Quickstart

Empromptu is an AI Accuracy Platform that helps you create and optimize your AI application to deliver 98% accuracy and reliability automatically.

We start optimizing your AI responses by refining the prompts that drive them. Instead of writing your prompts in plain text and tinkering with them, we use agents and AI to send your prompt to our prompt registry where we will send you a fully optimized prompt back in real-time with low latency.

Our goal is to ensure that for every user input into your system, you achieve at least 98% accurate outputs. This guide is designed to get you up and running quickly no matter if you are a hobbyist developer to the most sophisticated AI team.

Use this guide if you have an AI application

If you don't have an AI app and want to use our No-code UI, no problem, go here.

Step 0: Code Setup - New Environment

If you want to experiment with Empromptu in a new or demo environment, setting up that environment is easy. If you already have an application to install Empromptu's library you can skip to step 1.

1. First, install python 3.11 or higher::

sudo apt install python3.11 --upgrade

1. Start a new virtualenv (or whatever system you prefer):

virtualenv -p python3.11 venv && source venv/bin/activate

If you are setting up a new environment, check out our example code library here.

Step 1: Install Empromptu's Library

1. In your terminal:

pip install http://impromptu-packages.s3-website-us-west-2.amazonaws.com/impromptu-0.5.8-py3-none-any.whl

Step 2: Create an Empromptu account to generate your API Key

1. Go to the Empromptu Sign-Up Page → [Insert Sign-Up Link] and register using your email and create a password.

2. Head to the Settings page.

3. Copy your API key.

Step 3: Import Empromptu's Library

1. Using a code editor open the application where you want to use Empromptu

2. Find where you call your LLM

3. Import Empromptu’s Library and the debugger library by pasting the following lines at the top of the file.

   1. (Note: You can always turn off the debugger later)

from impromptu.prompt_tools import prompt_registry
from impromptu.model_utils import ModelUtils
import logging

1. Paste the debugger lines and your Empromptu API key at the top of the file.

No env file

logger = logging.getLogger("impromptu")
logger.setLevel(level=logging.INFO)
prompt_registry.set_api_key("xxxxxxxxxxxx")

Using an env file

logger = logging.getLogger("impromptu")
logger.setLevel(level=logging.INFO)
from dotenv import load_dotenv
import os
load_dotenv()
empromptu_api_key = os.getenv("EMPROMPTU_API_KEY")

Step 4: Update your application code to use Empromptu’s library

This is the most important step of Empromptu. We are going to replace where you define your prompts with an optimized prompt from Empromptu's prompt library.

If you get stuck on this step, please do not hesitate to chat with a technical expert here.

Using our UI

1. Go to our UI and create a Project and a Task that describes your AI Flow
2. Grab your custom Integration Code Snippet.

3. Find where you send your prompt to your LLMs.

4. Paste Empromptu’s Prompt Registry, Auto Scorer, and Auto Logger in the place where you call your prompts.

Without using our UI

1. Find where you send your prompt to your LLMs.

2. Paste the following code snippet, Empromptu’s Prompt Registry, Auto Scorer, and Auto Logger in the place where you call your prompts

prompt_context = prompt_registry.get_prompt(prompt_family_name,input_data)
prompt_registry.set_autolog()
ModelUtils.score_and_log(prompt_family_name,prompt_context)

Want to get fancy? You can discover and configure all of your options by visiting our config guide here.

Step 5: Run your code!

1. Once you run your code. You should see data show up on the dashboard and on the inputs page.

Are you stuck? We're on standby to help. Talk to a technical expert here.

What's Next

Lets review what you have accomplished:

You have installed Empromptu's library which includes, your new prompt registry, an auto-logger, our custom models to score all of your prompts to ensure that you are getting the most accurate and reliable prompts in real time with low latency.

• Create new prompts using

  • Empromptu's UI

  • In your code

• Define or upload your evaluations

• Optimize your input data

• Optimize your edge cases.