Cookbook Overview

Practical recipes for building AI applications with ABV

This cookbook provides copy-paste recipes for common ABV patterns. Each recipe includes both Python and JavaScript/TypeScript examples.

What’s in the Cookbook

Tracing Basics

Spans, generations, decorators, and nested observations

Gateway Patterns

Multi-provider LLM routing, streaming, and cost tracking

Guardrails

Content validation, safety checks, and custom rules

Evaluations

Datasets, scoring, experiments, and LLM-as-judge

Integration Patterns

Combining tracing + guardrails + scoring in workflows

Production Patterns

Chatbots, RAG systems, agents, and error handling

Prerequisites

Python

pip install abvdev openai

import os
os.environ["ABV_API_KEY"] = "sk-abv-..."
os.environ["OPENAI_API_KEY"] = "sk-proj-..."

JavaScript

npm install @abvdev/client @abvdev/tracing @abvdev/otel openai

// .env
ABV_API_KEY=sk-abv-...
OPENAI_API_KEY=sk-proj-...

Quick Reference

Pattern	Use When
@observe / observe()	Tracing existing functions without modification
Context managers	Manual control over span lifecycle
Manual spans	Fine-grained control, custom nesting
Gateway	Unified LLM API across providers
Guardrails	Input/output validation
Scores	Quality measurement and feedback