Skip to content

SCHOOLAI.md

Latest commit

 

History

History
73 lines (49 loc) · 2.16 KB

SCHOOLAI.md

File metadata and controls

73 lines (49 loc) · 2.16 KB
Raw

SchoolAI

This is SchoolAI's fork of Unstructured-IO/unstructured-api. We run it internally for document partitioning in the web-app RAG pipeline.

There is no CI/CD in this repo. Image builds are infrequent enough to do manually from a local machine.

Image

Staging and production pull:

us-central1-docker.pkg.dev/schoolai-global/docker/unstructured-api:latest-amd64

That reference lives in web-app at ops/unstructured-api/values-global.yaml. The cluster runs amd64 only (kubernetes.io/arch: amd64).

Build and publish

Prerequisites:

  • Docker with buildx
  • gcloud CLI with access to the schoolai-global project
gcloud auth login
gcloud auth configure-docker us-central1-docker.pkg.dev

From the repo root, build for linux/amd64 and push:

IMAGE=us-central1-docker.pkg.dev/schoolai-global/docker/unstructured-api

docker buildx build \
  --platform linux/amd64 \
  --build-arg PIPELINE_PACKAGE=general \
  -t "${IMAGE}:latest-amd64" \
  --push \
  .

On Apple Silicon, --platform linux/amd64 is required so the image runs on GKE. The build may be slow because it cross-compiles.

Optional versioned tag (useful for tracking what was deployed):

VERSION="$(date +'%Y_%m_%d-%H_%M')-$(git rev-parse --short HEAD)"
docker buildx build \
  --platform linux/amd64 \
  --build-arg PIPELINE_PACKAGE=general \
  -t "${IMAGE}:${VERSION}-amd64" \
  -t "${IMAGE}:latest-amd64" \
  --push \
  .

Roll out

Pushing latest-amd64 does not restart running pods by itself. To pick up the new image in staging:

  1. Trigger a web-app staging deploy (e.g. push to web-app main that touches ops/unstructured-api/**, or run the webapp staging deploy workflow manually), or
  2. Restart the unstructured-api deployment in the webapp namespace on the target cluster.

Local development

To build and run locally without publishing:

make docker-build
make docker-start-api

The API is available at http://localhost:8000. See the main README for API usage examples.

web-app's docker-compose.yml can also run the published image directly for local integration testing.