Codebase audit: stale package-rename references and a config env-var mismatch

[calebrob6]

This is a comprehensive audit of the repo on fix/initial-setup (post-hasteutils → hastelib directory rename and haste/hasteutils → hastegeo package rename). All Python module imports under api/, hastelib/src/hastegeo/, and docker/training/code/ are clean — but a number of stale references in code, scripts, and docs still need to be cleaned up. One of them is a runtime crash and another is a silent production-deploy break.

Filing as a single issue because the items are tightly related (all fallout from the same refactor). Happy to split into per-PR sub-issues if preferred.

---

Critical — runtime crash

1. hastelib/src/hastegeo/core/processors/artifacts.py:231 — hardcoded stale module path

command = '"python -m haste.workflows.zip_artifacts"'

ArtifactProcessor.submit_zip_job() builds the Azure Batch task command literally referencing the old package name. With the rename to hastegeo, every artifact-zip job will fail with ModuleNotFoundError: No module named 'haste' — both on Azure Batch and through the local Docker runner (runner_type=local).

For comparison, imagery.py:319 correctly uses the prepare-imagery CLI shim, which the Dockerfile maps to python -m hastegeo.workflows.prepare_imagery (fixed in commit ffdfb30).

Fix: Either change to python -m hastegeo.workflows.zip_artifacts or, for symmetry with imagery.py, switch to the zip-artifacts CLI shim (already provisioned in docker/imageryprep/Dockerfile).

---

High — production deploy silently broken

2. AZURE_BATCH_REGISTRY_SERVER_URL vs AZURE_BATCH_REGISTRY_SERVER env-var mismatch

The setup/deploy scripts and the local-settings example all set:

• setup/deploy_apps.sh:146 — AZURE_BATCH_REGISTRY_SERVER_URL=https://${ACR_NAME}.azurecr.io
• setup/setup_infra.sh:582 — same
• .github/scripts/deploy_apps.sh:110 — same
• local.settings.example.jsonc:41 — same

But the Config layer reads a different name:

• hastelib/src/hastegeo/core/config.py:411 — os.getenv(\"AZURE_BATCH_REGISTRY_SERVER\", \"<registry-name>.azurecr.io\")

…and the value flows straight through to Azure Batch:

• hastelib/src/hastegeo/core/runners/azure_batch.py:139 — self.batch_config[\"registry_server\"]

So in every production deployment, the registry server falls back to the placeholder \"<registry-name>.azurecr.io\" and Azure Batch will fail to authenticate to ACR for any training/imagery-prep job.

Fix: Pick one name everywhere. Cleanest is to rename the env var in all four deploy locations to AZURE_BATCH_REGISTRY_SERVER (and also drop the https:// prefix — the Azure Batch SDK's ContainerRegistry.registry_server expects a bare hostname).

---

High — broken docs / contributor guide

3. README.md references nonexistent path hastelib/src/haste_geo

Lines 224, 272, 300, 312, 325 all use hastelib/src/haste_geo (with underscore). The directory does not exist — the actual path is hastelib/src/hastegeo (no underscore). The hot-reload docker run -v examples will silently mount an empty directory, defeating the purpose. Lines 272/300/325 also mount to /usr/local/lib/python3.11/site-packages/haste, which doesn't match the current Dockerfiles (they copy the source to /app/hastegeo and use PYTHONPATH=/app).

Fix: Replace hastelib/src/haste_geo with hastelib/src/hastegeo everywhere; update the mount target to /app/hastegeo (or remove that block if it no longer represents a supported workflow).

4. docs/hastelib/*.md Sphinx automodule directives use old haste. package

Every page under docs/hastelib/ uses .. automodule:: haste.core.X and haste.workflows.X:

• docs/hastelib/processors.md, models.md, config.md, utils.md, data-layer.md, artifact-storage.md, runners.md, workflows.md, modules.md

The Jupyter Book build (run by .github/workflows/docs-deploy.yml) imports the renamed hastegeo package, so every automodule directive will produce empty / failed pages.

Fix: Bulk rename haste. → hastegeo. across docs/hastelib/*.md. Also fix docs/hastelib/modules.md title and tree (haste_geo → hastegeo) and docs/architecture.md:79-90 ("haste (hasteutils)", "haste v0.0.67", -e hasteutils/).

5. docs/requirements.txt:8 and :43 — stale wheel name in install instructions

Comments tell the user to install:

pip install --no-deps hastegeo @ https://researchlabwuopendata.blob.core.windows.net/haste-binaries/haste-0.0.67-py3-none-any.whl

That wheel filename hasn't existed since the rename — the current wheel is hastegeo-1.0.0-py3-none-any.whl (matching the references in api/*/requirements.txt).

Fix: Update the wheel URL to hastegeo-1.0.0-py3-none-any.whl.

---

Medium — misleading docs

6. docs/getting-started.md:31

Says "installs the haste core library in editable mode (-e hasteutils/)" — both names are wrong. Should be \"installs the hastegeo core library in editable mode (-e hastelib/)\".

7. docs/README.md:38

Lists hasteutils/ - Core library documentation — directory is now hastelib/.

8. THIRD-PARTY-NOTICES.md:15

References hasteutils/pyproject.toml. Path is now hastelib/pyproject.toml.

9. docker/api/Dockerfile and docker/ui/Dockerfile — dead files

Already filed separately in #20 — listed here for completeness.

---

Low — cosmetic

10. README.md:184

VSCode tip mentions haste.core modules — should be hastegeo.core.

11. .github/copilot-instructions.md:202-203

Documents the legacy hot-reload mount path /usr/local/lib/python3.11/site-packages/haste. Current Dockerfiles use /app/hastegeo.

12. docs/azure_functions_simple.py:119 and :160

Stale comments saying "(like hasteutils functions)" / "(like hasteutils style)". Cosmetic only.

13. ui/src/Components/LabelingTool/LabelingToolHelper.js:154

await apiPut(\"PutLabelsFromLabelTool?imageLayerId\", labelingToolDataRef.current);

Sends ?imageLayerId with no value. The server reads imageLayerId from the JSON body (function_app.py:1074), so the query parameter is unused — harmless but misleading.

Fix: Drop the ?imageLayerId fragment from the URL.

---

Out of scope / intentionally left

• The \"hastetraining\", \"hasteimageryprep\", \"hastefuncapi\", \"hastequeue\", \"hastestorageapi\" Azure resource names in setup/*.sh, setup/README.md, and .github/workflows/README.md are intentional infra naming conventions that were not part of the package rename.
• The HASTE = \"haste\" enum value in hastelib/src/hastegeo/core/models/training.py:10 is a model-source identifier, not a package reference.
• CHANGELOG.md:35 mentions hasteutils as a historical entry, which is appropriate for a changelog.