prompt-to-json 0.1.1

Convert natural language prompts to structured JSON using OpenAI

prompt-to-json

Convert natural language prompts to structured JSON using OpenAI's GPT models.

Installation

pip install prompt-to-json

Quick Start

from prompt_to_json import PromptToJSON

# Initialize with API key
converter = PromptToJSON(api_key="sk-...")  # or set OPENAI_API_KEY env var

# Convert natural language to structured JSON
prompt = "Summarize this article in 3 bullet points professionally"
result = converter.convert(prompt)

print(result)
# Output:
# {
#   "task": "summarize",
#   "input_data": {"type": "article", "source": "provided"},
#   "output_format": {"format": "bullet_points", "count": 3},
#   "constraints": {"max_points": 3},
#   "config": {"tone": "professional"}
# }

Features

• 🎯 Simple API - Just one method: convert()
• 🧠 Intelligent Parsing - Uses GPT to understand intent and structure
• 📦 Structured Output - Returns clean JSON ready for downstream processing
• 🔄 Batch Processing - Convert multiple prompts at once
• ⚡ Production Ready - Error handling and fallbacks included

Usage

Basic Usage

from prompt_to_json import PromptToJSON

converter = PromptToJSON()

# Simple prompt
result = converter.convert("Extract key points from this text")

Batch Processing

prompts = [
    "Summarize this article",
    "Translate to Spanish",
    "Extract data points"
]

results = converter.convert_batch(prompts)

Using Different Models

# Use GPT-4.1 for better accuracy (default)
converter = PromptToJSON()

# Use GPT-3.5 Turbo for speed and cost efficiency
converter = PromptToJSON(model="gpt-3.5-turbo")

Output Structure

The converter extracts and structures:

• task - Main action/verb (summarize, extract, generate, etc.)
• input_data - Data or content to process
• output_format - Expected format and structure
• constraints - Limitations (length, count, style, etc.)
• context - Background information or purpose
• config - Settings like tone, approach, style

Examples

# Content Generation
prompt = "Write a marketing email for our new AI product"
# Returns: {"task": "generate", "input_data": {"type": "marketing_email"}, ...}

# Data Extraction
prompt = "Extract all dates and amounts from these invoices"
# Returns: {"task": "extract", "input_data": {"type": "invoices"}, ...}

# Analysis
prompt = "Analyze customer feedback and identify top complaints"
# Returns: {"task": "analyze", "input_data": {"type": "customer_feedback"}, ...}

Advanced Usage

Error Handling, Rate Limits, and Retries

from prompt_to_json import PromptToJSON
from openai import OpenAI, OpenAIError
from time import sleep

converter = PromptToJSON()

for attempt in range(3):
    try:
        result = converter.convert("Summarize this article")
        break
    except OpenAIError as e:
        # basic exponential backoff on failure or rate limit
        sleep(2 ** attempt)
else:
    raise RuntimeError("Request failed after retries")

print(result)

Timeout Configuration

converter = PromptToJSON()
# underlying OpenAI client supports a timeout parameter
converter.client = OpenAI(api_key=converter.api_key, timeout=30)

JSON Schema and Validation

from prompt_to_json import PromptJSON

result = converter.convert("Extract key facts")
PromptJSON.model_validate(result)  # raises if the shape is wrong
schema = PromptJSON.to_schema()

Custom Models and Prompts

# Use a different model and a deterministic temperature
converter = PromptToJSON(model="gpt-4o-mini")

# Modify system instructions or add few-shot examples
custom = PromptToJSON(model="gpt-4o-mini")
custom_prompt = "You are a precise JSON formatter."
response = custom.client.responses.create(
    model=custom.model,
    instructions=custom_prompt,
    input="Translate to French",
    temperature=0,
)

Requirements

• Python 3.6+
• OpenAI API key

Environment Setup

export OPENAI_API_KEY="sk-your-api-key-here"

License

MIT

Contributing

Contributions welcome! See CONTRIBUTING.md for details.

Changelog

See CHANGELOG.md for release history.

Support

For issues or questions, please open an issue on GitHub.