repello-argus-client 0.1.0

The official Python SDK for the Repello Argus API.

Repello Argus Python SDK

The official Python SDK for the Repello Argus AI Guardrails API.

The Argus Python SDK provides a robust, high-performance interface for integrating Repello Argus into your Python applications. It allows developers to scan LLM prompts and responses in real-time to detect and mitigate a wide range of security, safety, and content risks.

This SDK is designed for both rapid prototyping with code-defined policies and enterprise-grade orchestration using the Argus Platform, enabling features like centralized policy management, detailed analytics, and conversation tracking.

---

Key Features

• Comprehensive Security Policies: Scan for PII, prompt injection, toxicity, secrets, banned topics, unsafe content, and more.
• Flexible Policy Management: Define policies directly in your code for rapid development, or leverage the Argus Platform to manage policies centrally in the UI without code changes.
• Full Platform Observability: Use Assets and Sessions to organize and track your data, providing deep insights into application usage and conversation context in the Argus dashboard.
• Developer-First Experience: Built with type-safe enums (Verdict, Action, etc.) and a clear exception hierarchy for robust and predictable integration.
• Resilient & Performant: Powered by httpx for modern, asynchronous-ready HTTP requests and tenacity for automatic, resilient retries on network failures.

Installation

Install the package from PyPI using pip:

pip install repello-argus-client

The SDK requires Python 3.8 or higher.

Quickstart

This example shows how to get started in under a minute. It defines a simple policy in code to block toxic prompts.

First, set your API key as an environment variable:

export ARGUS_API_KEY="your_api_key_here"

Then, run the following Python code:

# quickstart.py
import os
from repello_argus_client import ArgusClient, PolicyName, Action, Verdict

# 1. Load your API Key from the environment
API_KEY = os.environ.get("ARGUS_API_KEY")
if not API_KEY:
    raise ValueError("ARGUS_API_KEY environment variable not set.")

# 2. Define a simple policy in your code
my_policy = {
    PolicyName.TOXICITY: {"action": Action.BLOCK}
}

# 3. Create a client and perform a scan
argus_guard = None
try:
    argus_guard = ArgusClient.create(api_key=API_KEY, policy=my_policy)
    result = argus_guard.check_prompt("You are a piece of garbage.")

    # 4. Handle the result using the Verdict enum
    print(f"Verdict: {result['verdict']}")

    if result['verdict'] == Verdict.BLOCKED:
        print("✅ The prompt was blocked. Application should halt this request.")
    else:
        print("The prompt was allowed to proceed.")

finally:
    if argus_guard:
        argus_guard.close()

Documentation

For a comprehensive guide covering all features, advanced usage patterns, and a full API reference, please see our official documentation:

docs.repello.ai/argus/sdk/python (Placeholder URL)

Usage Example: Platform Workflow

Platform users (with a Runtime Security Key) can leverage policies managed directly in the Argus UI. This simplifies the code significantly.

# platform_usage.py
import os
from repello_argus_client import ArgusClient, Verdict

# This workflow assumes you have:
# 1. A Runtime Security Key (rsk_...).
# 2. An Asset created in the Argus UI with its own policy.
API_KEY = os.environ.get("ARGUS_RUNTIME_API_KEY")
ASSET_ID = "your_asset_id_from_the_dashboard"

argus_guard = None
try:
    # Connect to your asset and enable saving results to the platform.
    # Notice no local `policy` object is needed.
    argus_guard = ArgusClient.create(
        api_key=API_KEY,
        asset_id=ASSET_ID,
        save=True
    )

    # This scan will be evaluated against the policy configured in the Argus UI.
    result = argus_guard.check_prompt("This prompt will be checked against the UI policy.")

    print(f"Verdict from platform policy: {result['verdict']}")

finally:
    if argus_guard:
        argus_guard.close()

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

We welcome contributions! Please see our CONTRIBUTING.md for guidelines on how to submit pull requests, report issues, and suggest features.

Support

If you encounter any issues or have questions, please file an issue on our GitHub Issues page or contact our support team at [e-mail].