AgentMeter Python SDK v0.3.1

Modern Python SDK for usage tracking and billing in AI applications.

✨ Features

• 🎯 Resource-Based Architecture - Clean, organized API with dedicated resource managers
• ⚡ Async/Await Support - Full async support for modern Python applications
• 🔌 Rich Integrations - First-class support for LangChain, Coinbase AgentKit, and more
• 🛡️ Type Safety - Built with Pydantic v2 for robust data validation and type hints
• 🌐 Production Ready - Enterprise-grade reliability with retry logic and error handling
• 📊 Flexible Billing - Support for usage-based, token-based, and instant billing models
• 🔄 100% Backward Compatible - All v0.2.0 code continues to work unchanged

🚀 Quick Start

Installation

pip install agentmeter

Basic Usage

from agentmeter import AgentMeter

# Initialize client
with AgentMeter(api_key="your_api_key") as meter:
    # Create a meter type
    meter_type = meter.meter_types.create(
        name="api_calls",
        unit="requests",
        description="API usage tracking"
    )
    
    # Record usage event
    event = meter.meter_events.record(
        meter_type_id=meter_type.id,
        subject_id="user_123",
        quantity=1.0,
        metadata={"endpoint": "/api/search"}
    )
    
    # Get usage statistics
    stats = meter.meter_events.get_aggregations(
        meter_type_id=meter_type.id,
        subject_id="user_123"
    )
    print(f"Total usage: {stats.total_quantity} {meter_type.unit}")

🏗️ Core Concepts

Resource Managers

AgentMeter v0.3.1 organizes functionality into focused resource managers:

meter.projects          # Project management
meter.meter_types       # Meter type definitions  
meter.meter_events      # Usage event recording and querying
meter.users            # User-specific meter management

Async Support

from agentmeter import AsyncAgentMeter

async with AsyncAgentMeter(api_key="your_api_key") as meter:
    # All methods have async equivalents
    meter_type = await meter.meter_types.create(...)
    event = await meter.meter_events.record(...)
    stats = await meter.meter_events.get_aggregations(...)

🔌 Integrations

🚀 Coinbase AgentKit

Build monetized Web3 AI agents:

from agentmeter import AgentMeter

meter = AgentMeter(api_key="your_api_key") 

# Create meter types for different Web3 operations
wallet_queries = meter.meter_types.create(
    name="wallet_queries", 
    unit="queries",
    description="Blockchain wallet queries"
)

trading_operations = meter.meter_types.create(
    name="trading_ops",
    unit="trades", 
    description="Smart contract trading"
)

# Track usage in your AgentKit application
def get_wallet_balance(user_id: str, asset: str):
    # Your Web3 logic here
    balance = wallet.get_balance(asset)
    
    # Record billable event
    meter.meter_events.record(
        meter_type_id=wallet_queries.id,
        subject_id=user_id,
        quantity=1.0,
        metadata={"asset": asset, "operation": "balance_check"}
    )
    
    return balance

📖 Full Coinbase AgentKit Integration Example

🤖 LangChain

Automatic LLM usage tracking:

from agentmeter.langchain_integration import LangChainAgentMeterCallback

# Create callback with your meter
callback = LangChainAgentMeterCallback(
    agentmeter_client=meter,
    meter_type_id="llm_usage_meter_id"
)

# Add to any LangChain component
llm = ChatOpenAI(callbacks=[callback])
chain = ConversationChain(llm=llm, callbacks=[callback])

# Usage automatically tracked
result = chain.run("Analyze this data...")

📖 Full LangChain Integration Example

💰 Billing Models

Usage-Based Billing

Perfect for API calls, processing requests, or resource consumption:

# Create usage meter
api_meter = meter.meter_types.create(
    name="api_requests",
    unit="requests"
)

# Record usage
meter.meter_events.record(
    meter_type_id=api_meter.id,
    subject_id="user_123", 
    quantity=1.0
)

Token-Based Billing

Ideal for AI/LLM applications:

