Use OpenRouter reasoning object for gpt-5.5

Probing the OR endpoint directly: when tools are sent alongside the bare
reasoning_effort param, OpenRouter silently drops reasoning. Sending
reasoning as an object (reasoning: {effort: ...}) preserves both. Set
openrouter_reasoning_object=True so users picking a thinking level
actually get reasoning back when their workflow uses tools.

https://claude.ai/code/session_01WSZjtD8K2X51BBJgaq2c5M

Author: claude

libs/core/kiln_ai/adapters/ml_model_list.py

@@ -557,6 +557,10 @@ class KilnModel(BaseModel):
                 structured_output_mode=StructuredOutputMode.json_schema,
                 available_thinking_levels=GPT_5_4_OPENAI_THINKING_LEVELS,
                 default_thinking_level="none",
+                # Use OpenRouter's reasoning object so reasoning is preserved
+                # when tools are sent (the bare reasoning_effort param is
+                # silently dropped on tool calls for openai/gpt-5.5).
+                openrouter_reasoning_object=True,
                 suggested_for_evals=True,
                 suggested_for_data_gen=True,
                 supports_doc_extraction=True,