prompt-config 0.1.5

A dynamic prompt building library with CLI support.

prompt-config

prompt-config is a dynamic prompt building library designed for LLM applications. It allows you to define complex prompts in a structured YAML configuration, manage dynamic variables with Jinja2 templating, and construct prompts at runtime using a powerful tagging and filtering system.

Perfect for managing prompts as code, enabling environment-specific prompts, and ensuring reproducibility in your AI applications.

---

Features

• YAML Configuration: Centralize your prompt definitions in a clean, readable YAML file.
• Dynamic Variables: Use Jinja2 syntax ({{ variable }}) for dynamic content insertion.
• Tag-Based Composition: conditionally include or exclude prompt parts based on tags (e.g., dev, prod, python, web).
• Validation: Enforce required variables and provide default values.
• CLI Tool: Built-in promptcfg CLI for easy integration into scripts and CI/CD pipelines.
• Python API: seamless integration into your Python applications.

Installation

Requires Python 3.11 or higher.

pip install prompt-config

Quick Start

1. Initialize Configuration

Create a starter promptcfg.yaml file in your current directory:

promptcfg init

This will generate a file similar to:

version: "1.0"
prompts:
  - id: "system_intro"
    text: "You are a helpful AI assistant named {{ assistant_name }}."
    default: true
    variables:
      - name: "assistant_name"
        default: "Prompter"

  - id: "coding_capabilities"
    text: "You are an expert in {{ language }} programming."
    tags: ["coding"]
    variables:
      - name: "language"
        required: true

2. Build via CLI

Generate a prompt by activating specific tags and passing variables:

promptcfg build --tags coding --vars assistant_name=Jarvis,language=Python

Output:

--- Generated Prompt ---
You are a helpful AI assistant named Jarvis.
You are an expert in Python programming.
------------------------

3. Build via Python API

Integrate prompt-config directly into your application:

from promptcfg.config import PromptConfig
from promptcfg.builder import PromptBuilder

# Load the configuration
config = PromptConfig.load("promptcfg.yaml")

# Initialize the builder
builder = PromptBuilder(config)

# Build the prompt
prompt = builder.build(
    tags=["coding"],
    variables={
        "assistant_name": "Jarvis",
        "language": "Python"
    }
)

print(prompt)

CLI Usage

The promptcfg generic command line interface supports the following subcommands:

• init: Create a generic configuration file.
• build: Construct a prompt based on tags and variables.

Build Options

Option	Description
--config	Path to the YAML config file (default: promptcfg.yaml)
--tags	Comma-separated list of tags to include (e.g., dev,test)
--vars	Key-value pairs for variables (e.g., key=value,foo=bar)
--include-ids	Explicitly include specific prompt blocks by ID
--exclude-ids	Explicitly exclude specific prompt blocks by ID
--exclude-tags	Exclude prompt blocks matching these tags
--list-prompts	List all available prompt IDs and their status

# Example: Build with exclusions
promptcfg build --tags all --exclude-tags deprecated --vars user_id=123

Development

If you want to contribute or modify the library:

1. Clone the repository:

   git clone https://github.com/AKSHILMY/prompter.git
   cd prompter
   

2. Set up a virtual environment:

   python3.12 -m venv venv
   source venv/bin/activate
   

3. Install in editable mode:

   pip install -e "."
   

4. Run Tests:

   python -m unittest discover tests
   

License

This project is licensed under the terms of the Apache License 2.0.