argus-ai-sdk 0.2.0

Argus Python SDK - Instrumentation for AI-native observability

argus-python

Argus Python SDK — instrumentation for AI-native observability, monitoring, and security.

Installation

pip install argus-ai-sdk

With framework extras:

pip install argus-ai-sdk[fastapi]
pip install argus-ai-sdk[flask]

Quick Start

Initialize the SDK

import argus

argus.init(
    server_url="https://argus.example.com",
    api_key="your-api-key",
    service_name="my-service",
)

Framework Middleware

FastAPI:

from fastapi import FastAPI
from argus.middleware.fastapi import ArgusMiddleware

app = FastAPI()
app.add_middleware(ArgusMiddleware)

Flask:

from flask import Flask
from argus.middleware.flask import ArgusFlask

app = Flask(__name__)
ArgusFlask(app)

Trace Functions

from argus.decorators import trace

@trace("fetch_user")
async def fetch_user(user_id: str):
    # Automatically creates a span with timing and error capture
    ...

Capture Exceptions

try:
    risky_operation()
except Exception:
    argus.capture_exception()

Breadcrumbs

argus.add_breadcrumb(
    category="http",
    message="GET /api/users",
    data={"status_code": 200}
)

Custom Events

argus.event("payment_processed", {"amount": 99.99, "currency": "USD"})

Features

• Distributed Tracing — W3C Traceparent propagation across services
• Framework Middleware — FastAPI and Flask integrations
• Function Tracing — @trace decorator for sync and async functions
• Exception Capture — Automatic stack traces with breadcrumb context
• Breadcrumbs — Trail of events leading up to errors
• Custom Events — Send arbitrary telemetry data
• HTTP Instrumentation — Auto-instrument outgoing httpx requests
• Serverless Support — Auto-detects AWS Lambda, Vercel, GCP Functions, and more
• Non-blocking — Events are batched and flushed asynchronously

License

MIT — see LICENSE for details.