> ## Documentation Index
> Fetch the complete documentation index at: https://docs.responsibleailabs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Concepts Overview

> Core concepts behind the RAIL Score platform.

RAIL Score is organized into three areas: the **core concepts** you score with, the **governance** layer that decides what to do with a score, and **specialized checks** for regulatory and agent use cases. Each concept has its own page, API reference, and SDK guide.

## Core concepts

<CardGroup cols={2}>
  <Card title="Evaluation" icon="magnifying-glass" href="/concepts/evaluation">
    Score content across 8 RAIL dimensions. The foundation of everything else.
  </Card>

  <Card title="RAIL Framework" icon="shield" href="/concepts/rail-framework">
    The 8 dimensions in detail, with score anchors and examples.
  </Card>

  <Card title="Safe Regeneration" icon="rotate" href="/concepts/safe-regeneration">
    Automatically regenerate content that falls below your quality threshold.
  </Card>
</CardGroup>

## Governance

<CardGroup cols={2}>
  <Card title="Applications" icon="cubes" href="/concepts/applications">
    How API keys, environments, and the policy that governs them fit together.
  </Card>

  <Card title="Policy Engine" icon="gavel" href="/concepts/policy-engine">
    Turn scores into actions: log, block, or regenerate below your thresholds.
  </Card>

  <Card title="Middleware" icon="layer-group" href="/concepts/middleware">
    Drop-in provider wrappers that score every LLM response automatically.
  </Card>
</CardGroup>

## Specialized checks

<CardGroup cols={2}>
  <Card title="Compliance" icon="scale-balanced" href="/concepts/compliance">
    Check content against GDPR, HIPAA, EU AI Act, CCPA, India DPDP, and more.
  </Card>

  <Card title="Agent Evaluation" icon="robot" href="/concepts/agent-evaluation">
    Intercept tool calls, scan results, and detect prompt injection in AI agents.
  </Card>
</CardGroup>

## How the pieces fit together

```mermaid theme={null}
flowchart TD
    Input["User prompt + LLM response"] --> Middleware["Middleware wrapper\n(optional — automatic eval)"]
    Middleware --> Eval["Evaluation\nbasic or deep mode"]
    Eval --> Policy["Policy Engine\n(block / warn / flag / allow)"]
    Policy --> Session["RAILSession\n(aggregate tracking)"]
    Eval --> Regen["Safe Regeneration\n(if below threshold)"]
    Eval --> Compliance["Compliance check\n(regulatory frameworks)"]
```

Start with [Evaluation](/concepts/evaluation): every other concept builds on it.