Skip to content

Latest commit

 

History

History
69 lines (53 loc) · 1.85 KB

ROADMAP.md

File metadata and controls

69 lines (53 loc) · 1.85 KB

msgspec_ext SDK Roadmap

Planned features for future releases.

v1.1.0 - Sampling and Performance

Trace Sampling

Add support for trace sampling to reduce overhead in high-volume production environments.

Feature: MSGTRACE_TRACE_SAMPLE_RATE

Implementation:

# Environment variable
MSGTRACE_TRACE_SAMPLE_RATE=0.1  # Sample 10% of traces

# In tracer.py, add to _initialize_tracer():
from opentelemetry.sdk.trace.sampling import TraceIdRatioBased

sample_rate = float(os.getenv("MSGTRACE_TRACE_SAMPLE_RATE", "1.0"))
sampler = TraceIdRatioBased(sample_rate)
provider = TracerProvider(resource=resource, sampler=sampler)

Use cases:

  • Production environments with high traffic (sample 1-10%)
  • Development/staging with full sampling (100%)
  • Cost optimization by reducing trace volume
  • Performance optimization by reducing overhead

Configuration options:

  • 1.0 (default) - Capture 100% of traces
  • 0.5 - Capture 50% of traces
  • 0.1 - Capture 10% of traces
  • 0.0 - Disable tracing (alternative to MSGTRACE_TELEMETRY_ENABLED)

Technical details:

  • Uses OpenTelemetry's TraceIdRatioBased sampler
  • Deterministic sampling based on trace ID
  • Sampling decision propagated to child spans
  • No performance impact on non-sampled traces

Future Considerations

Batch Export Configuration

Fine-tune batch processor settings:

  • MSGTRACE_BATCH_MAX_QUEUE_SIZE
  • MSGTRACE_BATCH_SCHEDULE_DELAY
  • MSGTRACE_BATCH_MAX_EXPORT_BATCH_SIZE

Custom Samplers

Support for custom sampling strategies:

  • Parent-based sampling
  • Rate limiting sampler
  • Attribute-based sampling

Metrics Support

Add OpenTelemetry metrics alongside traces:

  • Token usage metrics
  • Cost metrics
  • Latency histograms
  • Error rates

Log Correlation

Automatic correlation between logs and traces:

  • Inject trace context into logs
  • Link log records to spans