Few-Shot Example Selector

Build a dynamic few-shot example selector for improved LLM performance.

## Task Description
{{task_description}}

## Example Pool
{{example_pool_description}}

## Selection Criteria
{{selection_criteria}}

Example: For a code generation task, given the query "Write a function to parse JSON", select the most relevant examples from the pool.

Implement:

```python
class FewShotSelector:
    def select(self, query: str, example_pool: List[Example], k: int) -> List[Example]:
        """
        Selection strategies:
        1. Semantic similarity to query
        2. Diversity among selected examples
        3. Difficulty matching
        4. Domain relevance
        """
        pass
    
    def format_examples(self, examples: List[Example], format_template: str) -> str:
        """Format examples for prompt inclusion"""
        pass
```

Include:
- Embedding-based selection
- MMR for diversity
- Example quality scoring
- Caching for common queries