last fix

Author: msdavidram

learn-pr/wwl/analyze-monitor-tune-ai-powered-business-solutions/includes/6-interpret-telemetry-data-performance-model-tuning.md

@@ -10,47 +10,47 @@ Telemetry provides data about how the system behaves in real time. It is essenti
 
 #### Operational Telemetry
 
-Latency and throughput
+* Latency and throughput
 
-Error rates and failure modes
+* Error rates and failure modes
 
-Resource consumption and throttling
+* Resource consumption and throttling
 
 #### ModelLevel Telemetry
 
-Token usage and cost patterns
+* Token usage and cost patterns
 
-Response consistency
+* Response consistency
 
-Drift indicators and degradation trends
+* Drift indicators and degradation trends
 
 #### Behavioral Telemetry
 
-User satisfaction and completion rates
+* User satisfaction and completion rates
 
-Prompt patterns and abandonment rates
+* Prompt patterns and abandonment rates
 
-Model alignment to intended tasks
+* Model alignment to intended tasks
 
 #### Governance and Compliance Signals
 
-Guardrail interventions
+* Guardrail interventions
 
-Blocked actions or restricted data access
+* Blocked actions or restricted data access
 
-Policy or sensitivity label conflicts
+* Policy or sensitivity label conflicts
 
 ## 2. Performance Signals and Interpretation
 
 Solution architects should focus on **patterns**, not isolated events.
 
 ### Performance Indicators
 
-**Increased latency**<br>Indicates heavy workloads, inefficient prompt structures, or connector delays.
+* **Increased latency**<br>Indicates heavy workloads, inefficient prompt structures, or connector delays.
 
-**Spikes in error rates**<br>Often point to broken integrations, incorrect environment configuration, or model instability.
+* **Spikes in error rates**<br>Often point to broken integrations, incorrect environment configuration, or model instability.
 
-**High token usage**<br>Suggests verbose outputs, unclear prompts, or an overly complex workflow.
+* **High token usage**<br>Suggests verbose outputs, unclear prompts, or an overly complex workflow.
 
 ### Performance Signal Map 
 
@@ -68,31 +68,31 @@ Model tuning focuses on improving the quality and reliability of responses.
 
 ### 3.1 Tuning Opportunities
 
-**Prompt Refinement**<br>Improving instructions, constraints, and expectations for predictable results.
+* **Prompt Refinement**<br>Improving instructions, constraints, and expectations for predictable results.
 
-**Knowledge Updates**<br>Adding, removing, or restructuring knowledge sources for better grounding.
+* **Knowledge Updates**<br>Adding, removing, or restructuring knowledge sources for better grounding.
 
-**Behavioral Adjustments**<br>Introducing fallback logic, clarifying actions, or refining orchestration flow.
+* **Behavioral Adjustments**<br>Introducing fallback logic, clarifying actions, or refining orchestration flow.
 
-**Cost Optimization**<br>Reducing unnecessary token usage and optimizing invocation structure.
+* **Cost Optimization**<br>Reducing unnecessary token usage and optimizing invocation structure.
 
 ## 4. TelemetryDriven Diagnosis Workflow
 
 A consistent workflow helps isolate issues quickly.
 
 ### StepbyStep Diagnostic Flow
 
-**Monitor Key Metrics**<br>Gather baseline information across latency, throughput, quality, and satisfaction.
+* **Monitor Key Metrics**<br>Gather baseline information across latency, throughput, quality, and satisfaction.
 
-**Identify Anomalies**<br>Look for deviations from expected patterns.
+* **Identify Anomalies**<br>Look for deviations from expected patterns.
 
-**Correlate Related Signals**<br>Combine user behavior, failures, and performance metrics.
+* **Correlate Related Signals**<br>Combine user behavior, failures, and performance metrics.
 
-**Determine Root Cause**<br>Validate if the issue is modelbased, integrationbased, or promptbased.
+* **Determine Root Cause**<br>Validate if the issue is modelbased, integrationbased, or promptbased.
 
-**Apply Targeted Tuning**<br>Update prompts, improve workloads, adjust knowledge, or change configuration.
+* **Apply Targeted Tuning**<br>Update prompts, improve workloads, adjust knowledge, or change configuration.
 
-**Validate Improvements**<br>Compare beforeandafter telemetry patterns to ensure successful tuning.
+* **Validate Improvements**<br>Compare beforeandafter telemetry patterns to ensure successful tuning.
 
 ## 5. Architecture Flow of Telemetry Analysis
 
@@ -103,15 +103,15 @@ A consistent workflow helps isolate issues quickly.
 
 Solution architects should define clear KPIs aligned to business goals:
 
-**Responsiveness**<br>Median response time within acceptable limits.
+* **Responsiveness**<br>Median response time within acceptable limits.
 
-**Accuracy & Relevance**<br>Model outputs aligned to task expectations.
+* **Accuracy & Relevance**<br>Model outputs aligned to task expectations.
 
-**Reliability**<br>Low failure rate across workflows.
+* **Reliability**<br>Low failure rate across workflows.
 
-**CostEffectiveness**<br>Balanced token usage and model selection.
+* **CostEffectiveness**<br>Balanced token usage and model selection.
 
