root-signals 1.0.0

The Python SDK for API of Root Signals

root-sdk

Robust AI Skill Management

This SDK provides a way to use Root Signals cloud service, which helps you evaluate your LLM pipelines. There are existing tools and frameworks that help you build these pipelines but evaluating it and quantifying your pipeline performance can be hard. This is where this comes in.

Root Signals provides you with the tools based on the latest research for evaluating LLM-generated text to give you insights about your LLM pipeline. It can be integrated with your CI/CD to provide continuous checks to ensure performance stays within acceptable limits.

Install

From pypi

The preferrable way of installing the SDK is from PyPI, using pip or other package installation of tool of choice. If using pip, this should be enough:

pip install root-signals

Quickstart

Please set your API key to environment variable ROOTSIGNALS_API_KEY, or to local .env file.

Retrieve an API key from https://app.rootsignals.ai/settings/api-keys

For example:

export ROOTSIGNALS_API_KEY=somethingreallysecretyougotfromtheweb

or, if you prefer using .env file:

echo ROOTSIGNALS_API_KEY=somethingreallysecretyougotfromtheweb >> .env

Minimal skill

{::comment}examples/minimal.py{:/comment}

from root import RootSignals

# Connect to the Root Signals API
client = RootSignals()

# Create a skill
skill = client.skills.create(
    "Find me good recipes for {{food_type}} food that are {{cuisine}}."
)

# Run it
response = skill.run({"food_type": "spicy", "cuisine": "Korean"})

print(response)

# llm_output="1. Kimchi Jjigae (Kimchi Stew): This spicy and flavorful
# stew is made with fermented kimchi, pork, tofu, and vegetables. ..."
# validation={'validator_results': [], 'is_valid': True}
# model='gpt-3.5-turbo' engine='gpt-3.5-turbo'
# execution_log_id='181eb95b-b972-4e96-8e30-ca7d3447d4fe'
# rendered_prompt='Find me good recipes for spicy food that are
# Korean.'

Documentation

For more details, please see the main SDK documentation.

Miscellaneous notes

Versioning policy

We follow semantic versioning; to point, major versions are not guaranteed to be backwards compatible, minor versions are, and patch versions only fix bugs.