Overview

LlamaAgent is a comprehensive AI agent framework designed for production environments. It provides a robust foundation for building intelligent agents that can reason, use tools, maintain memory, and integrate seamlessly with modern AI providers.

Key Features

• Multi-Provider LLM Support: OpenAI, Anthropic, Cohere, Together AI, and more
• Advanced Reasoning: ReAct pattern implementation with chain-of-thought capabilities
• Tool Integration: Extensible tool system with calculator, Python REPL, and custom tools
• Memory Management: Persistent memory with vector storage capabilities
• Production Ready: Comprehensive error handling, logging, and monitoring
• FastAPI Integration: RESTful API endpoints for web applications
• Docker Support: Containerized deployment with Kubernetes manifests
• Comprehensive Testing: 38+ tests with 100% pass rate

Quick Start

Installation

# Install from PyPI
pip install llamaagent

# Or install from source
git clone https://github.com/llamasearchai/llamaagent.git
cd llamaagent
pip install -e ".[dev]"

Basic Usage

from llamaagent.agents.react import ReactAgent
from llamaagent.agents.base import AgentConfig
from llamaagent.llm.providers.openai_provider import OpenAIProvider
from llamaagent.types import TaskInput

# Configure the agent
config = AgentConfig(
    name="MyAgent",
    description="A helpful AI assistant",
    tools_enabled=True
)

# Initialize LLM provider
provider = OpenAIProvider(api_key="your-api-key", model="gpt-4")

# Create the agent
agent = ReactAgent(config=config, llm_provider=provider)

# Execute a task
task = TaskInput(
    id="task-1",
    task="Calculate the square root of 144 and explain the process"
)

result = await agent.execute(task.task)
print(result.content)

Architecture

LlamaAgent follows a modular architecture designed for scalability and maintainability:

├── agents/          # Agent implementations (ReAct, reasoning chains)
├── llm/            # LLM provider integrations
├── tools/          # Tool system and implementations
├── memory/         # Memory management and storage
├── api/            # FastAPI web interfaces
├── monitoring/     # Observability and metrics
├── security/       # Authentication and validation
└── types/          # Core type definitions

Advanced Features

Tool System

from llamaagent.tools.calculator import CalculatorTool
from llamaagent.tools.python_repl import PythonREPLTool

# Register custom tools
agent.register_tool(CalculatorTool())
agent.register_tool(PythonREPLTool())

Memory Management

from llamaagent.memory.vector_memory import VectorMemory

# Configure persistent memory
memory = VectorMemory(
    embedding_model="text-embedding-3-large",
    storage_path="./agent_memory"
)
agent.set_memory(memory)

FastAPI Integration

from llamaagent.api.main import create_app

# Create web API
app = create_app()

# Run with: uvicorn main:app --host 0.0.0.0 --port 8000

Configuration

Environment Variables

# LLM Provider Keys
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key
COHERE_API_KEY=your_cohere_key

# Database Configuration
DATABASE_URL=postgresql://user:pass@localhost/db
REDIS_URL=redis://localhost:6379

# Monitoring
ENABLE_METRICS=true
LOG_LEVEL=INFO

Configuration File

# config/default.yaml
agent:
  name: "ProductionAgent"
  max_iterations: 10
  timeout: 300

llm:
  provider: "openai"
  model: "gpt-4"
  temperature: 0.7
  max_tokens: 2000

tools:
  enabled: true
  timeout: 30

memory:
  enabled: true
  type: "vector"
  max_entries: 10000

Deployment

Docker

# Build the image
docker build -t llamaagent:latest .

# Run the container
docker run -p 8000:8000 -e OPENAI_API_KEY=your_key llamaagent:latest

Kubernetes

# Deploy to Kubernetes
kubectl apply -f k8s/

Docker Compose

# Full stack deployment
docker-compose up -d

API Reference

Core Endpoints

• POST /agents/execute - Execute agent task
• GET /agents/{agent_id}/status - Get agent status
• POST /tools/execute - Execute tool directly
• GET /health - Health check endpoint

