Skip to content

v2.0.0

Latest
Latest

Choose a tag to compare

View all tags
@Roopan-Microsoft Roopan-Microsoft released this 23 Jul 15:32
· 6 commits to main since this release
v2.0.0
4708902
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

2.0.0 (2026-07-23)

Major release promoting the ground-up v2 architecture to main (#2326). The stack moves to a React + Vite frontend, a FastAPI backend, and an Azure Functions ingestion worker on Azure Container Apps, with a fully modular Bicep infrastructure, pluggable orchestration, and secretless managed-identity authentication. Deploys with a single azd up.

⚠ BREAKING CHANGES

  • Full v1 → v2 rewrite. The v1 application and infrastructure are replaced; existing deployments must be redeployed with azd up (no in-place upgrade).
  • Removed Streamlit UI/admin, Prompt Flow, Semantic Kernel, direct openai/AzureOpenAI SDK usage, Poetry (now uv), Key Vault for application secrets, the one-click "Deploy to Azure" ARM button (azd-only), Azure Bot Service, and the Teams extension.

Infrastructure Modularity

  • Restructure Bicep into a modular infra/ layout with reusable, single-responsibility modules composed from a single main.bicep (#2326)
  • Adopt Azure Verified Modules (AVM) for core resources and centralize resource naming through abbreviations.json (#2326)
  • Deploy the full workload to Azure Container Apps with per-service modules for the frontend, backend API, and ingestion worker (#2326)
  • Deploy-time backend choice between Azure AI Search + Cosmos DB and PostgreSQL + pgvector, selected by a single parameter (#2326)
  • Secretless infrastructure: one user-assigned managed identity plus Azure RBAC across Foundry, Search, PostgreSQL, Storage, Content Safety, Speech, and Container Registry — no Key Vault (#2326)
  • WAF-aligned hardened deployment variant for tenants with additional security restrictions (#2326)
  • Modular Azure Functions ingestion blueprints (batch_start, batch_push, add_url, blob_event, search_skill) triggered via Event Grid (#2326)
  • Rewritten .env.sample documenting the full v2 configuration surface, aligned with backend settings and infrastructure (#2326)

Features

  • New React + Vite single-page chat frontend replacing the Streamlit UI (#2326)
  • New FastAPI backend that runs headless and exposes a generated OpenAPI client for the frontend (#2326)
  • Streaming answers over typed SSE channels (reasoning, tool, answer, citation, error) with model reasoning shown in a collapsible panel (#2326)
  • Pluggable orchestration with LangGraph and Agent Framework orchestrators behind one shared async interface (#2326)
  • Registry-based provider model for credentials, LLMs, embedders, parsers, search, and chat history — resolved by key without if/elif dispatch (#2326)
  • Grounded answers with inline citations using a single shared citation contract across orchestrators and storage backends (#2326)
  • Admin experience merged into the frontend for ingesting, inspecting, and configuring the dataset and prompts (#2326)
  • Speech-to-text voice input on supported browsers (#2326)
  • Chat history persisted across sessions in Cosmos DB or PostgreSQL depending on the selected backend (#2326)

Other Updates

  • Migrate Python dependency management from Poetry to uv (#2326)
  • SDK-boundary resilience with structured logging, re-raise, and health checks classifying failures as pass / degraded / fail (#2326)
  • Refresh architecture, deployment, local development, admin, chat history, and managed identity documentation (#2326)
Assets 2
Loading