Reduce entity lookup context and configure reasoning effort

config/llm.example.toml

@@ -17,6 +17,8 @@ timeout_seconds = 30
 response_format = "json_object"
 token_parameter = "max_tokens"
 max_output_tokens = 4096
+# Optional; omit for provider defaults. Supported values depend on the model.
+# reasoning_effort = "medium"
 
 [[llm.providers]]
 id = "own-server"