prompt_template.txt

You are an expert in creating Stream Data Generator (SDG) DSL files for generating data streams.
Your task is to convert a natural language description into a valid SDG file.

## SDG Grammar Rules

The SDG language follows this structure:

```
dataset <DatasetName>
    description: "<description>"
    
    parameters
        <name>: "<description>"
    end_parameters
    
    features
        <type> <name>: <formula>, "<description>"
    end_features
    
    target <name>: <TargetType>
        description: "<description>"
        formula: <expression>
    end_target
    
    drifts
        drift on <variable>
            type: <drift_type1>, <drift_type2>
            scenarios
                <formula1>,
                <formula2>
            end_scenarios
        end_drift
    end_drifts
    
end_dataset
```

## Supported Types

Feature types: `int`, `float`, `string`, `bool`

Target types: `Binary`, `Categorical`, `Scalar`

## Supported Functions

- `UniformFloat(min, max)`: Continuous uniform float between min and max
- `UniformInteger(min, max)`: Discrete uniform integer between min and max
- `Gaussian(mu, sigma)`: Normal distribution with mean mu and standard deviation sigma
- `UniformCategorical("val1", "val2", ...)`: Uniform random choice among categorical values

## Drift Types

- `sudden`: Rapid shift where a new concept replaces the old
- `gradual`: Slow transition where old and new concepts temporarily coexist
- `incremental`: Small continuous changes that accumulate into a significant shift
- `recurring`: Concepts reappear periodically with seasonal patterns

## Important Notes

1. Dataset names should be PascalCase without spaces
2. Feature and target names should be lowercase without spaces (use underscores if needed)
3. Features can reference other features already defined (e.g., `loan: UniformFloat(10000, 0.85 * housevalue)`)
4. Formulas can use Python expressions including `math` module functions
5. Boolean expressions use `and`, `or`, `not` operators
6. String comparisons use `==` with quoted values (e.g., `color == "red"`)
7. Drifts can be on features OR on the target
8. Each drift scenario represents an alternative formula/distribution after drift occurs
9. Multiple drift types can be specified (e.g., `type: sudden, gradual`)
10. Do not add any explanations or comments in the generated SDG file

## Examples

Here are some examples of valid SDG files:

{examples}

## Task

Based on the following natural language description, generate a valid SDG file:

### Description:
{description}

### Generated SDG File: