py-ai-toolkit¶

A streamlined toolkit for building AI-driven software with LLMs.

What is py-ai-toolkit?¶

py-ai-toolkit bundles together essential utilities and patterns for working with Large Language Models into a cohesive, easy-to-use package. Instead of piecing together different tools and libraries, you get a unified interface for:

Prompt management - Template-based prompting with Jinja2
LLM interactions - Structured responses via instructor
Response models - Dynamic Pydantic model manipulation
Workflow orchestration - DAG-based task execution with grafo

Why Use This Toolkit?¶

Building AI-driven applications requires juggling multiple concerns: prompt formatting, HTTP requests to LLM APIs, structured output parsing, and complex workflow management. This toolkit consolidates these needs into simple, composable methods.

Rather than managing each piece separately, you can focus on your application logic while the toolkit handles the plumbing.

Key Features¶

Simple LLM Calling¶

ait = PyAIToolkit(main_model_config=LLMConfig())
response = await ait.chat(template="./prompt.md", message="Hello!")

Structured Responses¶

class Purchase(BaseModel):
    product: str
    quantity: int

response = await ait.asend(
    response_model=Purchase,
    template="Extract purchase from: {{ message }}",
    message="I want 5 apples"
)

Validated Task Execution¶

result = await ait.run_task(
    template="Extract a purchase from: {{ message }}",
    response_model=Purchase,
    kwargs=dict(message="I want 5 apples"),
    config=SingleShotValidationConfig(
        issues=["The purchase matches the user's request"]
    )
)

Workflow Building¶

workflow = BaseWorkflow(ai_toolkit=ait, error_class=WorkflowError)
task_tree = await workflow.create_task_tree(
    template="...",
    response_model=MyModel,
    kwargs={...},
    config=ValidationConfig()
)

Quick Start¶

Install the package:

uv add py-ai-toolkit

Then jump into the Getting Started guide.

Philosophy¶

This toolkit embraces composition over complexity. Each component serves a clear purpose and works independently, while integrating seamlessly when combined. You're not locked into a framework - you can use as much or as little as you need.