feat: add Azure AI Foundry provider support (CPD)

[ricofurtado]

What

Adds Azure AI Foundry as a model provider on the CPD release line, for both LLM and embedding roles.

Ported from azure-ai-rebase-main. The two lines diverged well before this feature — CPD forked roughly 234 main commits back — so this is a rebuilt port rather than a cherry-pick: the Azure-specific changes were kept and adapted to CPD's conventions, while unrelated main-side drift that happened to sit in the same hunks was left out.

Included

• Azure AI Foundry provider config, credential handling, and LiteLLM routing (azure_ai/ prefix, api-version baked into the base URL), plus Azure OpenAI env plumbing.
• POST /models/azure-ai-foundry with lightweight credential validation and optional real-inference testing.
• Settings/onboarding acceptance of Azure providers, provider removal with embedding-in-use conflict handling, and the OPENRAG_AZURE_AI_ENABLED feature flag (default on) gating the UI, endpoints, and credential acceptance.
• Frontend: provider tile, settings dialog/form, onboarding card, logo, and Azure entries in the LLM/embedding selectors.

Adapted to CPD

• Auth uses get_current_user, not main's require_permission — CPD has no providers:write RBAC scope.
• Provider removal clears the model to "", matching how the other providers behave on this branch, rather than main's _default_*_model.
• Kept CPD's plain expressions over main's useMemo/render-phase refactors, and its simpler models-route error handling.

Deliberately excluded: embedding_model_provider

Upstream adds this to the OpenSearch component so the embedding identifier is prefixed with its provider (OpenAI:text-embedding-3-small). That value is what gets written to the embedding_model doc field and the dynamic chunk_embedding_* vector field, so it changes the format of indexed data.

Search resolves models by aggregating embedding_model off the index and matching those values against the embedding object's raw deployment/model/model_id/model_name identifiers, which are never provider-prefixed. Prefixed documents therefore miss the lookup and get skipped, and if every document is prefixed the search raises.

Upstream changes only the write side and has no test coverage for it. Doing this properly needs prefix-aware matching plus a backfill for existing documents (scripts/migrate_embedding_model_field.py), which belongs in its own change. Leaving it out keeps CPD's current behaviour for same-named models across providers unchanged.

Also excluded, as CPD-inapplicable: the Langflow 1.11.2 → 1.11.3 flow re-export and langflowai base-image bump (CPD stays on its own Langflow); the component_index.json / custom_components/ restructure, which does not exist here; and VLM settings, workspace OAuth overrides, RBAC providers:write, the local provider, and main's provider-error sanitizers.

flows/ is untouched by this PR.

Testing

• 545 unit tests pass, up from 530, with the same 16 pre-existing failures as the untouched branch (all in test_settings_refresh_endpoint.py / unrelated suites — present before this change).
• New tests/unit/test_azure_ai_foundry.py covers credential validation, routing, the feature flag, and provider removal.
• Frontend typecheck, lint, and production build clean.
• docker-compose.yml parses.

Manual verification against a live Azure AI Foundry deployment has not been done — worth a pass before merge.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 142deb3c-1f4e-47c9-9695-c6b2653ff7a3

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}