Skip to content

Fix README→bundle warehouse drift; materialize VS index on first deploy#1

Merged
SatyKrish merged 1 commit intomainfrom
fix/warehouse-drift-and-vs-index
Apr 26, 2026
Merged

Fix README→bundle warehouse drift; materialize VS index on first deploy#1
SatyKrish merged 1 commit intomainfrom
fix/warehouse-drift-and-vs-index

Conversation

@SatyKrish
Copy link
Copy Markdown
Owner

@SatyKrish SatyKrish commented Apr 26, 2026

Summary

  • Plumb DOCINTEL_WAREHOUSE_ID through the bundle as --var warehouse_id in scripts/bootstrap-dev.sh and .github/workflows/deploy.yml, so the dashboard + serving env match what wait_for_kpis / agent/log_and_register.py use. Without this, the bundle falls back to its lookup: warehouse: Serverless Starter Warehouse default — which fails validation in workspaces lacking that named warehouse and silently picks a different ID otherwise.
  • Add an explicit databricks bundle run … index_refresh after stage-2 first-deploy in bootstrap-dev.sh. The table_update trigger does not fire retroactively on rows already produced by the stage-2 pipeline run, so the VS index never materialized on a fresh workspace until the next pipeline run. sync_index.py is create-if-missing/sync-if-exists, so this stays idempotent.
  • README fixes: include pytest in the install line so the documented pytest agent/tests/ works after install; add DOCINTEL_WAREHOUSE_ID and .venv/bin/python to the steady-state agent registration command (the script reads the var; macOS doesn't ship python); correct the env-var table description for DOCINTEL_WAREHOUSE_ID — it isn't read by evals/clears_eval.py.

Test plan

  • bash -n scripts/bootstrap-dev.sh
  • databricks bundle validate --strict -t devValidation OK!
  • databricks bundle validate --strict -t prod → fails as designed without service_principal_id
  • .venv/bin/python -m pytest agent/tests/ -q → 18 passed
  • End-to-end fresh-workspace ./scripts/bootstrap-dev.sh (requires real workspace; not run locally — review the diff against the runbook before merging)
  • Confirm GitHub Actions vars.DOCINTEL_WAREHOUSE_ID is set in the repo (the validate step now reads it)

🤖 Generated with Claude Code

Plumb DOCINTEL_WAREHOUSE_ID through bundle as --var warehouse_id in
bootstrap + CI so dashboard/serving env match wait_for_kpis and
log_and_register, instead of falling back to the lookup-by-name default
that fails in workspaces lacking "Serverless Starter Warehouse".

Add an explicit `bundle run index_refresh` after stage-2 first-deploy
because the table_update trigger doesn't fire retroactively on rows
already produced by the stage-2 pipeline run.

README fixes: include pytest in install line, add DOCINTEL_WAREHOUSE_ID
+ .venv/bin/python to the steady-state agent registration command, and
correct the env-var-table description (the var isn't read by the eval).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
@SatyKrish SatyKrish merged commit 5a855b6 into main Apr 26, 2026
@SatyKrish SatyKrish deleted the fix/warehouse-drift-and-vs-index branch April 26, 2026 02:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant