Implement Sentry instrumentation for performance and error tracking

[cyrilS-dev]

This Pull Request introduces Sentry integration for performance tracking and error monitoring.

Key changes:

1. FrameProcessor:

• Add conditional Sentry import and initialization check
• Implement custom instrumentation with Sentry spans in FrameProcessorMetrics to measure TTFB (Time To First Byte) and processing time when Sentry is available
• Maintain existing metrics functionality with MetricsFrame regardless of Sentry availability

2. DeepgramSTTService:

   1. Enable general metrics generation:

      • Implement can_generate_metrics method, returning True when VAD is enabled
      • This allows metrics to be collected and used by both Sentry and the metrics system in frame_processor.py

   2. Integrate Sentry-compatible performance tracking:

      • Add start_ttfb_metrics and start_processing_metrics calls in the VAD speech detection handler
      • Implement stop_ttfb_metrics call when receiving transcripts
      • Add stop_processing_metrics for final transcripts

   3. Enhance VAD support for metrics:

      • Add vad_enabled property to check VAD event availability
      • Implement VAD-based speech detection handler for precise metric timing

These changes allow for comprehensive performance monitoring and analysis using Sentry, providing valuable insights into STT, LLM and TTS services. The integration is designed to be non-intrusive, maintaining existing functionality while adding custom performance instrumentation.

Testing:

To validate the Sentry integration, the following steps were taken:

1. Sentry initialization was added to examples/simple-chatbot/demo.py:

sentry_sdk.init(
dsn="SENTRY_DSN",
integrations=[AsyncioIntegration()],
disabled_integrations=[HttpxIntegration()],
traces_sample_rate=1.0,
enable_tracing=True,
)

2. A Sentry transaction was created to encompass the main execution of the chatbot:

with sentry_sdk.start_transaction(op="pipecat", name="Pipecat"):
      await runner.run(task)

3. The vad_events parameter in the DeepgramSTTService configuration was set to True

4. The chatbot was run with this configuration, and the results were observed in the Sentry dashboard.

5. A screenshot of the Sentry performance monitoring results, demonstrating successful tracking of STT, LLM, and TTS services performance metrics :

These tests confirm that Sentry is correctly integrated and capable of monitoring both performance and errors across the pipeline components.

Notes:

It's crucial to initialize Sentry as early as possible in the application, and before from pipecat.pipeline.pipeline import Pipeline :

• To capture any potential errors that may occur during the imports
• To ensure that sentry_sdk.is_initialized() returns the correct state when called in frame_processor.py.

Future Improvements:

1. Usage Metrics in Sentry Spans:
   I plan to enhance the Sentry integration by adding "usage_metrics" to Sentry spans as tags.
   This improvement will provide more detailed and contextual information about resource usage within each service, allowing for better analysis and optimization of our pipeline components.

2. Review TTS Processing Time Measurement:
   I need to review and potentially refine the way we measure processing time for the Text-to-Speech (TTS) service.