feat(01-learn): add 21-observability-debugging

[Thamaraimeena]

New tutorial under 01-learn/21-observability-debugging that teaches developers how to observe and debug Strands agents using the SDK's built-in StrandsTelemetry API.

Includes 5 progressive notebooks:

• 01_tracing_setup: Configure StrandsTelemetry and run first traced invocation
• 02_trace_hierarchy: Understand Agent → Cycle → Model → Tool span hierarchy
• 03_debugging_tools: Debug tool failures and detect context window pressure
• 04_backend_export: Export traces to CloudWatch, Langfuse, Jaeger via OTLP
• 05_custom_metrics: Custom span attributes, metrics, and production config

Tested on SageMaker Studio with strands-agents 1.40.0

[Thamaraimeena]

@manoj-selvakumar5 This is ready for review.

[github-actions]

Latest scan for commit: 7f22e47 | Updated: 2026-06-06 19:02:56 UTC

✅ Security Scan Report (PR Files Only)

Scanned Files

• python/01-learn/21-observability-debugging/.gitignore
• python/01-learn/21-observability-debugging/01_tracing_setup.ipynb
• python/01-learn/21-observability-debugging/02_trace_hierarchy.ipynb
• python/01-learn/21-observability-debugging/03_debugging_tools.ipynb
• python/01-learn/21-observability-debugging/04_backend_export.ipynb
• python/01-learn/21-observability-debugging/05_custom_metrics.ipynb
• python/01-learn/21-observability-debugging/README.md
• python/01-learn/21-observability-debugging/requirements.txt
• python/01-learn/21-observability-debugging/trace_utils.py

---

Security Scan Results

Critical	High	Medium	Low	Info
0	0	0	0	0

Threshold: High

No security issues detected in your changes. Great job!

This scan only covers files changed in this PR.

[manoj-selvakumar5]

Hi @Thamaraimeena - Thanks for putting this together. My main feedback is about placement and framing rather than the code. This is a lot for 01-learn. There is a lot of OpenTelemetry engineering likesubclassing SpanExporter, reconstructing span trees, BatchSpanProcessor tuning, sampling ratios. That is good material, but a learn-track module should mostly be teaching the SDK. Here is what I'd suggest, please turn this into a technical use case where you can show some patterns that can be built on top of the spans that get created. Some example ideas like

• Tool reliability SLOs - Aggregate execute_tool spans across traces into per-tool error rate and p95 latency.
• Redaction / compliance pipeline. A custom SpanProcessor that scrubs attributes before export.

Rather than explaining to someone how to look at the spans, I would rather ask: what would you build on top of a trace stream? Please do come up with your own ideas for patterns like these, you are closer to this material than I am. When you build them, Python scripts would work better than notebooks.