langspend 1.0.0

LangSpend Python SDK - Track LLM costs and usage with customer attribution. Requires LangSpend API key for production use.

LangSpend Python SDK

Track LLM costs and usage with customer attribution for OpenAI, Anthropic, AWS Bedrock, and Google Vertex AI.

Features

• Customer Attribution: Track costs per customer and feature
• Zero Code Changes: Wrap your existing LLM clients
• Multi-Provider Support: OpenAI, Anthropic, AWS Bedrock, Google Vertex AI
• Streaming Support: Works with streaming responses
• Async Support: Full async/await compatibility
• Error Resilience: Never breaks your LLM calls

Installation

Using pip

pip install langspend

Using uv (recommended)

uv add langspend

For specific providers

# Using pip
pip install langspend[openai]      # OpenAI support
pip install langspend[anthropic]   # Anthropic support
pip install langspend[bedrock]     # AWS Bedrock support
pip install langspend[vertex]      # Google Vertex AI support
pip install langspend[all]         # All providers

# Using uv
uv add "langspend[openai]"         # OpenAI support
uv add "langspend[anthropic]"      # Anthropic support
uv add "langspend[bedrock]"        # AWS Bedrock support
uv add "langspend[vertex]"         # Google Vertex AI support
uv add "langspend[all]"            # All providers

Get Your API Key

LangSpend is a commercial SaaS product with a free tier available.

Get your LangSpend API key at langspend.com - Sign up here

Note: An API key is required for production use. The SDK includes a free tier for development and evaluation.

Quick Start

Option 1: Customer and Feature Cost Tracking (Recommended)

from langspend import LangSpend, wrap_openai
from openai import OpenAI

# Initialize LangSpend
langspend = LangSpend(api_key="lsp_your_api_key_here")

# Wrap your OpenAI client
openai = wrap_openai(
    OpenAI(api_key="your_openai_api_key"),
    langspend
)

# Use as normal - tracking happens automatically
response = openai.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    langspend_tags={
        "customer": {
            "id": "user_123",
            "name": "John Doe",
            "email": "[e-mail]"
        },
        "feature": {
            "name": "chat-api"
        }
    }
)

Option 2: Zero Code Setup (Quick, no customer level tracking)

from langspend import LangSpend, wrap_openai
from openai import OpenAI

# Initialize LangSpend
langspend = LangSpend(api_key="lsp_your_api_key_here")

# Wrap your OpenAI client
openai = wrap_openai(OpenAI(), langspend)

# Use as normal - tracking happens automatically
response = openai.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}]
)

Supported Providers

LangSpend supports OpenAI, Anthropic, and AWS Bedrock. See the examples directory for integration examples with each provider.

Demo Application

A complete Next.js + FastAPI demo is available in examples/nextjs-chat-demo/:

cd examples/nextjs-chat-demo/backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt
cp env.example .env
# Edit .env with your API keys
python main.py

Examples

See the examples/ directory for comprehensive usage examples:

• basic_usage.py - Basic usage examples
• customer_attribution.py - Customer attribution examples
• anthropic_usage.py - Anthropic integration
• bedrock_usage.py - AWS Bedrock integration
• flask_integration.py - Flask web app integration
• fastapi_integration.py - FastAPI integration
• streaming_example.py - Streaming responses
• multi_provider.py - Multiple providers
• nextjs-chat-demo/ - Complete Next.js + FastAPI demo

Configuration

from langspend import LangSpend

langspend = LangSpend(
    api_key="lsp_your_api_key",
    base_url="https://platform.langspend.com",  # Optional
    debug=False,  # Enable debug logging
    max_retries=3  # Max retry attempts
)

Error Handling

The SDK never throws errors for tracking failures. All tracking happens asynchronously and failures are silent by default.

Enable debug mode to log tracking errors:

langspend = LangSpend(api_key="lsp_your_api_key", debug=True)

License

MIT License - see LICENSE file for details.