Skip to content

DEPLOYMENT.md

Latest commit

 

History

History
45 lines (34 loc) · 1.27 KB

DEPLOYMENT.md

File metadata and controls

45 lines (34 loc) · 1.27 KB
Raw

Legacy CPU Development — Production Deployment Retired

Production uses one GPU-backed Cloud Run service:

  • Cloud Run service: prompt-compression
  • Artifact Registry image: prompt-compression-gpu
  • Production runbook: DEPLOYMENT_GPU.md

There is no production CPU service. Never deploy Dockerfile or cloudbuild.yaml to prompt-compression; doing so would replace the GPU revision behind the existing domain.

Dockerfile and cloudbuild.yaml remain only for local CPU development and historical reproducibility. The CPU Cloud Build image is deliberately named prompt-compression-cpu-retired so it cannot be mistaken for the production image.

Local CPU Development

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
pip install -r requirements-dev.txt
python -m pytest

Run directly:

uvicorn app.main:app --host 127.0.0.1 --port 8080 --reload

Or run the legacy CPU container locally:

docker build -t prompt-compression-cpu:local .
docker run --rm -p 8080:8080 `
  -e COMPRESSOR_DEVICE=cpu `
  prompt-compression-cpu:local

For any Google Cloud build, deploy, rollback, environment update, GPU sizing, traffic change, or production verification, stop here and follow DEPLOYMENT_GPU.md.