Practical overview
When it fits
Use Refinery when an NLP project needs labeled examples, when existing annotations need quality assessment, or when a team wants to focus limited annotation resources on the data that needs attention.
Who it is for
- Data scientists building NLP models who need to create or improve labeled training datasets.
- Teams that need to assess existing annotations or prioritize limited labeling effort.
Documented uses
- Create labeled NLP training data with semi-automated, human-in-the-loop workflows.
- Identify low-quality subsets in existing training datasets.
- Organize and monitor training data in one place.
Requirements & limits
The repository documents a Python package installation and a local multi-service setup. It also describes a mostly read-only online playground.
- The project does not replace manual labeling; its stated aim is to make annotation time more focused.
- The repository describes a multi-service application rather than a single standalone labeling script.
- The README identifies Hugging Face and spaCy for pre-built language models and Qdrant for neural search.
- Local data defaults to refinery/postgres-data; changing the path requires editing LOCAL_VOLUME in the operating system's start script.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install and open Refinery from the repository
Start the documented multi-service application locally and open its browser interface.
Before you start
- A cloned Refinery repository.
- A Mac/Linux or Windows environment, using the corresponding documented start script.
Steps
Clone https://github.com/code-kern-ai/refinery and change into the repository directory.
On Mac/Linux, run ./start; on Windows, run start.bat.
After setup, open http://localhost:4455 in a browser.
When finished, run ./stop on Mac/Linux or stop.bat on Windows.
Expected outcome
The local Refinery interface is available in a browser, and the documented stop script shuts down the services.
Guide limitations
- The repository says the setup runs multiple services and may take several minutes.
- By default, data is stored in refinery/postgres-data. Removing that volume irreversibly deletes its data.
Install Refinery with pip and start its service
Install the package and use its documented commands to start and stop Refinery.
Before you start
- A Python environment with pip.
Steps
Run pip install kern-refinery.
Go to the directory where you want the data stored.
Run refinery start.
When finished, run refinery stop.
Expected outcome
The package's documented command starts Refinery, with the data directory selected by the user.
Guide limitations
- The README does not specify further environment or version requirements in the installation instructions.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- code-kern-ai/refinery · Official homepage title · retained in its source language