parolo 0.1.0

A version-controlled prompt system

Parolo

A version-controlled prompt system for LLM workflows.

Parolo is an elegant prompt versioning and templating system designed specifically for LLM workflows. It enables you to:

• Automatically version prompts: Each change creates a new version (e.g., v0001.txt, v0002.txt, etc.)
• Maintain organized prompt storage: Prompts are stored in a configurable location.

---

Installation

uv pip install parolo

Quick Start

from parolo import Prompt

# Create a simple prompt
Prompt.create(name="summarize", prompt="Summarize this in three bullet points.")

# Create another version with different content
Prompt.create(name="summarize", prompt="Provide a concise summary in bullet points.")

# List all versions for a prompt
versions = Prompt.list_versions("summarize")
print(versions)  # ['v0001.txt', 'v0002.txt']

Prompt.create("greet", prompt="Hello, world!", metadata={"author": "demo", "type": "greeting"})

# Get overview of all prompts
overview = Prompt.overview()
print(overview)  # {'summarize': 2, 'greet': 1}

Advanced Usage

from parolo import Prompt

# Configure custom storage location
Prompt.set_base_dir("./project-prompts")

# Example of a complex multiline prompt
EXAMPLE_PROMPT = """You are a senior code reviewer with expertise in Python, security, and best practices.

Your task is to review the following code and provide feedback on:

1. **Code Quality**
   - Style and readability
   - Variable naming conventions
   - Function structure and organization

2. **Security Issues**
   - Input validation
   - Potential vulnerabilities
   - Data handling concerns

3. **Performance**
   - Algorithmic efficiency
   - Memory usage
   - Scalability considerations

4. **Best Practices**
   - Error handling
   - Documentation
   - Testing considerations

Please provide your feedback in the following format:

## Summary
Brief overview of the code's overall quality.

## Issues Found
List specific issues with line numbers and severity levels.

## Recommendations
Actionable suggestions for improvement.

## Code Rating
Rate the code from 1-10 with justification.

---

Code to review:
{code}
"""

# Create prompts for different use cases
Prompt.create(
    name="code_review_template",
    prompt=EXAMPLE_PROMPT,
    metadata={"author": "team", "type": "code_review", "complexity": "high"}
)

# Retrieve and use the stored prompt using Prompt methods
latest_prompt = Prompt.get_prompt("code_review_template")

# Use the prompt with actual code
sample_code = """
def calculate_total(items):
    total = 0
    for item in items:
        total += item
    return total
"""

# Format the prompt with the code using the built-in method
formatted_prompt = Prompt.format_prompt("code_review_template", code=sample_code)
print(formatted_prompt)

# Or get a specific version
version_prompt = Prompt.get_prompt("code_review_template", version="v0001")

# Get and format a specific version
formatted_v1 = Prompt.format_prompt("code_review_template", version="v0001", code=sample_code)

# Example usage in an application
def review_code(code: str, version: str = "latest") -> str:
    """Review code using the stored prompt template"""
    prompt = Prompt.format_prompt("code_review_template", version=version, code=code)
    # Send to your LLM API here
    return prompt

# Use in your application
result = review_code(sample_code)

# Or use a specific version for reproducibility
result_v1 = review_code(sample_code, version="v0001")

Configuration

By default, prompts are stored in ~/.parolo/prompts/. You can customize this in several ways:

# Method 1: Using set_base_dir()
Prompt.set_base_dir("/custom/path")

# Method 2: Environment variable
# export PAROLO_HOME="/custom/path"

Metadata and Version History

Parolo stores rich metadata with each version, similar to git commits:

from parolo import Prompt

# Create prompt with custom metadata
Prompt.create(
    name="code_review",
    prompt="Review the following code for security issues:\n\n{code}",
    metadata={"author": "security_team", "category": "security", "priority": "high"}
)

# List versions with metadata
Prompt.list_versions("code_review", show_metadata=True)

# Show version history (like git log)
Prompt.log("code_review")

# Get detailed version information
Prompt.show_version_info("code_review", "v0001")

# Retrieve metadata programmatically
metadata = Prompt.get_metadata("code_review", "v0001")
print(metadata["hash"])  # SHA-256 hash
print(metadata["timestamp"])  # ISO format timestamp
print(metadata["metadata"])  # Custom metadata

File Structure with Metadata

~/.parolo/prompts/
├── code_review/
│   ├── latest.txt
│   └── versions/
│       ├── v0001.txt           # Prompt content
│       ├── v0001.json          # Metadata (hash, timestamp, etc.)
│       ├── v0002.txt
│       └── v0002.json

Each .json file contains:

• hash: SHA-256 hash of the prompt content
• timestamp: ISO format creation timestamp
• version: Version identifier (e.g., "v0001")
• size: Size in bytes
• line_count: Number of lines
• previous_hash: Hash of the previous version
• metadata: Custom metadata provided by user

Features

• Automatic Versioning: Only creates new versions when content actually changes (using SHA-256 hashing)
• Rich Metadata: Stores timestamp, hash, size, and custom metadata with each version
• Version History: Git-like log functionality to view version history
• Latest File: Always maintains a latest.txt file with the current version
• Organized Storage: Clean directory structure for easy browsing
• Simple API: Intuitive class-based interface with static methods
• Flexible Configuration: Customizable storage location