Process a query using DeepSeek's R1 reasoning engine and prepare it for integration with Claude.
DeepSeek R1 leverages advanced reasoning capabilities that naturally evolved from large-scale
reinforcement learning, enabling sophisticated reasoning behaviors. The output is enclosed
within `<ant_thinking>` tags to align with Claude's thought processing framework.
Args:
query (dict): Contains the following keys:
- context (str): Optional background information for the query.
- question (str): The specific question to be analyzed.
Returns:
str: The reasoning output from DeepSeek, formatted with `<ant_thinking>` tags for seamless use with Claude.