-**User Outcome Completion**<br>Ability for users to complete tasks without manual intervention.
+* **User Outcome Completion**<br>Ability for users to complete tasks without manual intervention.
 
 ## References
 

learn-pr/wwl/analyze-monitor-tune-ai-powered-business-solutions/includes/8-module-summary.md

@@ -6,61 +6,61 @@ The module emphasized data driven decision making, structured analysis methods,
 
 ### 1. Monitoring Is Foundational to Reliable AI Operations
 
-Effective monitoring spans multiple layers: operational health, performance, quality, usage, and risk.
+* Effective monitoring spans multiple layers: operational health, performance, quality, usage, and risk.
 
-A strong monitoring operating model includes defined roles, standardized metrics, log review cadence, and clear expectations for agent behavior.
+* A strong monitoring operating model includes defined roles, standardized metrics, log review cadence, and clear expectations for agent behavior.
 
-Guardrails and threshold based alerts help detect anomalies, prevent misuse, and maintain governance alignment.
+* Guardrails and threshold based alerts help detect anomalies, prevent misuse, and maintain governance alignment.
 
 ### 2. Enterprise Tools Enable Deep Observability
 
-Solution architects rely on telemetry tools, analytics dashboards, and observability platforms to gain insight into agent performance.
+* Solution architects rely on telemetry tools, analytics dashboards, and observability platforms to gain insight into agent performance.
 
-Monitoring solutions help track errors, latency, usage trends, connector performance, and model behavior.
+* Monitoring solutions help track errors, latency, usage trends, connector performance, and model behavior.
 
-Centralized dashboards improve visibility across complex AI environments.
+* Centralized dashboards improve visibility across complex AI environments.
 
 ### 3. Backlogs and User Feedback Drive Continuous Improvements
 
-Backlogs capture enhancements, recurring issues, feature gaps, and governance concerns.
+* Backlogs capture enhancements, recurring issues, feature gaps, and governance concerns.
 
-Categorizing backlog items into domains like accuracy, knowledge, performance, UX, integration, and compliance enables structured prioritization.
+* Categorizing backlog items into domains like accuracy, knowledge, performance, UX, integration, and compliance enables structured prioritization.
 
-User feedback—volume, severity, sentiment, and friction points—provides critical signals for improvement planning.
+* User feedback—volume, severity, sentiment, and friction points—provides critical signals for improvement planning.
 
-Conversation transcripts reveal root causes that aren't always visible in raw telemetry.
+* Conversation transcripts reveal root causes that aren't always visible in raw telemetry.
 
 ### 4. Telemetry and Diagnostics Reveal Root Causes
 
-Key telemetry signals include latency spikes, error patterns, token consumption, drift patterns, and guardrail triggers.
+* Key telemetry signals include latency spikes, error patterns, token consumption, drift patterns, and guardrail triggers.
 
-A consistent diagnostic workflow helps isolate issues: monitor → detect anomalies → correlate patterns → identify root causes → tune → validate.
+* A consistent diagnostic workflow helps isolate issues: monitor → detect anomalies → correlate patterns → identify root causes → tune → validate.
 
 ### 5. Tuning Improves Accuracy, Performance, and User Experience
 
-Knowledge tuning: update, restructure, or remove outdated content.
+* Knowledge tuning: update, restructure, or remove outdated content.
 
-Behavioral tuning: clarify orchestration steps, refine prompts, and introduce fallback logic.
+* Behavioral tuning: clarify orchestration steps, refine prompts, and introduce fallback logic.
 
-Performance tuning: optimize workflows, reduce unnecessary actions, fix connector, or dependency bottlenecks.
+* Performance tuning: optimize workflows, reduce unnecessary actions, fix connector, or dependency bottlenecks.
 
-Governance aligned tuning ensures that DLP, permissions, and sensitivity rules remain intact.
+* Governance aligned tuning ensures that DLP, permissions, and sensitivity rules remain intact.
 
 ### 6. Performance and User Centered Metrics Track Effectiveness
 
-Core operational metrics: latency, throughput, error rate, and resource utilization.
+* Core operational metrics: latency, throughput, error rate, and resource utilization.
 
-Quality metrics: accuracy, knowledge coverage, and action effectiveness.
+* Quality metrics: accuracy, knowledge coverage, and action effectiveness.
 
-User centered metrics: satisfaction, abandonment rate, and task completion rate (the best indicator of user success).
+* User centered metrics: satisfaction, abandonment rate, and task completion rate (the best indicator of user success).
 
-Monitoring model and drift indicators ensure models remain reliable over time.
+* Monitoring model and drift indicators ensure models remain reliable over time.
 
 ### 7. Continuous Improvement Is an Ongoing Lifecycle
 
-AI agents require an iterative approach: monitor → analyze → tune → validate → release → monitor again.
+* AI agents require an iterative approach: monitor → analyze → tune → validate → release → monitor again.
 
-Regular updates to knowledge content, workflows, guardrails, and orchestration refine agent behavior.
+* Regular updates to knowledge content, workflows, guardrails, and orchestration refine agent behavior.
 
 
-Insights from telemetry and transcripts feed into improvement cycles and stakeholder communication.
+* Insights from telemetry and transcripts feed into improvement cycles and stakeholder communication.