Helicone's guides for building, optimizing, and analyzing LLM applications with Helicone.

​
How-to Guides

Practical, task-oriented guides for solving specific problems with Helicone. These guides assume you already know the basics and need to accomplish a particular task.

​
Data Management & Analytics

Debug your LLM app

Identify and fix errors in your LLM application using Helicone’s debugging tools.

ETL / Data extraction

Extract and export your LLM data for analysis and reporting.

Segment data with Custom Properties

Track costs and behaviors by environment, user type, and custom dimensions.

Label request data

Add labels to requests for easier searching and filtering.

Get user requests

Retrieve user-specific requests for monitoring and cost tracking.

Get session data

Access conversation threads and session history.

​
Advanced Features

Replay LLM sessions

Replay and analyze past LLM sessions for optimization.

Run experiments

A/B test prompts and model configurations.

Fine-tune models

Prepare datasets and track fine-tuning workflows.

Predefine request IDs

Set custom request IDs for better tracking.

​
Integration & Environment

Track environments

Separate dev, staging, and production environments.

GitHub Actions integration

Monitor LLM calls in your CI/CD pipelines.

Manual logger streaming

Implement custom streaming with the logger SDK.

​
Tutorials

Step-by-step guides for learning by building complete applications with Helicone. Perfect for understanding how different features work together.

Build an AI Agent System

Create a complete AI agent with tool calling, memory, and observability.

Customer Support Assistant

Build a multi-model assistant that routes queries based on complexity.

AI Debate Simulator

Create an interactive debate app showcasing different integration methods.

Evaluation System with Ragas

Implement comprehensive LLM evaluation using Helicone and Ragas.

Chatbot with Structured Outputs

Build a chatbot using OpenAI’s structured outputs and function calling.

Thinking Models Implementation

Work with reasoning models like DeepSeek R1 and OpenAI o1/o3.

​
Knowledge Base

Educational resources to deepen your understanding of LLM concepts and best practices.

​
Prompt Engineering

Master the art of crafting effective prompts for optimal LLM performance.

Overview

Learn the basics of prompt engineering and how to craft effective LLM prompts.

Prompt thinking models

Learn how to effectively prompt thinking models like DeepSeek R1 and OpenAI o1/o3.

Be specific and clear

Create concise prompts for better LLM responses.

Use structured formats

Format the generated output for easier parsing and interpretation.

Role-playing

Assign specific roles in system prompts to set the style, tone, and content.

Few-shot learning

Provide examples of desired outputs to guide the LLM towards better responses.

Use constrained outputs

Set clear rules for the model’s responses to improve accuracy and consistency.

Chain-of-thought prompting

Encourage the model to generate intermediate reasoning steps before arriving at a final answer.

Thread-of-thought prompting

Build on previous ideas to maintain a coherent line of reasoning between interactions.

Least-to-most prompting

Break down complex problems into smaller parts, gradually increasing in complexity.

Meta-Prompting

Use LLMs to create and refine prompts dynamically.