Feat: BYOM API flavor constants and factory resolution (#61)

cosminacho · claude · web-flow · commit 69d9cf36b67c · 2026-04-16T17:43:09.000+03:00
Co-authored-by: Claude Opus 4.6 (1M context) &lt;noreply@anthropic.com&gt;
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,6 +2,14 @@
 
 All notable changes to `uipath_llm_client` (core package) will be documented in this file.
 
+## [1.8.3] - 2026-04-16
+
+### Added
+- BYOM API flavor constants for discovery endpoint: `OpenAiChatCompletions`, `OpenAiResponses`, `OpenAiEmbeddings`, `GeminiGenerateContent`, `GeminiEmbeddings`, `AwsBedrockInvoke`, `AwsBedrockConverse`
+- `BYOM_TO_ROUTING_FLAVOR` mapping to resolve BYOM discovery flavors to routing-level API flavors
+- Extended `API_FLAVOR_TO_VENDOR_TYPE` with BYOM flavor entries for automatic vendor resolution
+- LiteLLM client now resolves BYOM discovery flavors to correct routing flavors and litellm providers
+
 ## [1.8.2] - 2026-04-13
 
 ### Fixed
diff --git a/packages/uipath_langchain_client/CHANGELOG.md b/packages/uipath_langchain_client/CHANGELOG.md
@@ -2,6 +2,11 @@
 
 All notable changes to `uipath_langchain_client` will be documented in this file.
 
+## [1.8.3] - 2026-04-16
+
+### Added
+- Factory functions (`get_chat_model`, `get_embedding_model`) now automatically resolve BYOM discovery API flavors to the correct client and routing flavor
+
 ## [1.8.2] - 2026-04-13
 
 ### Changed
diff --git a/packages/uipath_langchain_client/pyproject.toml b/packages/uipath_langchain_client/pyproject.toml
@@ -6,7 +6,7 @@ readme = "README.md"
 requires-python = ">=3.11"
 dependencies = [
     "langchain>=1.2.15",
-    "uipath-llm-client>=1.8.2",
+    "uipath-llm-client>=1.8.3",
 ]
 
 [project.optional-dependencies]
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/__version__.py b/packages/uipath_langchain_client/src/uipath_langchain_client/__version__.py
@@ -1,3 +1,3 @@
 __title__ = "UiPath LangChain Client"
 __description__ = "A Python client for interacting with UiPath's LLM services via LangChain."
-__version__ = "1.8.2"
+__version__ = "1.8.3"
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/clients/azure/chat_models.py b/packages/uipath_langchain_client/src/uipath_langchain_client/clients/azure/chat_models.py
@@ -6,6 +6,7 @@
 from uipath_langchain_client.base_client import UiPathBaseChatModel
 from uipath_langchain_client.clients.openai.utils import fix_url_and_api_flavor_header
 from uipath_langchain_client.settings import (
+    ApiFlavor,
     ApiType,
     RoutingMode,
     UiPathAPIConfig,
@@ -31,6 +32,7 @@ class UiPathAzureAIChatCompletionsModel(UiPathBaseChatModel, AzureAIOpenAIApiCha
         vendor_type=VendorType.AZURE,
         freeze_base_url=False,
     )
+    api_flavor: ApiFlavor | str | None = None
 
     # Override fields to avoid env var lookup / validation errors at instantiation
     endpoint: str | None = Field(default="PLACEHOLDER")
@@ -40,13 +42,16 @@ class UiPathAzureAIChatCompletionsModel(UiPathBaseChatModel, AzureAIOpenAIApiCha
 
     @model_validator(mode="after")
     def setup_uipath_client(self) -> Self:
+        if self.api_flavor is not None:
+            self.api_config.api_flavor = self.api_flavor
         base_url = str(self.uipath_sync_client.base_url).rstrip("/")
+        locked_flavor = str(self.api_config.api_flavor) if self.api_config.api_flavor else None
 
         def on_request(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         async def on_request_async(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         self.uipath_sync_client.event_hooks["request"].append(on_request)
         self.uipath_async_client.event_hooks["request"].append(on_request_async)
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/clients/openai/chat_models.py b/packages/uipath_langchain_client/src/uipath_langchain_client/clients/openai/chat_models.py
@@ -7,6 +7,7 @@
 from uipath_langchain_client.base_client import UiPathBaseChatModel
 from uipath_langchain_client.clients.openai.utils import fix_url_and_api_flavor_header
 from uipath_langchain_client.settings import (
+    ApiFlavor,
     ApiType,
     RoutingMode,
     UiPathAPIConfig,
@@ -31,6 +32,7 @@ class UiPathChatOpenAI(UiPathBaseChatModel, ChatOpenAI):  # type: ignore[overrid
         api_version="2025-03-01-preview",
         freeze_base_url=False,
     )
+    api_flavor: ApiFlavor | str | None = None
 
