Refactor: use model_fields_set to exclude unset UiPathChat params

Replace the _UNSET sentinel approach with pydantic's built-in
model_fields_set. Fields keep plain `int | None = None` defaults;
_default_params includes only fields that were explicitly set by
the caller. Explicit None passes through to the API as null.

No sentinel class, no type: ignore, no extra model_config needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: cosminacho

packages/uipath_langchain_client/CHANGELOG.md

@@ -5,7 +5,7 @@ All notable changes to `uipath_langchain_client` will be documented in this file
 ## [1.9.7] - 2026-04-22
 
 ### Changed
-- **Behavior change:** `UiPathChat` parameter fields now default to an internal `_UNSET` sentinel instead of `None`, and `_default_params` filters on the sentinel rather than on `None`. Unset fields are still omitted from the request payload, but passing an explicit `None` (e.g. `UiPathChat(temperature=None)`) now forwards `null` to the normalized API instead of being silently dropped. Previously, both "not passed" and "explicitly `None`" produced the same omission. Callers relying on `None` as shorthand for "omit" should switch to simply not passing the argument.
+- **Behavior change:** `UiPathChat._default_params` now uses pydantic's `model_fields_set` to decide which params to include in the request payload instead of filtering on `v is not None`. Fields that were not explicitly passed are omitted; fields explicitly set to `None` (e.g. `UiPathChat(temperature=None)`) now forward `null` to the API. Previously both "not passed" and "explicitly `None`" were silently dropped.
 
 ## [1.9.6] - 2026-04-22
 

packages/uipath_langchain_client/src/uipath_langchain_client/clients/normalized/chat_models.py

@@ -62,7 +62,7 @@
     convert_to_openai_tool,
 )
 from langchain_core.utils.pydantic import is_basemodel_subclass
-from pydantic import AliasChoices, BaseModel, ConfigDict, Field
+from pydantic import AliasChoices, BaseModel, Field
 
 from uipath_langchain_client.base_client import UiPathBaseChatModel
 from uipath_langchain_client.settings import ApiType, RoutingMode, UiPathAPIConfig
@@ -72,10 +72,6 @@
 _DictOrPydantic = Union[dict[str, Any], BaseModel]
 
 
-# Sentinel — params whose value is still _UNSET are omitted from the request payload.
-_UNSET: Any = object()
-
-
 def _oai_structured_outputs_parser(ai_msg: AIMessage, schema: type[BaseModel]) -> BaseModel:
     if not ai_msg.content:
         raise ValueError("Expected non-empty content from model.")
@@ -162,50 +158,48 @@ class UiPathChat(UiPathBaseChatModel):
         freeze_base_url=True,
     )
 
-    model_config = ConfigDict(validate_default=False)
-
     # Common
     max_tokens: int | None = Field(
-        default=_UNSET,
+        default=None,
         validation_alias=AliasChoices("max_tokens", "max_output_tokens", "max_completion_tokens"),
     )
-    temperature: float | None = _UNSET  # type: ignore[assignment]
-    top_p: float | None = _UNSET  # type: ignore[assignment]
-    top_k: int | None = _UNSET  # type: ignore[assignment]
+    temperature: float | None = None
+    top_p: float | None = None
+    top_k: int | None = None
     stop: list[str] | str | None = Field(
-        default=_UNSET,
+        default=None,
         validation_alias=AliasChoices("stop", "stop_sequences"),
     )
     n: int | None = Field(
-        default=_UNSET,
+        default=None,
         validation_alias=AliasChoices("n", "candidate_count"),
     )
-    frequency_penalty: float | None = _UNSET  # type: ignore[assignment]
-    presence_penalty: float | None = _UNSET  # type: ignore[assignment]
-    seed: int | None = _UNSET  # type: ignore[assignment]
+    frequency_penalty: float | None = None
+    presence_penalty: float | None = None
+    seed: int | None = None
 
     model_kwargs: dict[str, Any] = Field(default_factory=dict)
     disabled_params: dict[str, Any] | None = None
 
     # OpenAI
-    logit_bias: dict[str, int] | None = _UNSET  # type: ignore[assignment]
-    logprobs: bool | None = _UNSET  # type: ignore[assignment]
-    top_logprobs: int | None = _UNSET  # type: ignore[assignment]
-    parallel_tool_calls: bool | None = _UNSET  # type: ignore[assignment]
-    reasoning_effort: str | None = _UNSET  # type: ignore[assignment]
-    reasoning: dict[str, Any] | None = _UNSET  # type: ignore[assignment]
+    logit_bias: dict[str, int] | None = None
+    logprobs: bool | None = None
+    top_logprobs: int | None = None
+    parallel_tool_calls: bool | None = None
+    reasoning_effort: str | None = None
+    reasoning: dict[str, Any] | None = None
 
     # Anthropic
-    thinking: dict[str, Any] | None = _UNSET  # type: ignore[assignment]
+    thinking: dict[str, Any] | None = None
 
     # Google
-    thinking_level: str | None = _UNSET  # type: ignore[assignment]
-    thinking_budget: int | None = _UNSET  # type: ignore[assignment]
-    include_thoughts: bool | None = _UNSET  # type: ignore[assignment]
-    safety_settings: list[dict[str, Any]] | None = _UNSET  # type: ignore[assignment]
+    thinking_level: str | None = None
+    thinking_budget: int | None = None
+    include_thoughts: bool | None = None
+    safety_settings: list[dict[str, Any]] | None = None
 
     # Shared
-    verbosity: str | None = _UNSET  # type: ignore[assignment]
+    verbosity: str | None = None
 
     @property
     def _llm_type(self) -> str:
@@ -248,8 +242,9 @@ def _default_params(self) -> dict[str, Any]:
             "verbosity": self.verbosity,
         }
 
+        set_fields = self.model_fields_set
         return {
-            **{k: v for k, v in candidates.items() if v is not _UNSET},
+            **{k: v for k, v in candidates.items() if k in set_fields},
             **self.model_kwargs,
         }