GitHub

Getting Started with PomlSDK.jl

Welcome to the PomlSDK.jl documentation! This guide will help you get started with creating structured prompts for LLMs using the POML standard.

Understanding POML

POML (Prompt Orchestration Markup Language) is a standard developed by Microsoft for creating structured prompts. It provides a way to define complex prompt structures with metadata, examples, and multi-modal content, moving beyond simple string-based prompts.

Basic Concepts

PomlSDK.jl implements the following key concepts from the POML standard:

  • Prompt: The root object that contains all prompt elements
  • Hierarchy: A tree structure where elements can be nested within each other
  • Tags: Semantic elements like role, task, example, etc.
  • Content: Text, tables, lists, and other data that goes inside tags
  • Metadata: Information about the prompt, its source, and context

Your First Prompt

Let's create a simple prompt that instructs an AI assistant:

using PomlSDK

# 1. Create a new Prompt object
p = Prompt()

# 2. Add a role (system instructions)
role_node = role(p, caption="System")
add_node!(p, role_node)  # Make role_node the current parent
add_text(p, "You are a helpful AI assistant specialized in data analysis.")
pop_node!(p)  # Return to root

# 3. Add a task for the AI
task_node = task(p, caption="Data Summary")
add_node!(p, task_node)
add_text(p, "Summarize the provided sales data.")
pop_node!(p)

# 4. Serialize to POML format
poml_string = dump_poml(p)
println(poml_string)

This will produce XML similar to:

<poml>
  <role caption="System">You are a helpful AI assistant specialized in data analysis.</role>
  <task caption="Data Summary">Summarize the provided sales data.</task>
</poml>

Understanding the Hierarchy

PomlSDK uses a stack-based approach to manage the prompt hierarchy:

  • add_node!(p, node) makes node the current parent for subsequent additions
  • pop_node!(p) moves back up one level in the hierarchy

This allows you to create nested structures:

# Create a task with examples
task_node = task(p, caption="Data Analysis")
add_node!(p, task_node)

# Add examples
example_set_node = example_set(p)
add_node!(p, example_set_node)

example_node = example(p)
add_node!(p, example_node)
# Add input/output structure
input_node = tag(p, "input")
add_node!(p, input_node)
add_text(p, "Q1 Sales: $1.2M, Q2 Sales: $1.5M")
pop_node!(p)

output_node = tag(p, "output")
add_node!(p, output_node)
add_text(p, "Sales increased by 25% from Q1 to Q2.")
pop_node!(p)
pop_node!(p)  # Pop example
pop_node!(p)  # Pop example_set
pop_node!(p)  # Pop task