Skip to content

setup.md

Latest commit

 

History

History
221 lines (151 loc) · 5.72 KB

setup.md

File metadata and controls

221 lines (151 loc) · 5.72 KB

Setup and Local/Production Usage

Table of Contents

Installation (Python 3.12) using PDM

pdm venv create 3.12
pdm use
pdm venv activate
pdm sync --dev
wget -P training_data https://dl.fbaipublicfiles.com/fasttext/supervised-models/lid.176.bin

Using Docker

  1. Install Docker engine - https://docs.docker.com/engine/install/
  2. Start LanguageTool from Docker Hub (erikvl87/languagetool):
# Create an isolated network so containers can talk by name
docker network create nlp-net

# Start LanguageTool on port 8010
docker run -d --name lt --network nlp-net -p 8010:8010 erikvl87/languagetool:latest

LanguageTool API will be available at http://localhost:8010/v2

  1. Build and run the NLP API locally:
# Build the API image from the local Dockerfile
docker build -t nlpapi .

# Run the API and point it to the LanguageTool container
docker run -d --name nlp_api --network nlp-net -p 8080:8080 \
  -e LANGUAGETOOL_API=http://lt:8010/v2 \
  nlpapi

You should see the application running under http://localhost:8080/docs

Using docker compose (recommended for local multi-service setup)

A ready-made compose.yml is included to start both the NLP API and LanguageTool with one command. It:

  • Launches erikvl87/languagetool on port 8010
  • Builds and runs the API on port 8080
  • Mounts your local ./models directory read-only into the container (so local SetFit models can be used if present)
  • Sets LANGUAGETOOL_API inside the API container to point to the LanguageTool service

Quick start:

# Start both services in the background
docker compose up -d

# Check health
curl http://localhost:8080/health

# Visit http://localhost:8080/docs in your browser

Enable local context checker models (if you downloaded them):

# Option 1: temporarily set env when starting
CONTEXT_CHECKER_LOCAL=true docker compose up -d

# Option 2: uncomment CONTEXT_CHECKER_LOCAL in compose.yml and re-run

Updating models: any changes you make in ./models on the host are reflected in the container because of the bind mount. The mount is read-only (:ro) to prevent accidental writes from inside the container.

Stop and remove services:

docker compose down

Rebuild after code changes (forces image rebuild):

docker compose build --no-cache nlpapi
docker compose up -d

Tail logs:

docker compose logs -f nlpapi
docker compose logs -f languagetool

If you only need the API without LanguageTool, you can still use the single-container commands above or remove the languagetool service from the compose file.

Update dependencies locally

To update packages locally after pyproject.toml/pdm.lock was changed:

pdm sync --dev

Add new package

When adding a new package to the project, update pyproject.toml via:

pdm add <package_name>

Docker image

After making changes in the code or in the Dockerfile, build a fresh image:

pdm export --prod -o requirements.txt
DOCKER_BUILDKIT=0 docker build -t nlpapi . --no-cache
docker run nlpapi

Install Platform.sh CLI

  • Run platform login
  • Run platform project:set-remote
  • Run platform list to find out what commands are available
  • Run platform help [command] to find out details about a command

See https://docs.platform.sh/development/cli.html for details

Adjust size on platform.sh

Adjust server size:

platform e:curl -e main /deployments/next -X PATCH -d '{"webapps":
  {"app": {"resources": {"profile_size": "8"}},"languagetool": {"resources": {"profile_size": "4"}}}}'

Adjust instance count:

platform e:curl -e main /deployments/next -X PATCH -d '{"webapps":
  {"app": {"resources": {"instance_count": "2"}},"languagetool": {"resources": {"instance_count": "2"}}}}'

Run Locally

Note for Mac users. Set environment variables with the following snippet:

cp .env.development.mac .env

Start the dev server:

pdm run uvicorn app.main:app --reload

or

uvicorn app.main:app --reload

Open your browser to http://localhost:8000/docs to view the OpenAPI UI.

Alternative docs: http://localhost:8000/redoc

Profiling locally with Blackfire

pdm run blackfire-python uvicorn app.main:app --reload

Make sure you have a .blackfire.ini, get the settings from https://blackfire.io/docs/php/configuration

BLACKFIRE_SERVER_ID=""
BLACKFIRE_SERVER_TOKEN=""

Production Deployment

Set an env variable API_DOCS_AUTH_ENABLED to "true" and set API_DOCS_USERNAME and API_DOCS_PASSWORD for basic auth for the API docs.

If the build fails due to "No space left on device" while installing dependencies run:

platform project:clear-build-cache

See: https://docs.platform.sh/development/troubleshoot.html#clear-the-build-cache

See Also