Kreatyw - API Documentation

Overview

The Kreatyw API provides a simple REST endpoint for generating text continuations using N-gram language models. The API uses Markov chains trained on source texts to predict and generate coherent text sequences.

Base URL

http://localhost:8000

Endpoints

POST /api/predict

Generate text continuation based on a given prompt.

Request Body

Parameter	Type	Required	Description
prompt	string	Required	The starting text to continue from
n	integer	Optional	N-gram size (2-5). Default: 4. Higher values produce more coherent but less creative text.
temperature	float	Optional	Sampling temperature (0.1-2.0). Default: 1.6. Higher values increase randomness.
length	integer	Optional	Number of words to generate (1-500). Default: 5.

Response

Field	Type	Description
prediction	string	The generated text continuation

Example Request

curl -X POST http://localhost:8000/api/predict \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Once upon a time",
    "n": 4,
    "temperature": 1.2,
    "length": 20
  }'

Example Response

{
  "prediction": "in a kingdom far away there lived a brave knight who sought adventure..."
}

JavaScript Example

const response = await fetch('/api/predict', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    prompt: 'The old wizard',
    n: 3,
    temperature: 0.8,
    length: 15
  })
});

const data = await response.json();
console.log(data.prediction);

Python Example

import requests

response = requests.post('http://localhost:8000/api/predict',
    json={
        'prompt': 'In the beginning',
        'n': 4,
        'temperature': 1.0,
        'length': 25
    }
)

result = response.json()
print(result['prediction'])

Model Parameters

N-gram Size (n)

Controls the context window size. Higher values use more context words to predict the next word:

n=2 (Bigram): Uses 1 previous word for context. Very creative but less coherent.
n=3 (Trigram): Uses 2 previous words. Balanced creativity and coherence.
n=4 (Tetragram): Uses 3 previous words. More coherent, less random.
n=5 (Pentagram): Uses 4 previous words. Most coherent, closest to training data.

Temperature

Controls the randomness of predictions:

Low (0.1-0.5): More deterministic, picks most likely words.
Medium (0.6-1.0): Balanced between predictability and creativity.
High (1.1-2.0): More random and creative, may produce unexpected results.

Error Handling

The API returns standard HTTP status codes:

200 OK: Request successful
422 Unprocessable Entity: Invalid request parameters
500 Internal Server Error: Server error

Rate Limits

Currently, there are no rate limits imposed on the API. For production use, consider implementing appropriate rate limiting.