Introduction¶
Welcome to the CI/CD for Non-Agentic Workloads learning path. In this module, you will add observability to a traditional Azure provisioning pipeline — without modifying any application code — using Monocle's zero-code YAML configuration. When the pipeline fails, an AI-powered SRE agent automatically analyzes the traces and provides root cause analysis.
What you will learn¶
- How to instrument any Python application with zero code changes using
okahu.yaml - How Monocle captures function inputs, outputs, and errors as trace spans
- How to build a CI/CD workflow that triggers AI-powered root cause analysis on failure
- How the Kahu SRE Agent uses traces to diagnose deployment failures
Scenario¶
You are running a 4-step Azure provisioning pipeline that deploys Blob Storage, SQL Database, Kusto tables, and user accounts. Step 4 fails. Instead of manually digging through logs, you want the CI/CD pipeline to automatically send traces to an AI agent that diagnoses the root cause and posts the analysis to a GitHub issue.
Architecture overview¶
graph TD
GHA[GitHub Actions] --> Monocle[Monocle Zero-Code Runner]
Monocle --> App[deploy_app.py]
App --> S1[Step 1: Azure Blob]
App --> S2[Step 2: Azure SQL]
App --> S3[Step 3: Kusto Tables]
App --> S4[Step 4: User Accounts ❌]
Monocle --> File[.monocle/ traces]
Monocle --> Cloud[Okahu Cloud]
S4 -->|Failure| GHA
GHA --> Query[Query Okahu for Traces]
Query --> SRE[Kahu SRE Agent]
SRE --> Issue[GitHub Issue with RCA]
Source code¶
okahu-demos/cicd_azure_provisioning
Time to complete¶
~30 minutes