Examples

Real-world examples demonstrating MAIF capabilities.

More Examples on DeepWiki

For a comprehensive list of examples with auto-generated documentation, visit DeepWiki - Examples and Use Cases.

Featured Example

LangGraph Multi-Agent RAG System

Production-ready multi-agent research assistant with cryptographic provenance.

Location: examples/langgraph/

Features:

• Five specialized agents (Init, Retrieve, Synthesize, Fact-Check, Citation)
• Real ChromaDB vector search with 384-dim embeddings
• Gemini API for generation and verification
• LLM-based fact-checking with iterative refinement
• Complete audit trail in MAIF artifacts
• Interactive console interface
• Multi-turn conversation support

Quick Start:

cd examples/langgraph
echo "GEMINI_API_KEY=your_key" > .env
pip install -r requirements_enhanced.txt
python3 create_kb_enhanced.py
python3 demo_enhanced.py

Documentation: See LangGraph RAG Guide for complete details.

---

CrewAI Research Crew

Multi-agent research workflow with complete audit trails.

Location: Uses maif.integrations.crewai

Features:

• Two specialized agents (Researcher, Writer)
• Task and step-level provenance tracking
• Persistent agent memory with search
• Complete execution audit trail
• Error handling with logging

Quick Start:

from crewai import Crew, Agent, Task
from maif.integrations.crewai import MAIFCrewCallback

callback = MAIFCrewCallback("session.maif")
crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, write_task],
    task_callback=callback.on_task_complete,
    step_callback=callback.on_step,
)
result = crew.kickoff()
callback.finalize()

Documentation: See CrewAI Research Guide for complete details.

---

Available Examples

Hello World

Get Started →

The simplest possible MAIF agent. Perfect for understanding the basics.

What you'll learn:

• Creating MAIF artifacts
• Adding text blocks
• Saving and loading
• Basic verification

Time: 5 minutes

---

Multi-Agent System

View Example →

Multiple agents collaborating through shared MAIF artifacts.

What you'll learn:

• Agent coordination
• Shared memory patterns
• Provenance tracking
• Multi-agent workflows

Time: 15 minutes

---

Privacy & Security

View Example →

Privacy-preserving agent with encryption and anonymization.

What you'll learn:

• AES-GCM encryption
• Differential privacy
• Data anonymization
• Access control

Time: 10 minutes

---

Streaming Data

View Example →

High-throughput streaming with memory-mapped I/O.

What you'll learn:

• Streaming operations
• Memory-mapped I/O
• Performance optimization
• Large file handling

Time: 15 minutes

---

Financial Agent

View Example →

Privacy-compliant financial transaction analysis.

What you'll learn:

• Regulatory compliance
• Transaction analysis
• Audit trails
• Risk scoring

Time: 20 minutes

---

Distributed Processing

View Example →

Distributed agent systems with MAIF synchronization.

What you'll learn:

• Distributed coordination
• State synchronization
• Network protocols
• Fault tolerance

Time: 25 minutes

---

Quick Start Examples

Hello World Agent (30 seconds)

The simplest possible MAIF agent:

from maif_api import create_maif

# Create agent with memory
memory = create_maif("hello-agent")

# Add content
memory.add_text("Hello, MAIF world!", title="Greeting")

# Save with cryptographic signing
memory.save("hello.maif", sign=True)

print("Your first AI agent memory is ready!")

Privacy-Enabled Chat Agent (2 minutes)

A more realistic agent with memory and privacy:

from maif_api import create_maif, load_maif
import os

class PrivateChatAgent:
    def __init__(self, agent_id: str):
        self.agent_id = agent_id
        self.memory_path = f"{agent_id}_memory.maif"
        
        # Load or create memory with privacy
        if os.path.exists(self.memory_path):
            self.memory = load_maif(self.memory_path)
        else:
            self.memory = create_maif(agent_id, enable_privacy=True)
    
    def chat(self, message: str, user_id: str) -> str:
        # Store message with privacy protection
        self.memory.add_text(
            f"User {user_id}: {message}",
            title="User Message",
            encrypt=True,
            anonymize=True  # Remove PII automatically
        )
        
        # Search for relevant context
        context = self.memory.search(message, top_k=3)
        
        # Generate response (integrate your LLM here)
        response = f"I understand you're asking about: {message}"
        
        # Store response
        self.memory.add_text(
            f"Agent: {response}",
            title="Agent Response",
            encrypt=True
        )
        
        return response
    
    def save(self):
        self.memory.save(self.memory_path, sign=True)

# Usage
agent = PrivateChatAgent("support-bot")
response = agent.chat("How do I reset my password?", "user123")
print(response)
agent.save()

Example Categories

By Experience Level

Beginner:

• Hello World - Your first MAIF agent
• Privacy Demo - Basic privacy features

Intermediate:

• Multi-Agent - Agent coordination
• Streaming - High-performance I/O
• Financial Agent - Production patterns

Advanced:

• LangGraph RAG - Complete multi-agent system
• CrewAI Research - Multi-agent workflows with CrewAI
• Distributed - Distributed systems

By Use Case

AI/ML Applications:

• LangGraph RAG - Research assistant with fact-checking
• CrewAI Research - Research and documentation workflow
• Multi-Agent - Collaborative agents

Enterprise:

• Financial Agent - Regulatory compliance
• Privacy Demo - Data protection

Performance:

• Streaming - High throughput
• Distributed - Scale-out architecture

Running the Examples

All examples follow the same pattern:

# 1. Navigate to repository root
cd /path/to/maif

# 2. Install dependencies (if needed)
pip install -e .

# 3. Run the example
python3 examples/<category>/<example_file>.py

For the LangGraph example:

cd examples/langgraph
pip install -r requirements_enhanced.txt
python3 demo_enhanced.py

Example Structure

Each example includes:

• Complete, runnable code
• Comprehensive error handling
• Performance optimizations
• Security best practices
• Testing and validation
• Detailed documentation

Contributing Examples

Have a great example to share? We welcome contributions!

1. Create your example in examples/<category>/
2. Add documentation in docs/examples/
3. Include README with usage instructions
4. Submit a pull request

Support

• Documentation: See the User Guide
• API Reference: Check the API docs
• Issues: Report problems on GitHub

---

Every example is designed to be production-ready. Copy, modify, and deploy with confidence.