endurance-rai 1.0.0

Official Python SDK for Endurance RAI Engine

Endurance RAI SDK

Official Python SDK for Endurance RAI Engine - Responsible AI monitoring for government chatbots.

Installation

pip install endurance-rai

Or install from source:

git clone https://github.com/yourcompany/endurance-sdk.git
cd endurance-sdk
pip install -e .

Quick Start

import asyncio
from endurance import EnduranceClient, RAGDocument

async def main():
    # Initialize client
    client = EnduranceClient(
        base_url="https://lamaq-endurance-backend-4-hods.hf.space"
    )
    
    # Prepare RAG documents
    docs = [
        RAGDocument(
            source="FOI_Act_2000.pdf",
            content="Public authorities must respond within 20 working days...",
            page=1,
            similarity_score=0.95
        )
    ]
    
    # Evaluate chatbot response
    result = await client.evaluate(
        query="How long for FOI response?",
        response="Public authorities must respond within 20 working days.",
        service_id="uk_gov_chatbot",
        rag_documents=docs
    )
    
    # Check results
    print(f"Overall Score: {result.overall_score}/100")
    print(f"Flagged: {result.flagged}")
    print(f"Grounding: {result.dimensions.data_grounding}")

asyncio.run(main())

Features

• ✅ Async/await support for non-blocking evaluations
• ✅ Automatic retry logic with exponential backoff
• ✅ Type hints for IDE autocomplete
• ✅ Comprehensive error handling
• ✅ 9 RAI dimensions evaluated

API Reference

EnduranceClient

Main client for interacting with Endurance RAI Engine.

Parameters:

• base_url (str): Endurance backend URL
• api_key (str, optional): API key for authentication
• timeout (float, default=10.0): Request timeout in seconds
• max_retries (int, default=3): Maximum retry attempts

Methods:

• evaluate(): Evaluate chatbot response
• health_check(): Check service availability
• get_service_stats(): Get aggregate statistics

RAGDocument

Represents a retrieved document from RAG system.

Fields:

• source (str): Document source
• content (str): Document text
• page (int): Page number
• similarity_score (float): Retrieval score (0-1)

EvaluationResult

Result from RAI evaluation.

Fields:

• overall_score (float): Aggregate score (0-100)
• flagged (bool): Whether response requires review
• dimensions (DimensionScores): Individual dimension scores
• session_id (str): Unique session ID
• timestamp (str): Evaluation timestamp

Error Handling

from endurance import EnduranceClient, EnduranceError, RateLimitError

client = EnduranceClient()

try:
    result = await client.evaluate(...)
except RateLimitError:
    print("Rate limit exceeded")
except EnduranceError as e:
    print(f"Error: {e}")

License

MIT License

Support

• Documentation: https://endurance-rai.readthedocs.io
• Issues: https://github.com/yourcompany/endurance-sdk/issues
• Email: [e-mail]