Pull Request

Summary

This PR enables users to set custom inference IDs when using the @trace decorator, addressing the need for request correlation and integration with external systems. Previously, inference IDs were always auto-generated UUIDs, making it difficult to correlate traces with external request identifiers.

Changes

• Add dedicated inference_id field to Trace class for cleaner architecture
• Enable custom inference ID via update_current_trace(inferenceId="custom_id")
• Maintain backward compatibility with auto-generated UUID fallback
• Simplify post_process_trace logic by removing special metadata filtering
• Update examples to demonstrate custom inference ID usage patterns
• Add comprehensive test coverage for both custom and auto-generated IDs

Context

Users need the ability to set custom inference IDs for several critical use cases:

• Request Correlation: Link traces with external API request IDs
• Batch Processing: Group related requests with correlation patterns
• System Integration: Use existing request identifiers from upstream systems
• Audit Trails: Maintain consistent identifiers across distributed systems

The previous implementation stored inferenceId in trace metadata, which was semantically incorrect and required complex filtering logic. This PR refactors the architecture to use a dedicated field while maintaining full backward compatibility.

Closes: OPEN-7312

Testing

• Unit tests: Created comprehensive tests for custom and auto-generated inference IDs
• Manual testing: Verified functionality with live Openlayer API using provided credentials
• Integration testing: Tested request correlation patterns and batch processing scenarios
• Backward compatibility: Verified existing code continues to work unchanged

Usage Examples

Basic Custom Inference ID

@trace()
def process_request(user_id: str):
    custom_id = f"req_{user_id}_{int(time.time())}"
    update_current_trace(inferenceId=custom_id)
    return "processed"

Request Correlation

@trace()
def batch_process():
    batch_id = f"batch_{uuid.uuid4().hex[:8]}"
    update_current_trace(inferenceId=batch_id)
    # Process related items...

External System Integration

@trace()
def handle_api_request(external_request_id: str):
    update_current_trace(inferenceId=external_request_id)
    # Use existing request ID from external system

Breaking Changes

None - This change is fully backward compatible. Existing code will continue to work exactly as before, using auto-generated UUIDs when no custom inference ID is provided.