Argilla
Deploy Argilla, a collaboration tool for building high-quality datasets for AI
Overview
Argilla is a tool designed to help AI engineers and collaborators build high-quality datasets through labeling, curation, and more. Using Argilla, you can provide feedback to steer training for natural language processing (NLP), large language models (LLMs), and multimodal models (text to images, etc.). The goal of the project is to set up workflows that help you quickly iterate through the right data and models through programmatic evaluation.
This Starter deploys Argilla as a standalone server. Once deployed, you can use the web interface to administer your instance and connect to it from your applications to load data and process the results.
Configuration
During configuration, open the Health checks section and set Grace period to 300
. This is necessary in order to avoid deployment failures due to timeouts as Argilla launches.
The following environment variables are available with this image:
OWNER_USERNAME
: The owner account name for Argilla. This account has full access to all workspaces and functions. If unset, this defaults toowner
.OWNER_PASSWORD
: The password for the owner username. If unset, this defaults to12345678
.OWNER_API_KEY
: The API key that can be used with the Argilla Python library to interact with the application. If unset, this will default toowner.apikey
.ADMIN_USERNAME
: The admin account for Argilla. This account has permissions on the workspace it has been assigned and cannot assign other users to it. If unset, this defaults toadmin
.ADMIN_PASSWORD
: The password for the admin username. If unset, this defaults to12345678
.ADMIN_API_KEY
: The API key that can be used with the Argilla Python library to interact with the application. If unset, this will default toadmin.apikey
.ANNOTATOR_USERNAME
: The annotator account name for Argilla. This account can annotate and provide feedback on the datasets it has been assigned. If unset, this defaults toargilla
.ANNOTATOR_PASSWORD
: The password for the annotator username. If unset, this defaults to12345678
.
Once the Argilla instance is ready, you can visit the Koyeb service's public URL to log in with one of the above accounts. Follow the Argilla quickstart to learn how to load datasets and begin using Argilla.