fix: accept betas body field from Claude Code clients (#360)

Claude Code newer versions send a `betas` list in the request body (e.g.
"interleaved-thinking-2025-05-14"). This landed in `__pydantic_extra__`
and triggered `_openai_reject_native_only_top_level_fields`, surfacing
as "Invalid request sent to provider." for all OpenAI Chat upstreams
(kimi, nvidia_nim).

Declare `betas` on both `MessagesRequest` and `TokenCountRequest` with
`exclude=True` so it is accepted from clients but never forwarded to any
provider body.

api/models/anthropic.py

@@ -153,6 +153,8 @@ class MessagesRequest(BaseModel):
     output_config: dict[str, Any] | None = None
     mcp_servers: list[dict[str, Any]] | None = None
     extra_body: dict[str, Any] | None = None
+    # Beta feature flags sent by Claude Code as a body field; accepted but never forwarded.
+    betas: list[str] | None = Field(default=None, exclude=True)
 
 
 class TokenCountRequest(BaseModel):
@@ -169,3 +171,4 @@ class TokenCountRequest(BaseModel):
     context_management: dict[str, Any] | None = None
     output_config: dict[str, Any] | None = None
     mcp_servers: list[dict[str, Any]] | None = None
+    betas: list[str] | None = Field(default=None, exclude=True)

providers/kimi/request.py

@@ -19,7 +19,7 @@ def build_request_body(request_data: Any, *, thinking_enabled: bool) -> dict:
     try:
         body = build_base_request_body(
             request_data,
-            reasoning_replay=ReasoningReplayMode.DISABLED,
+            reasoning_replay=ReasoningReplayMode.REASONING_CONTENT,
         )
     except OpenAIConversionError as exc:
         raise InvalidRequestError(str(exc)) from exc