# Create token meter
token_meter = meter.meter_types.create(
    name="llm_tokens",
    unit="tokens"
)

# Record token usage
meter.meter_events.record(
    meter_type_id=token_meter.id,
    subject_id="user_123",
    quantity=1500.0,  # Total tokens used
    metadata={
        "input_tokens": 1000,
        "output_tokens": 500,
        "model": "gpt-4"
    }
)

Event-Based Billing

For feature unlocks or one-time charges:

# Create event meter  
feature_meter = meter.meter_types.create(
    name="premium_features",
    unit="unlocks"
)

# Record feature usage
meter.meter_events.record(
    meter_type_id=feature_meter.id,
    subject_id="user_123",
    quantity=1.0,
    metadata={
        "feature": "advanced_analytics",
        "tier": "premium"
    }
)

📊 Analytics & Reporting

Usage Aggregations

# Get aggregated usage statistics
stats = meter.meter_events.get_aggregations(
    meter_type_id=meter_type.id,
    subject_id="user_123",
    start_time="2024-01-01T00:00:00Z",
    end_time="2024-01-31T23:59:59Z"
)

print(f"Total usage: {stats.total_quantity}")
print(f"Event count: {stats.total_events}")
print(f"Period: {stats.period_start} to {stats.period_end}")

User Meter Management

# Set user spending limits
user_meter = meter.users.set_meter(
    user_id="user_123",
    threshold_amount=100.0  # $100 monthly limit
)

# Check current usage
current_usage = meter.users.get_meter(user_id="user_123")
print(f"Usage: ${current_usage.current_usage} / ${current_usage.threshold_amount}")

# Reset usage (e.g., monthly reset)
meter.users.reset_meter(user_id="user_123")

🛠️ Configuration

Environment Variables

export AGENTMETER_API_KEY="your_api_key"
export AGENTMETER_BASE_URL="https://api.agentmeter.money"  # Optional

Programmatic Configuration

from agentmeter import AgentMeter

meter = AgentMeter(
    api_key="your_api_key",
    base_url="https://api.agentmeter.money",  # Optional
    timeout=30.0,
    max_retries=3
)

🔄 Migration from v0.2.0

AgentMeter v0.3.1 maintains 100% backward compatibility. Your existing code continues to work:

# ✅ This still works (legacy API)
from agentmeter import AgentMeterClient
client = AgentMeterClient(api_key="key")
client.record_api_request_pay(api_calls=1, unit_price=0.10)

# ✨ New recommended approach (v0.3.1)
from agentmeter import AgentMeter
with AgentMeter(api_key="key") as meter:
    meter.meter_events.record(meter_type_id="mt_123", subject_id="user", quantity=1.0)

For a smooth migration:

1. Keep existing code working - No immediate changes required
2. Gradually adopt new API - Use v0.3.1 for new features
3. Migrate when convenient - Update modules one at a time

📖 See complete examples: v0.3.1 vs Legacy

🔧 Error Handling

from agentmeter import AgentMeter
from agentmeter.exceptions import AuthenticationError, ValidationError, ServerError

try:
    with AgentMeter(api_key="invalid") as meter:
        event = meter.meter_events.record(...)
        
except AuthenticationError:
    print("Invalid API key")
except ValidationError as e:
    print(f"Data validation failed: {e}")
except ServerError as e:
    print(f"Server error: {e}")

🧪 Examples & Testing

# Run the new v0.3.1 examples
python examples/v031_new_features.py
python examples/coinbase_agentkit_integration.py  
python examples/langchain_integration_meter.py
python examples/ecommerce_integration.py

# Run tests
pytest tests/

📚 Documentation

• API Reference
• Integration Guides
• 🗂️ Legacy Documentation (v0.2.0 and earlier)

🤝 Support

• GitHub Issues
• Documentation
• Email Support

📄 License

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

---

🚀 Ready to get started? Install AgentMeter and start tracking usage in minutes!

pip install agentmeter