ai-observability

name	ai-observability
description	Comprehensive guide to monitoring AI/ML systems in production, including Use when this capability is needed.
metadata	{"author":"amnadtaowsoam"}

Ai Observability

Skill Profile

(Select at least one profile to enable specific modules)

• DevOps
• Backend
• Frontend
• AI-RAG
• Security Critical

Overview

Comprehensive guide to monitoring AI/ML systems in production, including LLMs, RAG applications, and traditional ML models. This skill covers monitoring stack setup (Prometheus, Grafana, Jaeger, Phoenix), metrics tracking (latency, throughput, token usage, cost, errors), model performance monitoring, data drift detection, LLM interaction logging, tracing with LangSmith/Phoenix, alerting strategies, dashboards, A/B test monitoring, cost optimization, debugging patterns, and production deployment checklists.

Why This Matters

AI/ML observability is critical because:

• Cost Control: LLM APIs can be expensive; monitoring prevents budget overruns
• Performance: AI systems have unique latency and throughput characteristics
• Quality: Model performance degrades over time due to drift
• Debugging: AI systems are non-deterministic; tracing is essential
• Reliability: AI systems have unique failure modes requiring specialized monitoring

Core Concepts & Rules

1. Core Principles

• Follow established patterns and conventions
• Maintain consistency across codebase
• Document decisions and trade-offs

2. Implementation Guidelines

• Start with the simplest viable solution
• Iterate based on feedback and requirements
• Test thoroughly before deployment

Inputs / Outputs / Contracts

Skill Composition

• Depends on: None
• Compatible with: None
• Conflicts with: None
• Related Skills: None

Quick Start / Implementation Example

1. Review requirements and constraints
2. Set up development environment
3. Implement core functionality following patterns
4. Write tests for critical paths
5. Run tests and fix issues
6. Document any deviations or decisions

# Example implementation following best practices
def example_function():
    # Your implementation here
    pass

Assumptions

• Monitoring stack is accessible via Docker or Kubernetes
• LLM APIs provide token usage information
• Sufficient storage for logs and metrics
• Network connectivity for external observability services

Compatibility

• Python 3.8+
• Prometheus 2.30+
• Grafana 9.0+
• OpenTelemetry 1.0+
• LangChain 0.1+ (for LangSmith tracing)
• Phoenix 2.0+ (for AI-specific observability)

Test Scenario Matrix (QA Strategy)

Type	Focus Area	Required Scenarios / Mocks
Unit	Core Logic	Must cover primary logic and at least 3 edge/error cases. Target minimum 80% coverage
Integration	DB / API	All external API calls or database connections must be mocked during unit tests
E2E	User Journey	Critical user flows to test
Performance	Latency / Load	Benchmark requirements
Security	Vuln / Auth	SAST/DAST or dependency audit
Frontend	UX / A11y	Accessibility checklist (WCAG), Performance Budget (Lighthouse score)

Technical Guardrails & Security Threat Model

1. Security & Privacy (Threat Model)

• Top Threats: Injection attacks, authentication bypass, data exposure

• Data Handling: Sanitize all user inputs to prevent Injection attacks. Never log raw PII
• Secrets Management: No hardcoded API keys. Use Env Vars/Secrets Manager
• Authorization: Validate user permissions before state changes

2. Performance & Resources

• Execution Efficiency: Consider time complexity for algorithms
• Memory Management: Use streams/pagination for large data
• Resource Cleanup: Close DB connections/file handlers in finally blocks

3. Architecture & Scalability

• Design Pattern: Follow SOLID principles, use Dependency Injection
• Modularity: Decouple logic from UI/Frameworks

4. Observability & Reliability

• Logging Standards: Structured JSON, include trace IDs request_id
• Metrics: Track error_rate, latency, queue_depth
• Error Handling: Standardized error codes, no bare except
• Observability Artifacts:
  • Log Fields: timestamp, level, message, request_id
  • Metrics: request_count, error_count, response_time
  • Dashboards/Alerts: High Error Rate > 5%

Agent Directives & Error Recovery

(ข้อกำหนดสำหรับ AI Agent ในการคิดและแก้ปัญหาเมื่อเกิดข้อผิดพลาด)

• Thinking Process: Analyze root cause before fixing. Do not brute-force.
• Fallback Strategy: Stop after 3 failed test attempts. Output root cause and ask for human intervention/clarification.
• Self-Review: Check against Guardrails & Anti-patterns before finalizing.
• Output Constraints: Output ONLY the modified code block. Do not explain unless asked.

Definition of Done (DoD) Checklist

• Tests passed + coverage met
• Lint/Typecheck passed
• Logging/Metrics/Trace implemented
• Security checks passed
• Documentation/Changelog updated
• Accessibility/Performance requirements met (if frontend)

Anti-patterns

1. High Cardinality Metrics: Too many label combinations causing performance issues
2. Silent Monitoring: Monitoring without alerting
3. Over-Monitoring: Tracking everything without focus
4. Ignoring Context: Metrics without dimensional context
5. Alert Fatigue: Too many low-priority alerts
6. No Retention Policy: Unlimited log/metric storage

Reference Links & Examples

• Internal documentation and examples
• Official documentation and best practices
• Community resources and discussions

Versioning & Changelog

• Version: 1.0.0
• Changelog:
  • 2026-02-22: Initial version with complete template structure

---

Converted and distributed by TomeVault — claim your Tome and manage your conversions.