     # Override fields to avoid errors when instantiating the class
     openai_api_key: SecretStr | None | Callable[[], str] | Callable[[], Awaitable[str]] = Field(
@@ -39,13 +41,24 @@ class UiPathChatOpenAI(UiPathBaseChatModel, ChatOpenAI):  # type: ignore[overrid
 
     @model_validator(mode="after")
     def setup_uipath_client(self) -> Self:
+        if self.api_flavor is not None:
+            self.api_config.api_flavor = self.api_flavor
+            # Lock LangChain's use_responses_api to match the discovered flavor.
+            # Without this, features like reasoning={} or certain model names
+            # silently switch LangChain to the Responses API, which would fail
+            # if the model only supports chat-completions (or vice versa).
+            if self.api_flavor == ApiFlavor.CHAT_COMPLETIONS:
+                self.use_responses_api = False
+            elif self.api_flavor == ApiFlavor.RESPONSES:
+                self.use_responses_api = True
         base_url = str(self.uipath_sync_client.base_url).rstrip("/")
+        locked_flavor = str(self.api_config.api_flavor) if self.api_config.api_flavor else None
 
         def on_request(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         async def on_request_async(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         self.uipath_sync_client.event_hooks["request"].append(on_request)
         self.uipath_async_client.event_hooks["request"].append(on_request_async)
@@ -75,6 +88,7 @@ class UiPathAzureChatOpenAI(UiPathBaseChatModel, AzureChatOpenAI):  # type: igno
         api_version="2025-03-01-preview",
         freeze_base_url=False,
     )
+    api_flavor: ApiFlavor | str | None = None
 
     # Override fields to avoid errors when instantiating the class
     azure_endpoint: str | None = Field(default="PLACEHOLDER")
@@ -83,13 +97,20 @@ class UiPathAzureChatOpenAI(UiPathBaseChatModel, AzureChatOpenAI):  # type: igno
 
     @model_validator(mode="after")
     def setup_uipath_client(self) -> Self:
+        if self.api_flavor is not None:
+            self.api_config.api_flavor = self.api_flavor
+            if self.api_flavor == ApiFlavor.CHAT_COMPLETIONS:
+                self.use_responses_api = False
+            elif self.api_flavor == ApiFlavor.RESPONSES:
+                self.use_responses_api = True
         base_url = str(self.uipath_sync_client.base_url).rstrip("/")
+        locked_flavor = str(self.api_config.api_flavor) if self.api_config.api_flavor else None
 
         def on_request(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         async def on_request_async(request: Request) -> None:
-            fix_url_and_api_flavor_header(base_url, request)
+            fix_url_and_api_flavor_header(base_url, request, api_flavor=locked_flavor)
 
         self.uipath_sync_client.event_hooks["request"].append(on_request)
         self.uipath_async_client.event_hooks["request"].append(on_request_async)
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/clients/openai/utils.py b/packages/uipath_langchain_client/src/uipath_langchain_client/clients/openai/utils.py
@@ -5,22 +5,28 @@
 from uipath_langchain_client.settings import ApiFlavor
 
 
-def fix_url_and_api_flavor_header(base_url: str, request: Request) -> None:
-    """Detect API flavor from URL suffix and rewrite the URL to the base gateway URL.
+def fix_url_and_api_flavor_header(
+    base_url: str, request: Request, *, api_flavor: str | None = None
+) -> None:
+    """Set the API flavor header and rewrite the URL to the base gateway URL.
 
-    Inspects the outgoing request URL to determine whether it targets the
-    OpenAI *responses* or *chat completions* endpoint and sets the
-    ``X-UiPath-LlmGateway-ApiFlavor`` header accordingly.  The request URL
-    is then collapsed back to *base_url* so that the gateway receives a
-    clean path.
+    When *api_flavor* is provided (e.g. from the discovery endpoint), it is
+    used directly — the model only supports that specific flavor.  Otherwise
+    the flavor is inferred from the outgoing URL suffix (``/responses`` vs
+    ``/chat/completions``).
 
