Repository navigation
EN index
Follow the actual rerun record for the latest execution evidence. Paid infrastructure, private ingestion and Foundry/MCP subsystems have now been exercised and cleaned up. The backend continuation also cloned the actual saved agent and PDF through official PAC. These partial results do not certify the complete connected-agent exercises.
Use these guides alongside your authorized local workshop material. Do not treat a successful documentation check as a successful cloud deployment. The last review date is 2026-10-20; each tenant can expose different previews and regional capabilities.
| Lab | Guide | Evidence level |
|---|---|---|
| 0 | Prerequisites, licensing, environment, and permissions | Preflight and backend access; standard runtime incomplete |
| 1 | Licensing advisor, knowledge, and email tools | Core live path passed; cloned PDF hash matched; citation audit incomplete |
| 2 | Azure AI Search as knowledge | Private ingestion passed; licensing/Copilot integration incomplete |
| 3 | Application Insights monitoring | Synthetic ingestion/KQL passed; agent-originated telemetry incomplete |
| 4 | Bring your own model with a prompt | Direct paid model call passed; connected prompt incomplete |
| 5.1 | Deploy the advanced-search infrastructure | Corrected paid deployment and verified cleanup completed |
| 5 | Advanced querying with Azure AI Search | Real retrieval checks passed; connected Copilot answer incomplete |
| 6 | Foundry IQ and agentic retrieval | MCP/single-source passed; exact multisource quotations failed; Activity conversation not run |
Run 5.1 before 5 if the advanced-search resources do not already exist. Lab 4 depends on the Search index from Lab 2. The supplied monitoring document has an internal title that calls it Lab 4, and the BYOM document calls itself Lab 3. These guides follow the source filenames: monitoring is Lab 3 and BYOM is Lab 4.
The current Copilot Studio documentation distinguishes several agent harnesses. The original workshop's Topics, knowledge nodes, and prompt tools refer to the standard harness. A new agent in the newer experience is not automatically an equivalent replacement. Read platform changes first.
In particular:
- Standard-harness generative orchestration and the GitHub Copilot harness are different concepts.
- Current licensing uses Copilot Credits, not the original message terminology.
- A successful maker license assignment does not prove environment capacity or model quota.
- Azure AI Search feature availability depends on tier, region, model, identity, and index schema.
- File knowledge does not support encrypted or password-protected files.
- BYOM prompt integration does not replace the agent's orchestration model.
- Validation ledger and acceptance gates
- Screenshot provenance and evidence gallery
- Troubleshooting, safe substitutions, and cleanup
- Publication and repeatable wiki export
For each lab, start a fresh test conversation, record the exact input and output, capture the visible activity/tool result, and retain the relevant downstream evidence. Do not send email to a third party or create cloud resources without the lab owner's approval. These guides use a controlled test mailbox and a dedicated resource group.
Source screenshots are not new executions. Les captures d'origine ne sont pas de nouvelles exécutions. See validation / voir le registre.