Documentation
Python

Python SDK

rail-score-sdkv2.4.0 on PyPI

The official Python client for the RAIL Score API. Evaluate AI outputs across 8 responsibility dimensions with sync/async clients, agent evaluation, provider integrations, session tracking, middleware, and policy enforcement.

Quick Start — 30 seconds

pip install rail-score-sdk

from rail_score_sdk import RailScoreClient

client = RailScoreClient(api_key="YOUR_RAIL_API_KEY")
result = client.eval(content="Your AI-generated text here...", mode="basic")

print(result.rail_score.score)        # 8.4
print(result.dimension_scores)        # {fairness: 9.0, safety: 8.5, ...}

Migration note: This package replaces the deprecated rail-score package. Import path changed from rail_score.RailScore to rail_score_sdk.RailScoreClient.

Installation

Base package

pip install rail-score-sdk

With provider integrations

pip install "rail-score-sdk[openai]"        # OpenAI wrapper
pip install "rail-score-sdk[anthropic]"     # Anthropic wrapper
pip install "rail-score-sdk[google]"        # Google Gemini wrapper
pip install "rail-score-sdk[langfuse]"      # Langfuse observability
pip install "rail-score-sdk[litellm]"       # LiteLLM guardrail
pip install "rail-score-sdk[integrations]"  # All LLM provider wrappers

pip install "rail-score-sdk[agents]"        # CrewAI, LangGraph, AutoGen
pip install "rail-score-sdk[telemetry]"     # OpenTelemetry instrumentation

Requires: Python 3.8+

Core dependencies: httpx, requests

Client Initialization

Synchronous
from rail_score_sdk import RailScoreClient

client = RailScoreClient(api_key="YOUR_RAIL_API_KEY")

# Optional parameters
client = RailScoreClient(
    api_key="YOUR_RAIL_API_KEY",
    base_url="https://api.responsibleailabs.ai",  # default
    timeout=30,
)
Async
from rail_score_sdk import AsyncRAILClient
import asyncio

async def evaluate_batch():
    async with AsyncRAILClient(api_key="YOUR_RAIL_API_KEY") as client:
        results = await asyncio.gather(
            client.eval(content="First response...",  mode="basic"),
            client.eval(content="Second response...", mode="deep"),
        )
        for r in results:
            print(f"Score: {r['rail_score']['score']}")  # AsyncRAILClient returns raw dicts

asyncio.run(evaluate_batch())

Tip: AsyncRAILClient returns raw dicts — use result["rail_score"]["score"]. The sync RailScoreClient returns typed dataclasses.

In this SDK

Evaluation

client.eval() — basic, deep, selective dimensions, custom weights

Safe Regeneration

client.safe_regenerate() — iterate until quality thresholds are met

Compliance

client.compliance_check() — GDPR, HIPAA, EU AI Act, and more

Sessions & Policy

RAILSession, Policy enum, RAILMiddleware, PolicyEngine, error handling

Integrations

RAILOpenAI, RAILAnthropic, RAILGemini, RAILLangfuse, CrewAI, LangGraph, AutoGen

Agent Evaluation

client.agent — evaluate tool calls, scan results, detect injection, track session risk

Fairness
fairness
Safety
safety
Reliability
reliability
Transparency
transparency
Privacy
privacy
Accountability
accountability
Inclusivity
inclusivity
User Impact
user_impact

Each dimension scores 0–10. Privacy scores exactly 5.0 when not applicable. See the RAIL Framework page for full anchor descriptions.

Start with Evaluation →View on PyPI