Argilla on Ubuntu 24.04 LTS
by cloudimg
Argilla data annotation for NLP and LLM datasets: Elasticsearch, a worker and a per-VM API key
Argilla on Ubuntu 24.04 LTS
Argilla is the open source data annotation and curation platform for building the high-quality datasets that modern Natural Language Processing and Large Language Models depend on. Data, machine learning and domain teams label, review and curate text, chat and multi-field records through a collaborative web interface backed by a full REST API and a Python SDK. cloudimg packages it as a finished, production-ready appliance, not a bare pip install.
What you get
- argilla-server 2.8.0 with the bundled compiled web UI, in a Python 3.12 virtual environment run by an unprivileged service account
- A memory-tuned single-node Elasticsearch 8.x record and vector search backend, Redis, and the RQ background worker that Argilla 2.x requires to process dataset and record operations
- Served in production by uvicorn behind an nginx reverse proxy on port 80; Elasticsearch, Redis and the application server all bind to loopback only
- Text classification, token and span labelling, ranking and preference collection and free-form feedback, organised into workspaces with owner, admin and annotator roles
- Secure by default: Argilla's well-known default accounts and API key are never created; a unique owner password, a fresh API key and a fresh token signing key are generated on the first boot of every VM and the owner credentials written to a root-only file
What it is for
Argilla is the human-feedback and dataset layer teams use to curate instruction and preference sets for fine-tuning and alignment (RLHF and DPO), to review and correct model outputs, and to build labelled corpora for classification and extraction. Create datasets and log records from the Python SDK, then annotate them collaboratively in the web UI.
Licensing and support
Argilla is free and open source under the Apache 2.0 licence. The cloudimg charge (0.04 US dollars per vCPU per hour) covers packaging, security patching, image maintenance and 24/7 expert support. Every deployment is paired with a step-by-step guide tested against the exact image you deploy.