Python SDK

Alpha

Complete API reference for the Bear Lumen Python SDK. Track AI usage, manage end users, and analyze costs with full type hints.

Also available: Node.js SDK . Same API surface, with TypeScript support.

Installation & Configuration

Install the SDK from PyPI:

pip install bearlumen

Initialize the client with your API key:

import bearlumen

client = bearlumen.Client(api_key='your_api_key')

api_key is the only required parameter.

End Users

Manage end users with client.end_users.

Method	Description
end_users.create(external_id, email=None, name=None, metadata=None)	Create a new end user
end_users.update(end_user_id, ...)	Update an end user
end_users.list(status=None, limit=None, offset=None)	List end users
end_users.deactivate(end_user_id)	Soft-delete an end user
end_users.reactivate(end_user_id)	Restore a deactivated end user

Create an End User

result = client.end_users.create(
    external_id='user_123',
    name='AI Startup Inc',
    metadata={'ai_provider': 'openai'}
)
print(result['endUser']['id'])

Update an End User

result = client.end_users.update(
    'eu_12345',
    name='Updated Name',
    metadata={'plan': 'enterprise'}
)

List End Users

result = client.end_users.list(
    status='ACTIVE',   # Optional: 'ACTIVE' or 'DEACTIVATED'
    limit=50,          # Optional
    offset=0,          # Optional
)

Deactivate / Reactivate

# Soft delete (usage history retained)
client.end_users.deactivate('eu_12345')

# Restore access
client.end_users.reactivate('eu_12345')

Subscriptions

Manage subscriptions with client.subscriptions.

Method	Description
subscriptions.create(organization_id, plan_id, billing_cycle, start_date, ...)	Create a new subscription
subscriptions.retrieve(subscription_id)	Retrieve a subscription by ID
subscriptions.update(subscription_id, ...)	Update a subscription
subscriptions.cancel(subscription_id)	Cancel a subscription
subscriptions.list(organization_id=None, limit=None, offset=None)	List all subscriptions

Create a Subscription

subscription = client.subscriptions.create(
    organization_id='org_123',
    plan_id='plan_pro_monthly',
    billing_cycle='monthly',
    start_date='2024-01-01T00:00:00Z',
)

Retrieve, Update, Cancel, List

# Get by ID
sub = client.subscriptions.retrieve('sub_12345')

# Update fields
updated = client.subscriptions.update('sub_12345',
    plan_id='plan_enterprise',
)

# Cancel
cancelled = client.subscriptions.cancel('sub_12345')

# List with optional filters
result = client.subscriptions.list(
    organization_id='org_123',  # Optional
    limit=20,                   # Optional
    offset=0,                   # Optional
)

Usage Tracking

Track and query usage with client.usage.

Method	Description
usage.record(event_type, metric_name, quantity, ...)	Record a usage event
usage.record_batch(events)	Record multiple events (max 1000)
usage.query(subscription_id=None, ...)	Query usage events
usage.aggregation(subscription_id=None, ...)	Get aggregated usage by metric
usage.summary(subscription_id=None, ...)	Get high-level usage summary

Record a Usage Event

Pass agent_id and workflow_id as convenience parameters for cost attribution.

result = client.usage.record(
    event_type='token_usage',
    metric_name='llm_tokens',
    quantity=150000,
    idempotency_key='invoice-001-summary',
    agent_id='support-bot',
    workflow_id='customer-onboarding',
    metadata={
        'model': 'gpt-4-turbo',
        'provider': 'openai',
        'provider_cost_usd': 4.50,
    }
)

if not result['recorded']:
    print('Duplicate event skipped')

Record Batch Events

result = client.usage.record_batch([
    {
        'event_type': 'token_usage',
        'metric_name': 'llm_tokens',
        'quantity': 1000,
        'idempotency_key': 'batch-1-evt-1',
    },
    {
        'event_type': 'token_usage',
        'metric_name': 'llm_tokens',
        'quantity': 2000,
        'idempotency_key': 'batch-1-evt-2',
    },
])
print(f"Recorded: {result['recorded_count']}, Skipped: {result['skipped_count']}")

