Key Concepts & Terminology
Understanding Empromptu's core concepts will help you build and optimize complete AI applications effectively.
What makes Empromptu different: Empromptu builds complete AI applications with model, RAG, and infrastructure included, going beyond what existing app builder tools offer. Unlike simple builders creating static websites or complex ML tools requiring specialized teams, Empromptu delivers production-ready AI applications with 90%+ accuracy through built-in optimization.
What you'll learn ⏱️ 4 minutes
How Projects, Tasks, and Optimization work together
What Prompt Families are and why they matter
How Empromptu's accuracy scoring works
Key platform terminology you'll see throughout the interface
Projects and Tasks
Projects are your AI applications - like a Review Summarizer or Customer Support Chatbot. Each project can contain multiple Tasks that handle specific functions within that application.
Example: A Customer Support project might have separate tasks for:
Handling billing questions
Processing refund requests
Routing technical issues
When you use the Builder to create an application, Empromptu automatically generates a project with LLMOps capabilities built-in. No additional setup required.
Accuracy Scoring
Empromptu uses a 0-10 scoring system to measure how well your AI applications perform:
0-3: Low Score (Red) - Needs improvement
4-6: Medium Score (Orange) - Getting better
7-8: Good Score (Blue) - Performing well
9-10: Excellent Score (Green) - Optimal performance
Initial vs Current Accuracy
Initial Accuracy: The first score when your task runs for the first time
Current Accuracy: Your task's performance after optimization
Improvement: Shows the progress from initial to current performance
Prompt Families
A Prompt Family is a collection of prompts that work together to perform better than any single prompt could. Start with one prompt, and Empromptu will help you to build out the whole family.
Here's how it works:
Start with one prompt for your task
Empromptu helps you build out additional prompts
The family of prompts collaborates to handle different scenarios and edge cases
Results in higher accuracy and more reliable responses
Think of it like having a team of specialists rather than one generalist handling everything.
AI Response Optimization Features
In order to get the AI to respond well and as you expect it, you may need to make adjustments to your inputs, evaluations and prompts.
Task Actions
From your project dashboard, each task has an Actions button that provides access to optimization tools:
Prompt Optimization: Fine-tune your prompts for better results
Input Optimization: Improve how your application handles user inputs
Model Optimization: Test different AI models to find the best fit
Edge Case Detection: Identify and resolve problematic scenarios
Evaluations: Set up criteria to measure success
Click the Actions button on any task to access these optimization capabilities.
Events
Events are created whenever there's an API call in your system. When your AI processes an input and generates an output, that creates an event. You can review these events to understand how your application is performing and where improvements can be made.
Evaluations
Evaluations define what "success" looks like for your AI application. You can:
Write specific criteria for what good performance of that task means to you in natural language. It's really that simple. Empromptu does all of the heavy lifting under the hood.
Examples of evaluation criteria:
"Correct Sequence" - Information should appear in logical order
"Extracted Complete Bug Set" - All bugs mentioned should appear in output
"Accurate Details" - Extracted details match the original text
Platform Flow
Here's how everything connects:
Builder → Describe your AI application in natural language (like Lovable or Bolt)
Project Created → Empromptu builds your complete AI application with model, RAG, and infrastructure included
Projects Tab → View your projects and access LLMOps capabilities
Tasks & Actions → Click Actions on any task to access optimization tools
Deployment → Launch your optimized application from the Builder interface
Next Steps
Now that you understand the core concepts, you're ready to:
Create your first AI application using the Builder
Learn about optimization workflows to improve performance
Explore deployment options to launch your application
Last updated