We are thrilled to unveil Argilla 2.0, marking a significant milestone in our journey towards making data tools accessible for everyone. Argilla is easy to get started, easy to use, and is designed for everyone who wants to maintain high-quality AI projects. Argilla puts data at the center of your workflow.
For those interested in the backstory and motivations behind the 2.0 development, the previous introducing 2.0 post provides detailed insights.
Argilla for Everyone!
Argilla is an open-source collaboration platform for AI engineers and domain experts to build high quality datasets. Argilla aims for simplicity, flexibility, and inclusivity.
The SDK and the UI provide two ways to interact with your datasets. The Argilla team is committed to involving as many qualified professionals as possible who believe data quality is key to enhancing model performance. Everyone should contribute to AI.
Deploying Argilla on the Hugging Face Hub is easier than ever
Argilla 2.0 is designed to get started with a few clicks to deploy on Hugging Face Spaces.
It’s quick and you don't need to maintain a server or run any commands. By sharing your Argilla Space with your team or open it to the OSS community your can build human feedback collection projects in just a minutes.
Team collaboration is now possible
Argilla 2.0 is optimized for team collaboration. With minimal setup you can complete annotation tasks in record time by sharing the work.
As the admin of your project, you can define the minimum number of annotations for each record. Each team member will have access to the whole dataset. The task is completed once all the records receive the number of annotations.
Kids could almost configure a dataset from the Argilla SDK
The new Python SDK is designed for simplicity with a unified API for managing datasets, settings, users, and workspaces. It naturally integrates with libraries like pandas and datasets. You don’t need to write dozens of lines of code or deal with weird abstractions to configure your projects.
It enables you to add records on the server. You can easily add new ones and remove others during the whole process. You can build datasets with multiple fields and set a choice of questions to collect feedback from annotators. Argilla supports similarity search and metadata to let annotators quickly review specific dataset samples.
The new SDK is designed together with the community over six months. We shared the alpha version two months ago and it has received positive feedback from the community.
import argilla as rgclient = rg.Argilla()settings = rg.Settings( guidelines="guidelines.md", fields=[ rg.TextField( name="Prompt", ), ], questions=[ rg.LabelQuestion( name="label", labels=["good", "bad"] ), ],)dataset = rg.Dataset( name="my_dataset" settings=settings)dataset.create()Information architecture is designed for every need
Whether you're a developer, domain expert, annotator, team manager, or AI engineer, Argilla's flexible UI layout lets you to focus on what’s most important. The flexible layout lets you hide, reduce, or expand fields, questions, guidelines, and progress metrics based on your task.
Admins can easily manage settings from the UI, including question and field edition, labels configuration, value reordering, and markdown fields. Most of these settings can be changed even when projects have started!
Whether you're exploring a new dataset or quickly annotating large volumes of data, the UI is designed to optimize your workflow. Focus and bulk modes for annotating are enabled to complete tasks like classification or NER where you can label dozens of records in one go.
A guide for each of your steps
We like the Diátaxis framework and we applied the UX principles to the documentation. That means you don’t need to read every guide to try out Argilla. Follow the Quickstart and a basic guide, and you are ready to start collecting annotations.
Whether you are transitioning from Argilla 1.X datasets or just starting out, the guides are tailored to your needs. The more you read, the more you will know, but ultimately, you shape your own project with your own rules.
Argilla docs aims to support newcomers and experienced practitioners alike, whatever your profile is. Teams are diverse and we consider is crucial to embark everyone in contributing to AI projects.
If you are excited to try Argilla, you’re only a few clicks away: https://docs.argilla.io/latest/getting_started/quickstart/
