=0.05)

MetadataLogger

Comprehensive experiment tracking:

from agentbase import MetadataLogger

logger = MetadataLogger(experiment_name="my_experiment")
logger.log_hyperparameters({"lr": 0.001, "batch_size": 32})
logger.log_metrics({"loss": 0.5, "accuracy": 0.95})
logger.export_data("experiment_results.json")

ConceptDriftDetector

Monitor data distribution changes:

from agentbase import ConceptDriftDetector

detector = ConceptDriftDetector(algorithm="ks_test")
detector.add_reference_data([1, 2, 3, 4, 5])
drift_detected = detector.detect_drift([1.5, 2.5, 3.5, 4.5])

🔧 Configuration

AgentBase supports flexible configuration:

config = {
    "memory_cache": {
        "max_size": 10000,
        "eviction_policy": "lru",
        "ttl": 3600
    },
    "replay_buffer": {
        "max_size": 50000,
        "alpha": 0.6,  # For prioritized sampling
        "beta": 0.4
    },
    "drift_detector": {
        "algorithm": "ks_test",
        "threshold": 0.05,
        "window_size": 1000
    }
}

agent = AgentBase(config=config)

📊 Performance Monitoring

AgentBase includes built-in performance monitoring:

# Get cache statistics
stats = agent.cache.get_statistics()
print(f"Cache hit rate: {stats['hit_rate']:.2%}")

# Monitor memory usage
memory_info = agent.get_memory_usage()
print(f"Memory usage: {memory_info['total_mb']:.2f} MB")

# Export performance data
agent.export_performance_data("performance_report.json")

🔄 Persistence

Save and load agent state:

# Save agent state
agent.save_state("agent_checkpoint.pkl")

# Load agent state
agent = AgentBase.load_state("agent_checkpoint.pkl")

🧪 Testing

Run the test suite:

# Install development dependencies
pip install agentbase[dev]

# Run tests
pytest tests/

# Run with coverage
pytest --cov=agentbase tests/

📚 Documentation

• API Reference
• Examples
• Tutorial

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

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

🙏 Acknowledgments

• Built with modern Python best practices
• Inspired by the need for robust AI agent storage solutions
• Thanks to all contributors and the open-source community

📈 Roadmap

• Advanced visualization dashboard
• Distributed storage backend
• Integration with popular ML frameworks
• Real-time monitoring and alerts
• Advanced drift detection algorithms
• Multi-agent coordination features

🔗 Links

• Homepage: https://bestagents.github.io/agentbase/
• PyPI: https://pypi.org/project/agentbase/
• GitHub: https://github.com/bestagents/agentbase
• Documentation: https://agentbase.readthedocs.io/
• Issues: https://github.com/bestagents/agentbase/issues

---

Made with ❤️ by the AgentBase team