promptcraft-py 0.2.0

A lightweight toolkit for prompt engineering, optimization, and evaluation

馃幆 PromptCraft

A lightweight Python toolkit for prompt engineering, optimization, and evaluation.

PromptCraft helps developers build, test, and refine prompts for LLMs with a clean, composable API.

鉁?Features

• Prompt Builder 鈥?Chain instructions, examples, and variables with a fluent API
• Template System 鈥?Reusable prompt templates with variable interpolation
• Optimizer 鈥?Automatically refine prompts using scoring and iteration
• Evaluator 鈥?Score prompt quality across clarity, specificity, and completeness
• Multi-provider 鈥?Works with OpenAI, Anthropic, and any OpenAI-compatible API
• Zero dependencies 鈥?Core library has no external dependencies

馃殌 Quick Start

pip install promptcraft-py

from promptcraft import PromptBuilder, Template

# Build a prompt
prompt = (
    PromptBuilder()
    .system("You are a helpful coding assistant.")
    .instruction("Explain the following concept clearly.")
    .variable("concept", "dependency injection")
    .examples([
        {"input": "What is recursion?", "output": "Recursion is when a function calls itself..."}
    ])
    .build()
)

print(prompt)

Using Templates

from promptcraft import Template

# Define a reusable template
review_template = Template(
    name="code_review",
    system="You are a senior code reviewer.",
    instruction="Review the following {language} code for bugs and improvements.",
    variables=["language", "code"]
)

# Render with specific values
prompt = review_template.render(
    language="Python",
    code="def add(a, b): return a"
)

Optimizing Prompts

from promptcraft import Optimizer

optimizer = Optimizer(
    metric="relevance",
    iterations=5
)

optimized = optimizer.optimize(
    prompt="Explain {topic}",
    test_cases=[
        {"topic": "quantum computing", "expected_keywords": ["qubit", "superposition"]},
        {"topic": "machine learning", "expected_keywords": ["model", "training"]},
    ]
)

print(f"Score improved from {optimized.initial_score:.2f} to {optimized.final_score:.2f}")

Evaluating Prompts

from promptcraft import Evaluator

evaluator = Evaluator()
result = evaluator.score(
    prompt="Write a function that sorts a list",
    criteria=["clarity", "specificity", "completeness"]
)

print(f"Overall: {result.overall:.2f}")
print(f"Clarity: {result.scores['clarity']:.2f}")

馃摝 Installation

# From PyPI
pip install promptcraft-py

# From source
git clone https://github.com/USER/PromptCraft.git
cd PromptCraft
pip install -e .

馃И Running Tests

# Run all tests
pytest

# With coverage
pytest --cov=promptcraft --cov-report=term-missing

馃摎 Documentation

• Getting Started
• API Reference
• Examples
• Contributing

馃 Contributing

Contributions are welcome! Please read CONTRIBUTING.md for details.

馃搫 License

This project is licensed under the MIT License 鈥?see LICENSE for details.