Distributed Tracing Setup

Set up distributed tracing for our ML inference pipeline.

## Pipeline Components
{{pipeline_components}}

## Tracing Tool
{{tracing_tool}}

## Key Metrics to Capture
{{key_metrics}}

Implement comprehensive tracing:

```python
# Tracing decorator for pipeline stages
@trace_stage("embedding_generation")
async def generate_embedding(text: str) -> List[float]:
    pass

# Context propagation
class TraceContext:
    # Propagate trace ID across services
    pass

# Custom attributes for ML-specific data
class MLSpanAttributes:
    # Token counts, model versions, cache hits
    pass
```

Include:
- Sampling strategies for high-volume systems
- Cost-effective retention policies
- Alert rules based on trace data
- Dashboard configurations