Skip to content

v2.9.1

Choose a tag to compare

View all tags
@placerda placerda released this 14 Jun 14:55
v2.9.1
7fa7290
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

User and operator impact

This release completes the custom_metadata feature started in v2.9.0. Blob user-defined metadata can now be extracted during ingestion and stored on Azure AI Search chunks, so retrievers can filter by blob tags such as department, document type, or business unit.

It also makes deployment troubleshooting easier. If a shell-level APP_CONFIG_ENDPOINT disagrees with the active azd environment, the deploy scripts now warn clearly before deploying with the shell value.

Changed

  • Orchestrator pin bumped from v2.8.2 to v2.8.3: Adds the deploy-time APP_CONFIG_ENDPOINT drift warning for orchestrator deployments (Azure/GPT-RAG#491).
  • Ingestion pin bumped from v2.4.3 to v2.4.4: Completes Azure/GPT-RAG#487 by extracting user-defined blob metadata into the custom_metadata search field added in v2.9.0. It also restores the Content Understanding multimodal extraction path for PDF page/region rendering and Office embedded image extraction.
  • UI pin bumped from v2.3.10 to v2.3.11: Adds the same deploy-time APP_CONFIG_ENDPOINT drift warning for UI deployments.
  • Deploy scripts now warn when configuration sources disagree: The component deploy scripts compare the shell APP_CONFIG_ENDPOINT with the active azd environment value. If both exist and differ, the scripts show a yellow warning, say which value will be used, and show how to clear the shell override. Shell precedence is unchanged, so existing jumpbox and CI flows keep working. The matching docs update was published in Azure/GPT-RAG#492.

Component versions

Component Version
gpt-rag-ui v2.3.11
gpt-rag-orchestrator v2.8.3
gpt-rag-ingestion v2.4.4
infra / AI Landing Zone v2.0.16

Validation

  • The deploy-script warning is diagnostic only. It does not change provisioned resources.
  • Orchestrator validation: pytest, 166 tests passed.
  • Ingestion validation: pytest tests/, 7 tests passed for blob metadata extraction.
  • UI validation: no automated test suite in this repo, deploy-script behavior was validated through the shared script change used across the components.
  • A fresh azd provision and azd deploy regression was already completed for the same infrastructure baseline during v2.9.0 validation, so v2.9.1 did not repeat a full azd up.
Assets 2
Loading