OpenAI Compatible API

# Chat completions
curl -X POST "http://localhost:8000/v1/chat/completions" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Testing

# Run all tests
pytest tests/ -v

# Run with coverage
pytest tests/ --cov=src --cov-report=html

# Run specific test categories
pytest tests/unit/ -v          # Unit tests
pytest tests/integration/ -v   # Integration tests
pytest tests/e2e/ -v          # End-to-end tests

Monitoring and Observability

Metrics

LlamaAgent provides comprehensive metrics for production monitoring:

• Request/response times
• Success/failure rates
• Token usage and costs
• Agent performance metrics
• Tool execution statistics

Logging

import logging
from llamaagent.monitoring.logging import setup_logging

# Configure structured logging
setup_logging(level=logging.INFO, format="json")

Health Checks

# Check system health
curl http://localhost:8000/health

# Detailed diagnostics
curl http://localhost:8000/diagnostics

Security

Authentication

from llamaagent.security.authentication import APIKeyAuth

# Configure API key authentication
auth = APIKeyAuth(api_keys=["your-secret-key"])
app.add_middleware(auth)

Input Validation

from llamaagent.security.validator import InputValidator

# Validate and sanitize inputs
validator = InputValidator()
safe_input = validator.sanitize(user_input)

Contributing

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

Development Setup

# Clone the repository
git clone https://github.com/llamasearchai/llamaagent.git
cd llamaagent

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install development dependencies
pip install -e ".[dev]"

# Run pre-commit hooks
pre-commit install

Code Quality

# Format code
black src/ tests/
isort src/ tests/

# Lint code
ruff check src/ tests/

# Type checking
mypy src/

# Security scan
bandit -r src/

Examples

Basic Agent

# examples/basic_agent.py
import asyncio
from llamaagent.agents.react import ReactAgent
from llamaagent.agents.base import AgentConfig
from llamaagent.llm.providers.mock_provider import MockProvider
from llamaagent.types import TaskInput

async def main():
    config = AgentConfig(name="BasicAgent")
    provider = MockProvider(model_name="test-model")
    agent = ReactAgent(config=config, llm_provider=provider)

    task = TaskInput(
        id="example-1",
        task="Explain quantum computing in simple terms"
    )

    result = await agent.arun(task)
    print(f"Agent Response: {result.content}")

if __name__ == "__main__":
    asyncio.run(main())

Multi-Agent System

# examples/multi_agent.py
import asyncio
from llamaagent.spawning.agent_spawner import AgentSpawner
from llamaagent.orchestration.adaptive_orchestra import AdaptiveOrchestra

async def main():
    spawner = AgentSpawner()
    orchestra = AdaptiveOrchestra()

    # Spawn multiple specialized agents
    research_agent = await spawner.spawn_agent("researcher")
    analysis_agent = await spawner.spawn_agent("analyst")
    writer_agent = await spawner.spawn_agent("writer")

    # Orchestrate collaborative task
    result = await orchestra.execute_collaborative_task(
        task="Write a comprehensive report on AI safety",
        agents=[research_agent, analysis_agent, writer_agent]
    )

    print(f"Collaborative Result: {result}")

if __name__ == "__main__":
    asyncio.run(main())

Benchmarks

LlamaAgent includes comprehensive benchmarking against industry standards:

• GAIA Benchmark: General AI Assistant evaluation
• SPRE Evaluation: Structured Problem Reasoning
• Custom Benchmarks: Domain-specific performance testing

# Run benchmarks
python -m llamaagent.benchmarks.run_all --provider openai --model gpt-4

Roadmap

• Multi-modal agent support (vision, audio)
• Advanced reasoning patterns (Tree of Thoughts, Graph of Thoughts)
• Federated learning capabilities
• Enhanced security features
• Performance optimizations
• Extended tool ecosystem

License

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

Support

• Documentation: https://llamaagent.readthedocs.io
• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: nikjois@llamasearch.ai

Acknowledgments

Built with love by Nik Jois and the LlamaSearch AI team.

Special thanks to the open-source community and all contributors who make this project possible.