Kerb Documentation

The complete toolkit for developers building LLM applications.

Built to drive production ML systems at ApX Machine Learning (apxml.com), available open source.

Overview

Simple

Advanced LLM techniques made simple. Clean, easy-to-use interfaces for complex operations.

Lightweight

Only install what you need. Kerb is modular, no unnecessary dependencies.

Compatible

Works with any LLM project. Kerb is a toolkit, not a framework. Use it alongside your existing stack.

Installation

# Install everything
pip install kerb[all]

# Or install specific modules
pip install kerb[generation] kerb[embeddings] kerb[evaluation]

Quick Start

from kerb.generation import generate, ModelName, LLMProvider
from kerb.prompt import render_template

# Generate with any provider, easy config change.
response = generate(
    "Explain quantum computing",
    model=ModelName.GPT_4O_MINI,
    provider=LLMProvider.OPENAI
)

print(f"Response: {response.content}")
print(f"Tokens: {response.usage.total_tokens}")
print(f"Cost: ${response.cost:.6f}")

Modules

Everything you need to build LLM applications.

Indices and tables