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

# Quality Scoring

> Get a confidence score with every response. Auto-retry if it's too low.

Add `include_quality: true` to get a heuristic confidence score (0–1) with every response. Combine with `min_quality` to automatically retry with a better model when the score is too low.

## Get a quality score

```json theme={null}
{
  "model": "gpt-5",
  "messages": [{"role": "user", "content": "Write a poem about the ocean"}],
  "include_quality": true
}
```

```json theme={null}
{
  "choices": [{"message": {"role": "assistant", "content": "..."}}],
  "quality": {
    "confidence": 0.94,
    "flags": [],
    "suggested_retry": false
  }
}
```

## Auto-retry on low quality

Set `min_quality` to a threshold. If the response scores below it, the system automatically retries with a higher-quality model.

```json theme={null}
{
  "model": "gpt-5-mini",
  "messages": [{"role": "user", "content": "Explain the Riemann hypothesis"}],
  "include_quality": true,
  "min_quality": 0.8
}
```

If the initial response scores below 0.8, the response includes:

```json theme={null}
{
  "quality": {"confidence": 0.91},
  "auto_retry": {
    "triggered": true,
    "original_model": "gpt-5-mini",
    "retry_model": "gpt-5",
    "original_quality": 0.52
  }
}
```

## Parameters

| Parameter         | Type    | Default | Description                                                               |
| ----------------- | ------- | ------- | ------------------------------------------------------------------------- |
| `include_quality` | boolean | `false` | Include a quality score in the response.                                  |
| `min_quality`     | number  | —       | If the score falls below this threshold (0–1), retry with a better model. |

## Billing

You're charged for each attempt. If auto-retry triggers, you pay for both the original and the retry.