Deploying Lettuce

Deploying Lettuce

Lettuce provides a server for HTTP endpoints. If you have successfully run the quickstart then you have the correct setup for running a Lettuce server.

💡

Making Lettuce easier to deploy is the next milestone for development, so watch this space!

The code for the server runs from api.py

uv run --env-file .env api.py

After running this command, you should see messages like these in your terminal:

INFO:     Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
INFO:     Started reloader process [73042] using StatReload
INFO:     Started server process [73353]
INFO:     Waiting for application startup.
INFO:     Application startup complete.

This means that the server is running, and is able to accept HTTP requests. A request can be made using e.g. curl

curl -X POST "http://127.0.0.1:8000/pipeline/" -H "Content-Type: application/json" -d '{"names": ["Betnovate Scalp Application", "Panadol"]}'

Endpoints

The endpoints will soon be re-arranged once the experimental pipelines have been stabilised. Consult the routes documentation.

The following endpoints will take HTTP POST requests:

  • /pipeline/: generates a suggested formal name for a source term(s) and searches the OMOP-CDM for that suggestion
  • /pipeline/db: searches the OMOP-CDM for a provided source term(s)
  • /pipeline/vector_search: searches a vector database for concepts semantically similar to the provided source term(s)
  • /pipeline/vector_llm: searches a vector database for concepts semantically similar to the source term(s) and uses these for retrieval-augmented generation with the LLM

Request options

Requests to the API endpoints must have a body containing a list of informal names ("names"), and can optionally define options for the pipeline, (pipeline_options).

Once you have the API accepting requests, you can use the GUI to more easily make requests.

Lettuce quickstartLettuce API reference