LLM Output Parser Builder

Build robust output parsers for structured LLM responses.

## Expected Output Format
{{output_format}}

## Common Failure Modes
{{failure_modes}}

## Recovery Requirements
{{recovery_requirements}}

Implement a resilient parsing system:

```python
class OutputParser:
    def parse(self, raw_output: str) -> ParsedOutput:
        """
        1. Attempt strict parsing
        2. Apply fuzzy matching for near-misses
        3. Use LLM-based repair for broken outputs
        4. Validate against schema
        5. Return structured result or detailed error
        """
        pass
    
    def repair(self, broken_output: str, error: ParseError) -> str:
        """Use LLM to fix malformed output"""
        pass
```

Include:
- Regex and JSON parsing strategies
- Schema validation with Pydantic
- Retry logic with modified prompts
- Metrics for parse success rates