Automatically syncs Microsoft Azure and M365 roadmap updates from official RSS feeds into Azure DevOps Epics. An AI agent generates structured work items — including impact summaries and recommended actions — and routes them to the correct ADO boards based on product.
Note: This repo uses real org/project names (
hobbitfeetado/Hobbit-Dev) as working examples. These are identifiers only — no credentials are stored here. To adapt for your own org, updateroadmap-sync-config.json,configure-logic-app.sh, andsetup-ado-boards.shwith your ADO organisation URL and project name.
Logic Apps (weekday 07:00 UTC)
└─▶ Azure AI Foundry Agent (GPT-4o)
├─▶ fetch_roadmap Function — fetches RSS, filters, resolves board routing
└─▶ ado_operations Function — deduplicates and creates Epics in ADO
└─▶ Azure DevOps (multiple boards)
- Logic Apps triggers the agent on a weekday schedule
- The agent calls
fetch_roadmapto get filtered roadmap items with board routing resolved - For each item, the agent checks ADO for an existing
RoadmapId:<guid>tag (duplicate detection) - New items get a generated Epic with an Impact Summary, Recommended Actions, and the original description
- The agent reports a summary of what was created, skipped, and any errors
| File | Purpose |
|---|---|
roadmap-sync-config.json |
Product-to-board mappings, RSS filter settings (feeds, globalFilters, boardMappings). Edit this to change which products sync to which boards. |
.env.example |
Template for local credentials — copy to .env and fill in before running any scripts. Never committed. |
| File | Purpose |
|---|---|
agent-instructions.md |
Foundry Agent system prompt. Defines the sync workflow, work item template (title, HTML description, tags), and routing rules. Deployed via create-foundry-agent.py. |
create-foundry-agent.py |
Creates or updates the Foundry Agent in Azure AI Foundry. Registers fetch_roadmap and ado_operations as OpenAPI tools and pushes the latest instructions. |
| File | Purpose |
|---|---|
functions/function_app.py |
Python Azure Functions entry point. Contains two HTTP-triggered functions: fetch_roadmap (fetches RSS feeds, filters by date/status/product, resolves board routing) and ado_operations (searches for duplicate work items by tag and creates new Epics in ADO). |
functions/requirements.txt |
Python dependencies for the function app. |
functions/host.json |
Azure Functions host configuration. |
| File | Purpose |
|---|---|
deploy-azure-resources.sh |
Provisions all Azure resources (Resource Group, Storage, App Insights, Function App, Azure OpenAI with GPT-4o, Logic App scaffold) and deploys the function code. Idempotent — safe to re-run. |
setup-ado-boards.sh |
Creates ADO teams and area paths from boardMappings in the config. Updates roadmap-sync-config.json with the resolved areaPath values. |
configure-logic-app.sh |
Deploys the full Logic App workflow definition: recurrence trigger, agent thread/run creation, poll-until-complete loop, and final message retrieval. Re-run after any config change. |
| File | Purpose |
|---|---|
SETUP.md |
Step-by-step provisioning guide covering all six setup stages. |
SOP.md |
Operator runbook — common tasks like changing the sync window, adding boards, rotating the ADO PAT, and diagnosing failures. |
playground-test-prompts.md |
Structured test prompts for the Azure AI Foundry Playground to validate each component before enabling the schedule. |
prompts.txt |
Example invocation prompts for the GitHub Copilot interactive agent. |
| File | Purpose |
|---|---|
.github/agents/roadmap-sync.agent.md |
Copilot agent definition for on-demand interactive triage sessions — an alternative to the automated Logic App flow. |
Items are routed to ADO boards based on product. All boards live in a single ADO project under a Roadmap area path hierarchy.
| Board | Products |
|---|---|
| M365 Collaboration | Teams, SharePoint, OneDrive, Outlook, Exchange, Planner, Viva, OneNote, PowerPoint |
| Security and Compliance | Microsoft Purview, Defender for Office 365, Information Protection |
| Identity and Access | Microsoft Entra |
| Endpoint Management | Microsoft Intune, Windows 365 |
| M365 Platform | Microsoft 365, M365 admin center, Microsoft Copilot (M365) |
| General | Fallback for all other products |
For a complete walkthrough, see SETUP.md. It covers every provisioning step in detail — including Azure resource configuration, Foundry Hub and Project setup, agent tool registration, Logic App deployment, and end-to-end verification. The steps below are a condensed overview.
- Azure subscription with Contributor access
- Azure DevOps project
- Azure CLI (
brew install azure-cli) - jq (
brew install jq) - Python 3.10+
Edit roadmap-sync-config.json to set your ADO organization and adjust product-to-board groupings if needed.
chmod +x setup-ado-boards.sh && ./setup-ado-boards.shCreates a team and area path per board mapping. Updates roadmap-sync-config.json with resolved area paths.
Add your credentials to .env (copy from .env.example), then:
chmod +x deploy-azure-resources.sh && ./deploy-azure-resources.shProvisions: Resource Group, Storage Account, Application Insights, Function App, Azure OpenAI (GPT-4o), Logic App scaffold.
Create manually at ai.azure.com — link to the resource group and Azure OpenAI resource created in step 3. See SETUP.md §3 for details.
pip3 install azure-ai-projects azure-ai-agents azure-identity python-dotenv
python3 create-foundry-agent.pyRegisters the agent with both function tools. Save the printed Agent ID for the Logic App.
Open la-roadmap-sync in the Azure Portal and configure the recurrence trigger (Mon–Fri, 07:00 UTC) and the Foundry Agent action. See SETUP.md §4.
Use the prompts in playground-test-prompts.md to validate each component before enabling the schedule.
Each Epic created in ADO includes:
- Title — cleaned roadmap item title (status prefix removed, max 120 chars)
- Impact Summary — AI-generated description of what the change means for an IT organisation
- Recommended Actions — status-dependent guidance (monitor / pilot / adopt)
- Details table — source feed, status, published date, products, link to Microsoft roadmap
- Tags —
Roadmap,RoadmapId:<guid>, feed name, status, product names,Needs-Review(if In preview)
roadmap-sync-config.json controls filtering and routing:
Test the function locally before deploying:
cd functions
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
func startcurl -X POST http://localhost:7071/api/fetch_roadmap \
-H "Content-Type: application/json" \
-d '{"config": <paste roadmap-sync-config.json>, "daysBack": 14}'Under $5/month. The main costs are Azure OpenAI token usage ($1–3/month at 50K tokens/day) and Logic Apps consumption ($0.30/month). The Function App runs on a Flex Consumption plan and is effectively free at this invocation volume.
{ "feeds": ["azure", "m365"], // which RSS feeds to fetch "globalFilters": { "statuses": ["In preview", "In development"], "excludeTypes": ["Retirements"], "daysBack": 7 }, "boardMappings": [ // ordered — first match wins { "name": "M365 Collaboration", "products": ["Microsoft Teams", "SharePoint"], "ado": { "organization": "https://dev.azure.com/yourorg/", "project": "Your-Project", "workItemType": "Epic", "areaPath": "Your-Project\\Area\\Roadmap\\M365 Collaboration" } } ], "defaultBoard": { ... } // fallback for unmatched products }