EventFlow Devin Integration

The AI-powered incident response layer for the EventFlow demo. Connects Azure Monitor alerts to the Devin API for automated root cause analysis and fix generation.

Architecture Role

Azure Monitor Alert
        ↓
┌───────────────────┐
│  Azure Function   │  ← alert-function/
│ (webhook handler) │
└────────┬──────────┘
         ↓
┌───────────────────┐
│    Devin API      │  Creates a session with context about all EventFlow repos
│   /v1/sessions    │
└────────┬──────────┘
         ↓
┌───────────────────┐
│   Devin Agent     │  Uses MCP to query Azure Log Analytics
│ (investigation)   │
└────────┬──────────┘
         ↓
┌───────────────────┐
│   Pull Request    │  Fix PR on app_eventflow-payment-service
│  (auto-generated) │
└───────────────────┘

Components

1. Alert Function (alert-function/)

An Azure Function (Python, Consumption plan) that:

• Receives webhook payloads from Azure Monitor alert rules
• Parses the Common Alert Schema to extract error details
• Calls the Devin API to create a new investigation session
• Passes structured context: affected service, error logs, repo URLs

2. MCP Server (mcp-server/)

A Model Context Protocol server that gives Devin access to:

• Azure Log Analytics — query application logs, exceptions, traces
• Azure Monitor Metrics — query error rates, response times, availability
• Provides tools: query_logs, get_exceptions, get_metrics, get_traces

3. Shared Schemas (schemas/)

JSON schemas for the events flowing through the system:

• order_created.json — Event published by Order Service
• payment_processed.json — Event published by Payment Service
• alert_payload.json — Common Alert Schema from Azure Monitor

4. Demo Runbook (docs/)

Step-by-step instructions for running the full demo narrative.

Setup

Prerequisites

• Python 3.11+
• Azure Functions Core Tools v4
• A Devin API key (from https://app.devin.ai/settings)

Alert Function

cd alert-function
pip install -r requirements.txt

# Local testing
func start

# Deploy to Azure
func azure functionapp publish <function-app-name>

MCP Server

cd mcp-server
pip install -r requirements.txt

# Run locally
python server.py

Environment Variables

Alert Function

Variable	Description
DEVIN_API_KEY	Devin API authentication token
DEVIN_API_URL	Devin API base URL (default: https://api.devin.ai/v1)
GITHUB_ORG	GitHub org name (default: Cognition-Partner-Workshops)
REPOS	Comma-separated repo names for Devin context

MCP Server

Variable	Description
AZURE_LOG_ANALYTICS_WORKSPACE_ID	Log Analytics workspace customer ID
AZURE_TENANT_ID	Azure AD tenant ID
AZURE_CLIENT_ID	Service principal client ID
AZURE_CLIENT_SECRET	Service principal secret

Demo Flow

1. Trigger: JPY order submitted to Order Service
2. Crash: Payment Service throws ValueError (zero-decimal currency bug)
3. Detect: Application Insights captures exception, alert rule fires
4. Invoke: Azure Function receives alert webhook, calls Devin API
5. Investigate: Devin connects to Log Analytics via MCP, reads exception details
6. Fix: Devin opens PR on app_eventflow-payment-service with currency fix + test
7. Deploy: CI passes, CD deploys fixed Payment Service
8. Verify: Same JPY order now succeeds