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

# Core Concepts

> Understand the fundamental concepts and architecture of HITL.sh for human-in-the-loop AI workflows.

## Understanding HITL.sh

HITL.sh is a platform that adds human oversight to AI workflows. It acts as a bridge between your AI systems and human reviewers, ensuring quality, safety, and compliance in automated processes.

## Key Concepts

### Human-in-the-Loop (HITL)

Human-in-the-loop refers to a workflow where AI systems perform tasks but require human oversight for critical decisions. This approach combines the efficiency of automation with the judgment and expertise of humans.

**Benefits:**

* **Quality control**: Human reviewers catch errors and improve outputs
* **Risk management**: Critical decisions require human approval
* **Compliance**: Meet regulatory requirements for human oversight
* **Learning**: Collect feedback to improve AI performance

### Review Workflow

A review workflow in HITL.sh follows this pattern:

1. **AI generates content** - Your AI system creates output (text, images, decisions)
2. **Review request** - Send the output to HITL.sh for human review
3. **Human review** - Reviewers examine and approve/reject/modify content
4. **Result processing** - Your workflow receives the review result and continues

> [card] [Workflow Examples](https://docs.hitl.sh/examples)
>
> See real-world examples of review workflows in action.

## Architecture Components

### Review Templates

Templates define the structure of what gets reviewed and what input you collect from humans.

**Template elements:**

* **Fields**: Input areas for reviewers (text, checkboxes, file uploads)
* **Layout**: How the review interface is organized
* **Validation**: Rules for required fields and data formats
* **Styling**: Custom appearance and branding

> [card] [Workflow Designer](https://docs.hitl.sh/guides/workflow-designer.md)
>
> Learn how to create and customize review templates.

### Agent Inbox

The Agent Inbox is the central hub where reviewers access pending reviews.

**Features:**

* **Unified view**: All reviews in one place
* **Filtering**: Sort by status, assignee, priority
* **Bulk actions**: Process multiple reviews efficiently
* **Mobile optimized**: Review on any device

### Integration Methods

HITL.sh integrates with your existing systems through multiple channels:

> [card] [SDKs](https://docs.hitl.sh/integrations/overview.md)
>
> Python and TypeScript SDKs for direct integration

> [card] [Workflow Tools](https://docs.hitl.sh/integrations/n8n.md)
>
> Native nodes for n8n, Make.com, and other platforms

> [card] [HTTP API](https://docs.hitl.sh/api-reference/introduction.md)
>
> RESTful API for custom integrations

> [card] [Webhooks](https://docs.hitl.sh/api-reference/webhooks.md)
>
> Receive review results in real-time

## Review Types

### Content Review

Review AI-generated content before publication or distribution.

**Use cases:**

* Blog posts and articles
* Marketing copy
* Social media content
* Product descriptions

### Decision Approval

Approve AI-generated decisions or recommendations.

**Use cases:**

* Financial transactions
* Risk assessments
* Compliance decisions
* Strategic recommendations

### Quality Assurance

Ensure AI outputs meet quality standards.

**Use cases:**

* Code generation
* Data analysis
* Image generation
* Translation services

### Training Data

Collect human feedback to improve AI models.

**Use cases:**

* Model fine-tuning
* Dataset creation
* Performance evaluation
* Bias detection

## Workflow Patterns

### Simple Approval

The most basic pattern: AI generates content, human approves/rejects.

```
AI Generation → HITL Review → Human Decision → Continue/Stop
```

### Iterative Improvement

AI generates content, human provides feedback, AI regenerates.

```
AI Generation → HITL Review → Human Feedback → AI Regeneration → Final Review
```

### Multi-Step Review

Complex workflows with multiple review checkpoints.

```
AI Step 1 → Review 1 → AI Step 2 → Review 2 → AI Step 3 → Final Review
```

### Conditional Review

Only review content that meets certain criteria.

```
AI Generation → Quality Check → Conditional Review → Continue
```

## Security & Privacy

### Data Protection

* **Encryption**: All data encrypted in transit and at rest
* **Access control**: Role-based permissions for team members
* **Audit logs**: Complete history of all review activities
* **Compliance**: GDPR and CCPA compliant

### Integration Security

* **API keys**: Secure authentication for integrations
* **Webhook verification**: Signed payloads for security
* **Rate limiting**: Protection against abuse
* **IP restrictions**: Optional IP whitelisting

## Performance & Scalability

### Review Processing

* **Real-time updates**: Instant notifications and status changes
* **Batch processing**: Handle multiple reviews efficiently
* **Priority queuing**: Important reviews get processed first
* **Load balancing**: Automatic scaling based on demand

### Integration Performance

* **Low latency**: API responses in milliseconds
* **High throughput**: Handle thousands of reviews per minute
* **Reliability**: 99.9% uptime guarantee
* **Global distribution**: Low-latency access worldwide

## Best Practices

### Template Design

* **Keep it simple**: Focus on essential review criteria
* **Clear instructions**: Help reviewers understand what to look for
* **Efficient layout**: Minimize clicks and scrolling
* **Mobile first**: Ensure good experience on mobile devices

### Workflow Integration

* **Async processing**: Don't block your main workflow
* **Error handling**: Plan for review failures or timeouts
* **Fallback options**: Provide alternatives when reviews aren't available
* **Monitoring**: Track review completion times and success rates

### Team Management

* **Clear roles**: Define who reviews what
* **Training**: Help reviewers understand the process
* **Feedback loops**: Continuously improve the review process
* **Automation**: Automate routine review assignments

## Next Steps

> [card] [Quickstart Guide](https://docs.hitl.sh/quickstart.md)
>
> Get hands-on experience with your first review workflow.

> [card] [Integration Guide](https://docs.hitl.sh/integrations/overview.md)
>
> Learn how to integrate HITL.sh with your existing systems.

> [card] [Examples](https://docs.hitl.sh/examples)
>
> See real-world implementations and use cases.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.