Welcome to Query Vary's comprehensive guide to our prompt evaluation resources!

Query Vary makes it easy for developers to fine-tune large language models (LLMs) and prepare them for production.

This guide provides links to all of Query Vary's educational resources on prompt evaluation with Claude, OpenAI and other models. The linked materials cover core concepts, best practices, debugging checklists, prompt examples, and interactive materials. Whether you're new to prompt engineering or an experienced prompter, these resources serve to help you craft better prompts.

Follow this documentation and we promise that by the end of it, you will create your very own personal responsive system.

🚦Evaluate LLM output

Query Vary gives developers the tools to design, test, and refine prompts in a systematic way. Ensuring reliability, reducing latency, and optimizing costs. This is the process flow.

• Divide and Conquer: Many approaches to solve non-linear problems, involve reducing the problem a smaller state, where linearity appear. We do that with templates. Templates allow users to concentrate uniquely on aspects of the response that might get affected by additional constraints or rules. For example, a we do not want to prescribe medicine without having an assertive diagnosis.
• System Testing: Inside templates, the first step is to describe your variables! These variables operate as environment rules, guardrails and constraints. They can be introduced and tested into a finalized structured text, modified and fine-tuned solely by the developer.
• Prompt User Tests: The most tedious part. But don't worry, we have you covered. By auto-generating variations of prompts in an excel-style format, we reduce plenty of your valuable time. We also know these variations aren't always perfect, so you can tweak them here and there. Once ready, we also allow the generation of responses (you can select which model to utilize) and through our verification form, you will be able to determine if the answers are up to the standard. What's more, is that you will be able to compare answers from different models, enabling you gather a better picture of what model to utilize for your end product.
• Add Snippets: The coolest thing about Query Vary is you can add and integrate the finalized model right to your app, by simply copying the snippet. Trust us, it's magical!

Prompt Engineering Documentation

This comprehensive guide provides valuable insights and techniques for troubleshooting, prompt formatting, and effectively communicating with Claude, Anthropic's advanced AI model. To navigate through the documentation, simply use the menu on the left-hand side to access different sections.

Here are some curated links that you may find useful:

• Configuring GPT prompts for Claude: This guide offers step-by-step instructions on tailoring prompts specifically designed for our powerful GPT models, ensuring optimal performance with Claude.
• Introduction to prompt design: If you prefer a more concise overview of prompt engineering techniques, this section provides essential information along with practical examples to help you understand the concepts better.
• Constructing a prompt: This section offers a high-level overview of various intermediate techniques for prompt construction. While not exhaustive, it includes inline examples that illustrate the described concepts.
• Optimizing your prompt: Discover how to approach and implement an iterative process for evaluating and refining your prompts. This section provides valuable insights on enhancing prompt effectiveness.

Thank you for choosing Query Vary. We hope this documentation empowers you to make the most out of our AI model and achieve your desired results.

📈 Know your users

One of the best ways to know if you're nailing the development experience is checking out how your users are interacting with your app.

• Testing Performance Metrics Quality of response, latency and number of tokens (cost), are the key measurements we want to have ready for you.
• API Production Metrics These metrics are the essential ones for your work. Discover how much you are spending on solving key problems. Determine if the templates are effective, even if they passed testing.

💬 We're here to help!

Query Vary has many ways to make your apps behave seamlessly and become the envy of any AI bot. If you get stuck, shoot us an email or use the Intercom widget on the bottom right of any page to talk to us.

We're excited to have you onboard!