Query Usage Events

events = client.usage.query(
    subscription_id='sub_123',
    metric_name='llm_tokens',
    start_date='2024-01-01',
    end_date='2024-01-31',
    limit=100,
)

Get Aggregated Usage

aggregation = client.usage.aggregation(
    subscription_id='sub_123',
    start_date='2024-01-01',
    end_date='2024-01-31',
)

Get Usage Summary

summary = client.usage.summary(
    subscription_id='sub_123',
    start_date='2024-01-01',
    end_date='2024-01-31',
)

Cost Attribution

Analyze costs by AI agent or workflow with client.cost_attribution.

Method	Description
cost_attribution.by_agent(start_date, end_date)	Cost breakdown grouped by agent
cost_attribution.by_workflow(start_date, end_date)	Cost breakdown grouped by workflow
cost_attribution.list_agents(start_date, end_date)	List unique agent identifiers
cost_attribution.list_workflows(start_date, end_date)	List unique workflow identifiers

Cost Breakdown by Agent

result = client.cost_attribution.by_agent(
    start_date='2024-01-01T00:00:00Z',
    end_date='2024-02-01T00:00:00Z',
)

for item in result['items']:
    print(f"{item['attributionValue']}: ${item['totalCost']}")
print(f"Total: ${result['totalCost']}")

Cost Breakdown by Workflow

result = client.cost_attribution.by_workflow(
    start_date='2024-01-01T00:00:00Z',
    end_date='2024-02-01T00:00:00Z',
)

List Unique Agents

result = client.cost_attribution.list_agents(
    start_date='2024-01-01T00:00:00Z',
    end_date='2024-02-01T00:00:00Z',
)
print('Active agents:', result['values'])

List Unique Workflows

result = client.cost_attribution.list_workflows(
    start_date='2024-01-01T00:00:00Z',
    end_date='2024-02-01T00:00:00Z',
)
print('Active workflows:', result['values'])

Exception Classes

The Python SDK provides a typed exception hierarchy for precise error handling:

Exception Hierarchy

BearLumenError- Base exception class

└InvalidEmailError- Email format is invalid

└DuplicateCustomerError- End user with this external ID already exists

└InvalidRequestError- Request parameters are invalid

└AuthenticationError- API key is invalid or missing

└NotFoundError- Requested resource does not exist

└RateLimitError- Too many requests (has retry_after attribute)

└NetworkError- Network connectivity issue

Error Handling Example

from bearlumen import (
    Client,
    InvalidEmailError,
    RateLimitError,
    AuthenticationError,
    BearLumenError,
)

try:
    result = client.end_users.create(
        external_id='user_123',
        name='AI Startup Inc',
    )
except InvalidEmailError:
    print('Invalid email format')
except RateLimitError as e:
    print(f'Rate limit hit, retry after {e.retry_after} seconds')
except AuthenticationError:
    print('Invalid API key')
except BearLumenError as e:
    print(f'Error: {e.message}')

Import Exceptions

from bearlumen import (
    Client,
    InvalidEmailError,
    DuplicateCustomerError,
    InvalidRequestError,
    AuthenticationError,
    NotFoundError,
    RateLimitError,
    NetworkError,
    BearLumenError,
)

Type Hints

The SDK includes full type hints for better IDE support:

from bearlumen import Client, EndUser, Subscription, UsageRecordInput, CostBreakdownResponse

client: Client = Client(api_key='your_key')

Available Types

ClientEndUserSubscriptionUsageRecordInputCostBreakdownResponseBearLumenErrorInvalidEmailErrorDuplicateCustomerErrorInvalidRequestErrorAuthenticationErrorNotFoundErrorRateLimitErrorNetworkError

Requirements

Python >= 3.9