Getting Started with PromptLoop
Welcome to PromptLoop, where we empower you with cutting-edge AI tools to automate and enhance your data tasks. This works by allowing you a simple way to create tasks that can accomplish your work including research, web scraping, text analysis and building out valuable company datasets.
PromptLoop allows your team to build business data to power your growth, doing more with less.
- Flexible - Access to hundreds of models and functions under a simple task based editor. Describe what you want without having to learn a new tool or chat with an AI model.
- Accurate - PromptLoop tasks are designed to stay within the bounds of what has been tested to work well. Whether extracting information from a website or labeling text, you will receive formatted and sourced answers from reliable models only using the relevant input data.
- Transparent - PromptLoop tasks and capabilities are transparent to the teams that rely on them. See exactly what your tasks are accomplishing and the steps they take to get there. This allows for peace of mind and helps you improve them with edits.
PromptLoop is built to help you accomplish repeatable AI research and analysis tasks, taking in sets of inputs (usually in the form of rows in a dataset) and returning formatted responses based on the task that you select.
Introduction
PromptLoop is a powerful set of tools that powers dataset creation for fast growing and forward leaning companies.
To make our services transparent and open to all, we offer several options to getting started, including a generous free trial.
For teams that need support growing their business with reliable tools and data, we offer team packages that are customized with both support and model capabilities.
To learn more and answer questions about your specific business you can book a demo here.
Core Concepts
- Tasks - Tasks are an action for our systems to accomplish for you. These are made up of inputs (like a website or search item) and return outputs which will be represented as new columns or rows of data.
- Datasets - Datasets are the repository for your information that you will be running tasks on. You can search, filter, save and share inputs and outputs from AI tasks up to hundreds of thousands of rows. For custom integrations with data sources, reach out to the team.
- Integrations - Spreadsheets - In most cases running data on your PromptLoop web portal is the recommended course, however, all tasks are integrated into both Excel and Google Sheets for cell specific use.
PromptLoop is a functional tool. We have a set of templates and building blocks to help you get started. Below you will find a few steps that we recommend when getting started.
Steps to Get Started
1. Define your goal and create a task
To get started with a new account, your first step is usually to create a new task.
A [task]](custom-tasks) is a function that you can run over and over with different inputs, rows of data, to get a set of outputs.
Tasks are designed to quickly carry out a series of steps. They can return multiple items, from a website crawl for instance. But are limited to what our models can accomplish. You can read more here
We have an extensive task library to get started with that has many building blocks, as well as the option to copy several base templates that you can then edit further.
If you have a custom task in mind, you can use the editor to help guide you through the process. This will let you identify inputs and the type of actions you want the model(s) to take.
2. Test and edit your task
Once you have created a task, PromptLoop makes it easy to modify and improve it to suit your needs.
Each task has a custom generated test page for it that will let you see what outputs will look like before deciding if it is correct. We suggest using a handful of test cases and variety of inputs that are reflective of the types of inputs you will be using it with.
If you need to modify the task, use the editor to do so. Edits can be submitted immediately and will change the instructions for the model. When adding edits be specific about things like formatting and style. These edits are not a prompt to the model but will update overall instructions. If you need to make major changes to the task, we recommend creating a new task instead of trying to edit an existing one.
3. Run a dataset on your task
Datasets are the basis for getting value from PromptLoop. Once you have a task, it is saved to your account for repeated use and ready for thousands of inputs.
Datasets, like tasks, are sharable throughout your organization. You can save edit and rename versions.
Start by selecting an excel or CSV file that has the type of data you want to use as inputs to the model. Upload this file on the datasets page, or select launch a new job and select the task that you want to launch.
Once you have selected a task you will identify the columns that will be used as input(s) to the model. You can then launch the job with one click.
It's off! Dataset jobs will run in the background.
4. Review, filter search and export
Your results will save as a new version in the dataset. You can search through them, noting the new column results to get a sense of what data was returned.
When satisfied, export all or a portion of the data. This flow is the core of our platform and end to end can take a few minutes but yield dozens of hours of work each run.