march-history 0.2.0

Python SDK for March AI Conversation History API

March Conversation History SDK

A Python SDK for the March AI Conversation History API with full type safety, automatic retries, and pagination support.

Features

✨ Type-Safe: Full type hints with Pydantic v2 models 🔄 Auto-Retry: Exponential backoff with configurable retry logic 📄 Auto-Pagination: Iterate over large result sets effortlessly 🚀 Sync & Async: Support for both synchronous and asynchronous operations 🧹 Auto-Cleanup: No context managers needed - automatic resource management 🎯 Best Practices: Follows Python SDK design patterns used by major APIs

Installation

pip install march-history

Or install from source:

git clone https://github.com/march/march-history-sdk
cd march-history-sdk
pip install -e .

Quick Start

from march_history import MarchHistoryClient
from march_history.models import MessageRole, ConversationStatus

# Create client - no context manager needed!
client = MarchHistoryClient(base_url="http://localhost:8000")

# Create a conversation
conv = client.conversations.create(
    tenant_name="acme-corp",
    title="Customer Support Session",
    metadata={"user_id": "user-123", "session_id": "sess-456"}
)

# Add messages
msg1 = client.messages.create(
    conversation_id=conv.id,
    role=MessageRole.USER,
    content="Hello, I need help with my order"
)

msg2 = client.messages.create(
    conversation_id=conv.id,
    role=MessageRole.ASSISTANT,
    content="I'd be happy to help! What's your order number?"
)

# List all messages in conversation
messages = client.messages.list(conversation_id=conv.id)
for msg in messages:
    print(f"{msg.role}: {msg.content}")

# Auto-pagination - iterate over all conversations
for conversation in client.conversations.list_iter(tenant_name="acme-corp"):
    print(f"Conversation: {conversation.title}")

# Search conversations
results = client.conversations.search(
    q="support",
    status=ConversationStatus.ACTIVE
)

# Archive conversation when done
client.conversations.archive(conv.id)

# No need to call client.close() - automatic cleanup!

Usage Guide

Creating Conversations

# Basic conversation
conv = client.conversations.create(
    tenant_name="acme-corp",
    title="Support Chat"
)

# With full metadata
conv = client.conversations.create(
    tenant_name="acme-corp",
    title="Sales Inquiry",
    agent_identifier="sales-bot-v2",
    status=ConversationStatus.ACTIVE,
    metadata={
        "user_id": "user-123",
        "source": "web_chat",
        "priority": "high"
    }
)

Working with Messages

# Single message
msg = client.messages.create(
    conversation_id=conv.id,
    role=MessageRole.USER,
    content="What are your business hours?",
    metadata={"client_ip": "192.168.1.1"}
)

# Batch create messages
msgs = client.messages.create_batch(
    conversation_id=conv.id,
    messages=[
        {"role": "user", "content": "Hello"},
        {"role": "assistant", "content": "Hi! How can I help?"},
        {"role": "user", "content": "I need information"},
    ]
)

# Search messages in a conversation
results = client.messages.search(
    conversation_id=conv.id,
    q="business hours"
)

# Search across all conversations
all_results = client.messages.search_global(
    q="error",
    role=MessageRole.ASSISTANT
)

Pagination

# Manual pagination - single page
conversations = client.conversations.list(
    tenant_name="acme-corp",
    offset=0,
    limit=50
)

# Auto-pagination - iterate over all items
for conv in client.conversations.list_iter(
    tenant_name="acme-corp",
    page_size=100  # Items per page
):
    print(conv.title)

# Limit total items fetched
for conv in client.conversations.list_iter(
    tenant_name="acme-corp",
    max_items=500  # Stop after 500 items
):
    print(conv.title)

Searching

# Search conversations by title
results = client.conversations.search(q="customer support")

# Search by metadata
results = client.conversations.search(
    metadata_key="environment",
    metadata_value="production"
)

# Combined search with filters
results = client.conversations.search(
    q="order",
    status=ConversationStatus.ACTIVE,
    tenant_name="acme-corp",
    agent_identifier="support-bot-v1"
)

# Search with auto-pagination
for conv in client.conversations.search_iter(q="urgent"):
    print(conv.title)

Managing Tenants

# List all tenants
tenants = client.tenants.list()

# Get tenant by ID
tenant = client.tenants.get(tenant_id=1)

# Get tenant by name
tenant = client.tenants.get_by_name("acme-corp")

# Iterate over all tenants
for tenant in client.tenants.list_iter():
    print(tenant.name)

Configuration

Basic Configuration

client = MarchHistoryClient(
    base_url="http://localhost:8000",
    timeout=60.0,  # Request timeout in seconds
    max_retries=5,  # Maximum retry attempts
)

