Fix: discovered BYOM api_flavor takes precedence and applies to embeddings

- Discovered BYOM api_flavor now overrides user-supplied api_flavor in both
  the LangChain chat factory and the LiteLLM client
- get_embedding_model now reads discovered_api_flavor consistently (not just
  as a vendor fallback when vendor is None)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: cosminacho

packages/uipath_langchain_client/src/uipath_langchain_client/factory.py

@@ -161,9 +161,11 @@ def get_chat_model(
         raise ValueError("No vendor type or api flavor found in model info")
     discovered_vendor_type = discovered_vendor_type.lower()
 
-    # For BYOM models, derive routing api_flavor from the discovered BYOM flavor
-    if api_flavor is None and discovered_api_flavor is not None:
-        api_flavor = BYOM_TO_ROUTING_FLAVOR.get(discovered_api_flavor)
+    # Discovered BYOM api_flavor takes precedence over user-supplied api_flavor
+    if discovered_api_flavor is not None:
+        routing_flavor = BYOM_TO_ROUTING_FLAVOR.get(discovered_api_flavor)
+        if routing_flavor is not None:
+            api_flavor = routing_flavor
 
     match discovered_vendor_type:
         case VendorType.OPENAI:
@@ -328,10 +330,9 @@ def get_embedding_model(
         )
 
     discovered_vendor_type = model_info.get("vendor")
-    if discovered_vendor_type is None:
-        discovered_api_flavor = model_info.get("apiFlavor")
-        if discovered_api_flavor is not None:
-            discovered_vendor_type = API_FLAVOR_TO_VENDOR_TYPE.get(discovered_api_flavor)
+    discovered_api_flavor = model_info.get("apiFlavor")
+    if discovered_vendor_type is None and discovered_api_flavor is not None:
+        discovered_vendor_type = API_FLAVOR_TO_VENDOR_TYPE.get(discovered_api_flavor)
     if discovered_vendor_type is None:
         raise ValueError(
             f"No vendor type found in model info for embedding model '{model_name}'. "

src/uipath/llm_client/clients/litellm/client.py

@@ -220,11 +220,17 @@ def _discover_and_build_api_config(
             raise ValueError(f"Cannot determine vendor for model '{self._model_name}'")
 
         resolved_vendor = str(vendor_type or discovered_vendor).lower()
-        resolved_flavor = str(api_flavor) if api_flavor is not None else discovered_flavor
 
-        # Resolve BYOM discovery flavors to routing-level flavors
-        if resolved_flavor is not None:
-            resolved_flavor = BYOM_TO_ROUTING_FLAVOR.get(resolved_flavor, resolved_flavor)
+        # Discovered BYOM api_flavor takes precedence over user-supplied api_flavor
+        routing_flavor = (
+            BYOM_TO_ROUTING_FLAVOR.get(discovered_flavor) if discovered_flavor is not None else None
+        )
+        if routing_flavor is not None:
+            resolved_flavor: str | None = routing_flavor
+        elif api_flavor is not None:
+            resolved_flavor = str(api_flavor)
+        else:
+            resolved_flavor = discovered_flavor
 
         # OpenAI defaults to chat-completions when no flavor is discovered
         if resolved_flavor is None and resolved_vendor in ("openai", "azure"):