     Args:
         base_url: The UiPath gateway base URL to rewrite the request to.
         request: The outgoing httpx request (mutated in place).
+        api_flavor: Locked API flavor from discovery. When set, overrides
+            dynamic detection from the URL path.
     """
-    url_suffix = str(request.url).split(base_url)[-1]
-    if "responses" in url_suffix:
-        request.headers["X-UiPath-LlmGateway-ApiFlavor"] = ApiFlavor.RESPONSES.value
+    if api_flavor is not None:
+        request.headers["X-UiPath-LlmGateway-ApiFlavor"] = api_flavor
     else:
-        request.headers["X-UiPath-LlmGateway-ApiFlavor"] = ApiFlavor.CHAT_COMPLETIONS.value
+        url_suffix = str(request.url).split(base_url)[-1]
+        if "responses" in url_suffix:
+            request.headers["X-UiPath-LlmGateway-ApiFlavor"] = ApiFlavor.RESPONSES.value
+        else:
+            request.headers["X-UiPath-LlmGateway-ApiFlavor"] = ApiFlavor.CHAT_COMPLETIONS.value
     request.url = URL(base_url)
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/factory.py b/packages/uipath_langchain_client/src/uipath_langchain_client/factory.py
@@ -28,6 +28,7 @@
 )
 from uipath_langchain_client.settings import (
     API_FLAVOR_TO_VENDOR_TYPE,
+    BYOM_TO_ROUTING_FLAVOR,
     ApiFlavor,
     RoutingMode,
     UiPathBaseSettings,
@@ -160,11 +161,16 @@ def get_chat_model(
         raise ValueError("No vendor type or api flavor found in model info")
     discovered_vendor_type = discovered_vendor_type.lower()
 
+    # Discovered 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
+        else:
+            api_flavor = discovered_api_flavor
+
     match discovered_vendor_type:
         case VendorType.OPENAI:
-            if api_flavor == ApiFlavor.RESPONSES:
-                model_kwargs["use_responses_api"] = True
-
             if is_uipath_owned:
                 from uipath_langchain_client.clients.openai.chat_models import (
                     UiPathAzureChatOpenAI,
@@ -173,6 +179,7 @@ def get_chat_model(
                 return UiPathAzureChatOpenAI(
                     model=model_name,
                     settings=client_settings,
+                    api_flavor=api_flavor,
                     byo_connection_id=byo_connection_id,
                     **model_kwargs,
                 )
@@ -184,6 +191,7 @@ def get_chat_model(
                 return UiPathChatOpenAI(
                     model=model_name,
                     settings=client_settings,
+                    api_flavor=api_flavor,
                     byo_connection_id=byo_connection_id,
                     **model_kwargs,
                 )
@@ -323,10 +331,9 @@ 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}'. "
diff --git a/packages/uipath_langchain_client/src/uipath_langchain_client/settings.py b/packages/uipath_langchain_client/src/uipath_langchain_client/settings.py
@@ -24,8 +24,10 @@
 )
 from uipath.llm_client.settings.constants import (
     API_FLAVOR_TO_VENDOR_TYPE,
+    BYOM_TO_ROUTING_FLAVOR,
     ApiFlavor,
     ApiType,
+    ByomApiFlavor,
     RoutingMode,
     VendorType,
 )
@@ -39,6 +41,8 @@
     "ApiType",
     "RoutingMode",
     "ApiFlavor",
+    "ByomApiFlavor",
     "VendorType",
     "API_FLAVOR_TO_VENDOR_TYPE",
+    "BYOM_TO_ROUTING_FLAVOR",
 ]
diff --git a/src/uipath/llm_client/__version__.py b/src/uipath/llm_client/__version__.py
@@ -1,3 +1,3 @@
 __title__ = "UiPath LLM Client"
 __description__ = "A Python client for interacting with UiPath's LLM services."
-__version__ = "1.8.2"
+__version__ = "1.8.3"
diff --git a/src/uipath/llm_client/clients/litellm/client.py b/src/uipath/llm_client/clients/litellm/client.py
@@ -27,8 +27,10 @@
 from uipath.llm_client.settings.base import UiPathAPIConfig
 from uipath.llm_client.settings.constants import (
     API_FLAVOR_TO_VENDOR_TYPE,
+    BYOM_TO_ROUTING_FLAVOR,
     ApiFlavor,
     ApiType,
+    ByomApiFlavor,
     RoutingMode,
     VendorType,
 )
@@ -87,6 +89,14 @@
     "converse": "bedrock",
     "invoke": "bedrock",
     "anthropic-claude": "vertex_ai",
+    # BYOM discovery flavors
+    ByomApiFlavor.OPENAI_CHAT_COMPLETIONS: "openai",
+    ByomApiFlavor.OPENAI_RESPONSES: "openai",
+    ByomApiFlavor.OPENAI_EMBEDDINGS: "openai",
+    ByomApiFlavor.GEMINI_GENERATE_CONTENT: "gemini",
+    ByomApiFlavor.GEMINI_EMBEDDINGS: "gemini",
+    ByomApiFlavor.AWS_BEDROCK_INVOKE: "bedrock",
+    ByomApiFlavor.AWS_BEDROCK_CONVERSE: "bedrock",
 }
 
