Accompanying code to www.shanelynn.ie tutorial on FastAPI NLP API creation.
In this simple API, we're building a web application that can process text in HTTP requests to extract sentiment and entities automatically.
Start a virtual environment (python3 -m venv venv) and install the requirements in that virtual environment (FastAPI, Spacy, Flair, Uvicorn[standard]). This tutorial was built with Python 3.9 originally.
pip install -r requirements.txt
You'll also need to install language models for Spacy:
python -m spacy download en_core_web_sm
Start the main API with:
uvicorn simple_api:app --reload
The first run may take longer than usual if Flair has to download any models to run.
There are some basic examples (from the blog) in the examples/ directory showing:
- Basic GET requests with FastAPI
simple_api.py - Basic POST API with FastAPI
simple_post_route.py - Sentiment detection with Flair alone
flair_sentiment_sample.py - Named entity recognition (NER) with Spacy
spacy_ner_example.py
Run these with simple python examples/<example_filename.py>.
You can test the application with curl commands. Send a request with the text of your choosing to http://localhost:8000 when the API is running.
curl -X POST -H "Content-Type: application/json" -d '{"query": "I absolutely love building Python applications with FastAPI and Shane's blog."}' http://localhost:8000/analysis/