Advanced Retry Configuration

from march_history.config import RetryConfig

client = MarchHistoryClient(
    base_url="http://localhost:8000",
    retry_config=RetryConfig(
        max_retries=5,
        backoff_factor=2.0,  # Exponential backoff multiplier
        retry_status_codes=(408, 429, 500, 502, 503, 504),
        max_backoff_seconds=120.0  # Max backoff time
    )
)

API Key Authentication

# Using api_key parameter (recommended)
client = MarchHistoryClient(
    base_url="http://localhost:8000",
    api_key="your-api-key-123"  # Sent as X-API-Key header
)

Custom Headers

client = MarchHistoryClient(
    base_url="http://localhost:8000",
    api_key="your-api-key",  # API key authentication
    custom_headers={
        "X-Custom-Header": "value"  # Additional custom headers
    }
)

Error Handling

The SDK provides a comprehensive exception hierarchy:

from march_history.exceptions import (
    MarchHistoryError,  # Base exception
    APIError,  # Base for API errors
    ValidationError,  # 422 validation errors
    NotFoundError,  # 404 not found
    ConflictError,  # 409 conflict (e.g., duplicate sequence number)
    BadRequestError,  # 400 bad request
    ServerError,  # 500+ server errors
    NetworkError,  # Network/connection errors
    RetryError,  # Max retries exceeded
)

try:
    conv = client.conversations.get(999)
except NotFoundError as e:
    print(f"Conversation not found: {e.message}")
    print(f"Status code: {e.status_code}")
    print(f"Error details: {e.details}")
except ValidationError as e:
    print(f"Validation failed: {e.message}")
    for detail in e.details:
        print(f"  - {detail['field']}: {detail['message']}")
except RetryError as e:
    print(f"Max retries exceeded: {e}")
    print(f"Last error: {e.last_exception}")
except APIError as e:
    print(f"API error: {e}")

Advanced Usage

Optional Context Manager

While not required, you can use context managers for explicit resource control:

# Guaranteed cleanup even if exceptions occur
with MarchHistoryClient(base_url="http://localhost:8000") as client:
    conv = client.conversations.create(
        tenant_name="acme-corp",
        title="Chat"
    )
    # ... work with client ...
# Connections automatically closed

Accessing Response Metadata

All models provide access to timestamps and IDs:

conv = client.conversations.create(
    tenant_name="acme-corp",
    title="Support"
)

print(f"ID: {conv.id}")
print(f"Tenant ID: {conv.tenant_id}")
print(f"Created: {conv.created_at}")
print(f"Updated: {conv.updated_at}")
print(f"Metadata: {conv.metadata}")

Working with Metadata

Metadata fields accept arbitrary JSON:

# Store custom data
conv = client.conversations.create(
    tenant_name="acme-corp",
    title="Support",
    metadata={
        "user": {
            "id": "user-123",
            "email": "[e-mail]",
            "tier": "premium"
        },
        "source": "mobile_app",
        "version": "1.2.3",
        "tags": ["urgent", "billing"]
    }
)

# Update metadata
conv = client.conversations.update(
    conv.id,
    metadata={
        **conv.metadata,
        "resolved": True,
        "resolution_time": 300
    }
)

Type Safety

The SDK is fully typed for excellent IDE support:

from march_history.models import MessageRole, ConversationStatus

# Type-safe enums
role: MessageRole = MessageRole.USER
status: ConversationStatus = ConversationStatus.ACTIVE

# Pydantic models validate at runtime
try:
    client.messages.create(
        conversation_id=1,
        role="invalid_role",  # Will raise ValidationError
        content="Hello"
    )
except ValidationError as e:
    print(f"Invalid role: {e}")

Examples

See the examples/ directory for more examples:

• basic_usage.py - Complete workflow example
• pagination_example.py - Pagination patterns
• error_handling.py - Error handling strategies

Requirements

• Python 3.11+
• httpx >= 0.25.0
• pydantic >= 2.0.0

Development

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

# Run tests
pytest

# Run tests with coverage
pytest --cov=march_history --cov-report=html

# Run type checking
mypy src/march_history

# Run linting
ruff check src/march_history

# Format code
ruff format src/march_history

Testing

The SDK includes comprehensive tests:

• Unit tests: Models, exceptions, configuration
• Integration tests: Resources, API interactions (mocked with respx)
• Pagination tests: Auto-pagination, page iterators
• Retry tests: Exponential backoff, retry configuration

Run all tests:

pytest

Run specific test files:

pytest tests/test_models.py
pytest tests/test_resources.py
pytest tests/test_retry.py

Run with verbose output:

pytest -v

Test coverage:

pytest --cov=march_history --cov-report=term-missing

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

For issues and questions:

• GitHub Issues: https://github.com/march/march-history-sdk/issues
• Documentation: https://github.com/march/march-history-sdk#readme