Practical overview
When it fits
Use it when you want to run GLiNER as a self-hosted API and need its documented queue, cache, monitoring and load-testing components.
Who it is for
- Developers integrating named-entity recognition into application backends.
- Teams running text-extraction jobs through asynchronous workers.
- Operators who want task monitoring and load-testing tools alongside a self-hosted model API.
Documented uses
- Expose GLiNER predictions through a FastAPI service.
- Distribute prediction tasks across Celery workers.
- Cache task results using Redis.
- Monitor Celery tasks with Flower.
- Load-test the API with Locust.
Requirements & limits
The repository is public and licensed under Apache-2.0. The documented setup runs with Docker Compose; the separate Streamlit client has extra dependencies.
- The documented startup path uses Docker Compose and includes FastAPI, Redis and Celery.
- Redis serves as both the Celery message broker and the task-result cache.
- The Streamlit client is independent and optional for running the API, but requires extra dependencies.
- The README documents local testing and does not provide a deployment procedure beyond the described setup.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the Docker Compose service
Build the documented images and start the API, Redis and Celery containers for local use.
Before you start
- Docker and Docker Compose
- A copy of the repository
- An environment file copied from .env.example, with REDIS_URL=redis://redis:6379/0 for local testing
Steps
Copy .env.example to .env and set the documented Redis URL.
Build the CPU image with docker build . -f Dockerfile.torch-cpu --pull -t torch-cpu:latest.
Build the GLiNER image with docker build . -f Dockerfile.gliner -t gliner-service:latest.
Start the containers with docker compose up -d.
Use python scripts/open_browsers.py to open the documented Swagger UI, Locust and Flower interfaces; the README also provides ./scripts/test-predict.sh for a prediction test.
Expected outcome
A local containerized GLiNER API setup with Redis and Celery, plus documented interfaces and a prediction test script.
Guide limitations
- The documented deployment procedure covers Docker Compose startup and local testing; it does not describe deployment beyond this setup.
Start the separate Streamlit client
Run the documented browser-based client for trying the prediction service.
Before you start
- The repository and its separately required Streamlit client dependencies
- The API service, for connecting the client to the prediction service
Steps
Follow the repository's client dependency setup.
Run ./scripts/start-ui.sh.
Expected outcome
A Streamlit browser interface for the prediction service.
Guide limitations
- The README identifies the UI as an independent client with extra dependencies; it is not required to run the API.
- The README does not detail the client dependency installation steps.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 16 October 2026
- Original website title
- henrikalbihn/gliner-as-a-service · Official homepage title · retained in its source language