Problem
Settings.host and Settings.port (api/src/temporal_model/api/settings.py:44-45) are documented in the README's configuration list (HOST, PORT) but never read anywhere in src/ — every entry point hardcodes the bind on the uvicorn CLI instead:
api/Dockerfile:35 — CMD [..., "--host", "0.0.0.0", "--port", "8000"]api/Makefile serve — uvicorn defaults (127.0.0.1:8000)api/Makefile run-gpu — --host 0.0.0.0 --port 8000
So a user setting TEMPORAL_API_PORT=8080 (e.g. to dodge the docker-compose port collision noted in the README's GPU section) silently gets 8000 anyway.
Surfaced during the review of #52.
Options
- Drop the two settings and remove
HOST/PORT from the README config list — smallest change; bind stays a deployment concern (CLI flags / image CMD).
- Wire them up — e.g. a
python -m temporal_model.api entry point that calls uvicorn.run(app, host=settings.host, port=settings.port), used by the Dockerfile and Makefile targets.
Option 1 seems right unless someone actually needs runtime-configurable binds in the container.
Problem
Settings.hostandSettings.port(api/src/temporal_model/api/settings.py:44-45) are documented in the README's configuration list (HOST,PORT) but never read anywhere insrc/— every entry point hardcodes the bind on the uvicorn CLI instead:api/Dockerfile:35—CMD [..., "--host", "0.0.0.0", "--port", "8000"]api/Makefileserve— uvicorn defaults (127.0.0.1:8000)api/Makefilerun-gpu—--host 0.0.0.0 --port 8000So a user setting
TEMPORAL_API_PORT=8080(e.g. to dodge the docker-compose port collision noted in the README's GPU section) silently gets 8000 anyway.Surfaced during the review of #52.
Options
HOST/PORTfrom the README config list — smallest change; bind stays a deployment concern (CLI flags / image CMD).python -m temporal_model.apientry point that callsuvicorn.run(app, host=settings.host, port=settings.port), used by the Dockerfile and Makefile targets.Option 1 seems right unless someone actually needs runtime-configurable binds in the container.