This project is a FastAPI-based sentiment analysis service using a Hugging Face transformers model, containerized with Docker and deployable on Kubernetes, featuring automated testing and MLflow integration for performance tracking.
- FastAPI for REST API: High performance, easy to use framework for building APIs.
- Hugging Face Transformers: Utilize powerful pre-trained models for sentiment analysis.
- Docker and Kubernetes: Containerization and orchestration support for easy deployment and scalability.
- Poetry for Dependency Management: Reproducible builds and straightforward dependency management.
- Automated Testing: Includes tests for API functionality and model performance evaluation.
- MLflow Integration: Track model performance metrics for continuous improvement.
-
Clone the Repository
git clone
Ensure that Poetry is installed on your system. Then, in the project directory, run:
poetry install
Create a .env file in the root directory and specify the following variables:
MODEL_NAME: The name of the pre-trained model from Hugging Face.MLFLOW_URL: The URL for MLflow tracking server.PREDICT_URL: The URL for the FastAPI prediction endpoint.
Use the following command to run the FastAPI application:
poetry run uvicorn main:app --reload
Refer to the Kubernetes manifests (fastapi_deployment.yaml, fastapi_hpa.yaml, fastapi_service.yaml) for deploying the application to a Kubernetes cluster. Ensure Docker is installed and configured for building and pushing the container image.
To run the tests, execute the following command:
poetry run pytest