 _ANTHROPIC_FAMILY = "anthropicclaude"
@@ -210,7 +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
+
+        # 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"):
diff --git a/src/uipath/llm_client/clients/openai/utils.py b/src/uipath/llm_client/clients/openai/utils.py
@@ -11,10 +11,12 @@ def __init__(
         model_name: str,
         client_settings: UiPathBaseSettings,
         byo_connection_id: str | None = None,
+        api_flavor: ApiFlavor | str | None = None,
     ):
         self.model_name = model_name
         self.client_settings = client_settings
         self.byo_connection_id = byo_connection_id
+        self.locked_api_flavor = str(api_flavor) if api_flavor else None
         self.base_api_config = UiPathAPIConfig(
             routing_mode=RoutingMode.PASSTHROUGH,
             vendor_type=VendorType.OPENAI,
@@ -37,13 +39,15 @@ def _apply_routing(self, request: Request, api_config: UiPathAPIConfig) -> None:
 
     def fix_url_and_headers(self, request: Request):
         if request.url.path.endswith("/completions"):
+            flavor = self.locked_api_flavor or ApiFlavor.CHAT_COMPLETIONS
             api_config = self.base_api_config.model_copy(
-                update={"api_flavor": ApiFlavor.CHAT_COMPLETIONS, "api_type": ApiType.COMPLETIONS}
+                update={"api_flavor": flavor, "api_type": ApiType.COMPLETIONS}
             )
             self._apply_routing(request, api_config)
         elif request.url.path.endswith("/responses"):
+            flavor = self.locked_api_flavor or ApiFlavor.RESPONSES
             api_config = self.base_api_config.model_copy(
-                update={"api_flavor": ApiFlavor.RESPONSES, "api_type": ApiType.COMPLETIONS}
+                update={"api_flavor": flavor, "api_type": ApiType.COMPLETIONS}
             )
             self._apply_routing(request, api_config)
         elif request.url.path.endswith("/embeddings"):
diff --git a/src/uipath/llm_client/settings/__init__.py b/src/uipath/llm_client/settings/__init__.py
@@ -30,7 +30,13 @@
 from typing import Literal
 
 from uipath.llm_client.settings.base import UiPathAPIConfig, UiPathBaseSettings
-from uipath.llm_client.settings.constants import ApiFlavor, ApiType, RoutingMode, VendorType
+from uipath.llm_client.settings.constants import (
+    ApiFlavor,
+    ApiType,
+    ByomApiFlavor,
+    RoutingMode,
+    VendorType,
+)
 from uipath.llm_client.settings.llmgateway import LLMGatewaySettings
 from uipath.llm_client.settings.platform import PlatformSettings
 
@@ -96,4 +102,5 @@ def get_default_client_settings(
     "RoutingMode",
     "VendorType",
     "ApiFlavor",
+    "ByomApiFlavor",
 ]
diff --git a/src/uipath/llm_client/settings/constants.py b/src/uipath/llm_client/settings/constants.py

Original file line number	Diff line number	Diff line change
`@@ -6,7 +6,7 @@ readme = "README.md"`
`6`	`6`	`requires-python = ">=3.11"`
`7`	`7`	`dependencies = [`
`8`	`8`	`"langchain>=1.2.15",`
`9`		`- "uipath-llm-client>=1.8.2",`
	`9`	`+ "uipath-llm-client>=1.8.3",`
`10`	`10`	`]`
`11`	`11`
`12`	`12`	`[project.optional-dependencies]`
Original file line number	Diff line number	Diff line change
`@@ -24,8 +24,10 @@`
`24`	`24`	`)`
`25`	`25`	`from uipath.llm_client.settings.constants import (`
`26`	`26`	`API_FLAVOR_TO_VENDOR_TYPE,`
	`27`	`+ BYOM_TO_ROUTING_FLAVOR,`
`27`	`28`	`ApiFlavor,`
`28`	`29`	`ApiType,`
	`30`	`+ ByomApiFlavor,`
`29`	`31`	`RoutingMode,`
`30`	`32`	`VendorType,`
`31`	`33`	`)`
`@@ -39,6 +41,8 @@`
`39`	`41`	`"ApiType",`
`40`	`42`	`"RoutingMode",`
`41`	`43`	`"ApiFlavor",`
	`44`	`+ "ByomApiFlavor",`
`42`	`45`	`"VendorType",`
`43`	`46`	`"API_FLAVOR_TO_VENDOR_TYPE",`
	`47`	`+ "BYOM_TO_ROUTING_FLAVOR",`
`44`	`48`	`]`