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

# Structured Outputs

In Lyzr, **Structured Outputs** allow an agent’s response to be returned in a **machine-readable format** such as JSON, YAML, or even tabular data. These are crucial when your agent needs to interact with other systems, APIs, databases, or dashboards.

Instead of returning free-form text — which is flexible but difficult to extract meaning from — structured outputs make the agent’s responses actionable, testable, and easier to plug into production-grade workflows.

***

## ✨ Why Structured Outputs Matter

### ✅ Machine Interoperability

Structured outputs can be directly consumed by APIs, connected to automation tools like Zapier, integrated into frontend dashboards, or stored in structured databases like PostgreSQL or MongoDB.

**Example:**
Imagine a marketing agent returning campaign stats:

```json theme={null}
{
  "campaign_name": "July Blast",
  "clicks": 4231,
  "conversions": 372,
  "budget_remaining": 129.5
}
```

This response can be fed into an analytics dashboard or automation rule.

### ✅ Predictable Parsing

Instead of using regular expressions or NLP techniques to extract meaning from agent output, you always know what structure to expect. This drastically reduces the complexity of downstream logic.

### ✅ Enhanced UI Rendering

You can automatically render JSON objects as tables, charts, forms, or timelines in the UI — useful for dashboards, forms, reports, and more.

### ✅ Validation & Error Detection

You can define strict schemas for agent responses — including required fields, data types, and constraints. This enables robust validation and early detection of hallucinations or formatting mistakes.

***

## 🧠 Use Cases with Real Examples

### 🎯 Lead Scoring Agent

```json theme={null}
{
  "lead_name": "Jane Doe",
  "lead_source": "LinkedIn",
  "score": 91,
  "confidence": "High"
}
```

### 🛠️ Tech Support Ticket Generator

```json theme={null}
{
  "issue": "Password reset not working",
  "urgency": "High",
  "customer_id": "CUST-29387",
  "assigned_to": "tech-support@xyz.com"
}
```

### 📬 Email Generator for Sales Outreach

```json theme={null}
{
  "subject": "Re: Interest in Data Platform",
  "body": "Hi Mark,\nJust following up on our previous conversation...",
  "call_to_action": "Schedule a demo"
}
```

### 📊 Data Summary Agent

```json theme={null}
{
  "summary": "Revenue increased by 15% QoQ",
  "top_regions": ["India", "US", "UK"],
  "growth_drivers": ["Product X", "Campaign Y"]
}
```

***

## 🏗️ How to Use Structured Outputs in Lyzr Studio

### 1. Agent Configuration

When creating or editing an agent in Lyzr Studio:

* Scroll to the **Expected Output Format** section.
* Toggle **Structured Output** to `ON`.

### 2. Define Output Structure

Specify a sample output format or a JSON schema.

**Example Output Format:**

```json theme={null}
{
  "name": "string",
  "email": "string",
  "score": "integer",
  "comments": "string"
}
```

### 3. Prompt Engineering

Embed structure instructions directly in your prompt:

```
Return the result in the following JSON format. Do not include any explanation or markdown:
{
  "issue": "<string>",
  "priority": "<Low|Medium|High>",
  "assignee": "<email>"
}
```

### 4. Enable Validation (Optional)

You can enable schema validation within Lyzr to:

* Reject malformed responses
* Retry generation
* Trigger fallback behavior or alerts

***

## ✅ Best Practices

| Tip                | Description                                                           |
| ------------------ | --------------------------------------------------------------------- |
| 💬 Be explicit     | Tell the agent exactly what structure to follow, with a JSON snippet. |
| 🔄 Add retries     | On schema mismatch, set up retries with more specific prompts.        |
| ✂️ Avoid overkill  | Only include necessary fields — don’t make structures too verbose.    |
| 🧪 Test edge cases | Try inputs like nulls, large numbers, or ambiguous prompts.           |
| ⚙️ Define schemas  | Use JSON Schema for validation when needed.                           |

***

## 🚫 When *Not* to Use Structured Outputs

* When you're building **conversational**, **storytelling**, or **creative** agents.
* When the output is unpredictable, such as philosophical insights or user interviews.
* When human readers are the only consumers of the agent’s response.

Use free-form text in those cases.

***

## 📌 Summary

| Attribute       | Value                                                     |
| --------------- | --------------------------------------------------------- |
| Format          | JSON, YAML, key-value pairs, tables                       |
| Ideal for       | Automation, dashboards, pipelines, data pipelines         |
| Setup           | Agent creation > Enable Structured Output                 |
| Prompting Style | Strict instruction + Output structure                     |
| Integration Use | High — works with webhooks, APIs, databases, frontend UIs |

***

## 🔚 Final Note

By using Structured Outputs in your Lyzr agents, you're enabling automation, reducing complexity, and building reliable integrations. Treat your agents not just as conversational entities — but as programmable, structured interfaces into your systems.