Repository navigation
Maintenance: Mark the Azure AI provider as legacy and point to Custom - #1359
Merged
Merged
Conversation
ArtifactsProduced during runtime |
The provider dialog offers Azure AI under the label "Azure AI (legacy
deployment-based endpoints)" and only accepts deployment URLs. That provider
cannot do semantic search, while the Custom (OpenAI Compatible) provider can,
so the type list now carries the new label and sends Azure admins to Custom
with their resource's v1 endpoint, including the embedding deployment they
need for semantic search. The Custom entry in the type list now spells the
provider name like its dialog label.
Claim -> source anchors:
- Label "Azure AI (legacy deployment-based endpoints)":
app/assets/javascripts/app/lib/app_post/ai_provider/azure_ai.coffee:3
- Type list entry "Custom (OpenAI Compatible)" matches that provider's label:
ai_provider/custom.coffee:3
- No model fields and no model listing, the deployment URL names the model:
azure_ai.coffee:9-11 (model_fields: [], supports_model_listing: false)
- No semantic search: lib/ai/provider/azure.rb:78-79 raises
NotImplementedError from #embeddings, the base class returns false from
#supports_embeddings? (lib/ai/provider.rb:172) and the "Use for semantic
search" action is gated on it
(_ai/provider_connections.coffee:90)
- Custom appends /chat/completions and /embeddings to the configured URL
(lib/ai/provider/custom_open_ai.rb:39, :73), so the documented URL takes no
trailing slash. Its embeddings call sends model: embedding_model!
(custom_open_ai.rb:76), which is why the Embedding Model field is the
remaining step for semantic search.
- Azure v1 API: base URL https://<resource>.openai.azure.com/openai/v1, no
api-version, model is the deployment name, API key accepted in the
authorization header (Microsoft Learn: Azure OpenAI v1 API, and the Foundry
REST reference for POST {endpoint}/openai/v1/embeddings)
- Type list order matches the providers' prio values (zammad_ai 1000 ...
custom 7000)
The legacy label has shipped since 7.0.2 (cherry-picked to stable-7.0), so
this is a maintenance fix for the current docs as well, not a 7.3 feature.
ralf401
force-pushed
the
rsc/maintenance/ai-azure-legacy-provider
branch
from
November 10, 2026 13:49
862ea6e to
50f15b6
Compare
ralf401
changed the title
ralf401
added a commit
that referenced
this pull request
Nov 10, 2026
Backport of the same change made for pre-release (50f15b6, PR #1359). The label has shipped since 7.0.2 (introduced on develop by 281f39ead8, cherry-picked to stable-7.0 as aa2a818162), so the current docs describe a UI that no longer exists: the provider dialog has been offering "Azure AI (legacy deployment-based endpoints)" for the whole 7.0/7.1/7.2 line. The blob of the touched page is identical on main, on pre-release and on the commit the change was written against (db7cbd6), so the port is the same edit with no drift: - Type list entry "Azure AI" -> "Azure AI (legacy deployment-based endpoints)", and "Custom (OpenAI compatible)" -> "Custom (OpenAI Compatible)" to match the dialog label - New paragraph sending Azure admins to Custom with their resource's v1 endpoint, plus the embedding deployment needed for semantic search Claim -> source anchors (zammad, unchanged between 7.2.2 and develop): - ai_provider/azure_ai.coffee:3 (label), :9-11 (model_fields: [], supports_model_listing: false) - ai_provider/custom.coffee:3 (label), :8 (supports_embeddings: true) - lib/ai/provider/azure.rb:78-79 (embeddings raises NotImplementedError), lib/ai/provider.rb:169 in 7.2.x (#supports_embeddings? -> false) - lib/ai/provider/custom_open_ai.rb:39, :73 (appends /chat/completions and /embeddings), :76 (embeddings send model: embedding_model!) The catalog was regenerated on main rather than carried over: its only changes are the strings on this page plus line-number churn.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The provider dialog offers Azure AI under the label "Azure AI (legacy deployment-based endpoints)" and only accepts deployment URLs. That provider cannot do semantic search, while the Custom (OpenAI compatible) provider can, so the type list now carries the new label and sends Azure admins to Custom with their resource's v1 endpoint.
Claim -> source anchors: