proxle 0.0.1

Proxle SDK for Python - LLM observability and cost optimization

Proxle Python SDK

LLM observability and cost optimization SDK for Python. Drop-in replacements for OpenAI, Anthropic, Azure OpenAI, Cohere, and Gemini clients that route through the Proxle proxy for automatic request logging, smart caching, and cost tracking.

Installation

pip install proxle

Install with provider-specific dependencies:

# For OpenAI / Azure OpenAI support
pip install proxle[openai]

# For Anthropic support
pip install proxle[anthropic]

# For all providers
pip install proxle[all]

Cohere and Gemini clients use httpx (included by default) and don't require additional packages.

Quick Start

Configuration

Set your Proxle API key via environment variable or pass it directly:

export PROXLE_API_KEY=pk_live_your_key_here
export PROXLE_URL=https://api.proxle.dev  # optional, this is the default

OpenAI

from proxle import OpenAI

client = OpenAI(api_key="sk-your-openai-key", proxy_key="pk_live_...")

response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={"feature": "chat_assistant", "user_id": "user_123"}
)
print(response.choices[0].message.content)

Anthropic

from proxle import Anthropic

client = Anthropic(api_key="sk-ant-your-key", proxy_key="pk_live_...")

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={"feature": "summarizer"}
)
print(response.content[0].text)

Azure OpenAI

from proxle import AzureOpenAI

client = AzureOpenAI(
    api_key="your-azure-key",
    azure_endpoint="https://my-resource.openai.azure.com",
    api_version="2024-02-01",
    proxy_key="pk_live_...",
)

response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={"feature": "chat"}
)

Cohere

from proxle import Cohere

client = Cohere(api_key="your-cohere-key", proxy_key="pk_live_...")

response = client.chat(
    message="Hello!",
    model="command-r-plus",
    metadata={"feature": "chat"}
)
print(response["text"])

Gemini

from proxle import Gemini

client = Gemini(api_key="your-gemini-key", proxy_key="pk_live_...")

response = client.generate_content(
    model="gemini-1.5-pro",
    contents=[{"role": "user", "parts": [{"text": "Hello!"}]}],
    metadata={"feature": "chat"}
)
print(response["candidates"][0]["content"]["parts"][0]["text"])

Metadata

All provider methods accept an optional metadata parameter for cost attribution:

metadata = {
    "feature": "chat_assistant",  # Track cost by feature
    "user_id": "user_123",        # Track cost by user
    "environment": "production",  # Any custom tags
}

Metadata is sent as the X-Metadata header and stored with the request log in your Proxle dashboard.

Streaming

OpenAI Streaming

stream = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
    metadata={"feature": "chat"}
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Anthropic Streaming

with client.messages.stream(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={"feature": "chat"}
) as stream:
    for text in stream.text_stream:
        print(text, end="")

Cohere/Gemini Streaming

for event in cohere_client.chat(message="Hello!", stream=True, metadata={"feature": "chat"}):
    print(event.get("text", ""), end="")

Async Support

OpenAI, Anthropic, and Azure providers have async variants:

from proxle import AsyncOpenAI, AsyncAnthropic, AsyncAzureOpenAI

client = AsyncOpenAI(api_key="sk-...", proxy_key="pk_live_...")
response = await client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}],
    metadata={"feature": "chat"}
)

API Reference

Providers

Class	Wraps	Requires
OpenAI / AsyncOpenAI	openai.OpenAI	pip install proxle[openai]
Anthropic / AsyncAnthropic	anthropic.Anthropic	pip install proxle[anthropic]
AzureOpenAI / AsyncAzureOpenAI	openai.AzureOpenAI	pip install proxle[openai]
Cohere	httpx (built-in)	None
Gemini	httpx (built-in)	None

Constructor Parameters

All providers accept:

Parameter	Type	Description
api_key	str	Your provider API key
proxy_key	str (optional)	Proxle API key (or set PROXLE_API_KEY env var)
proxy_url	str (optional)	Proxle proxy URL (or set PROXLE_URL env var)

Azure additionally requires azure_endpoint and api_version.

Development

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

# Run tests
pytest

# Lint
ruff check .

License

MIT