Skip to content

Support

TL;DR: the Azure Agentic AI Solution Accelerator is community-supported, best-effort. Partners own the customer deployment end-to-end. Microsoft does not operate the pager for customer solutions built on this accelerator.

What support looks like

1. Repo-level (what Microsoft provides)

  • GitHub issues on this repo — bug reports, pattern questions, content gaps, scenario requests.
  • Releases + release notes for the accelerator template (flagship code, scenario framework, Bicep modules, accelerator-lint, chat modes).
  • Security vulnerabilities — report privately per SECURITY.md.
  • Office hours (if your partner org is onboarded) — periodic sync with the accelerator engineering team.

What Microsoft does NOT provide via this repo: - Production support for customer deployments. - 24×7 incident response. - Legal / contractual support commitments. - Customer-specific architecture review.

2. Engagement-level (what the partner provides)

Partners own: - Sev-1 response + customer comms. - Runbook execution. - Waiver tracking within the customer engagement. - Baseline upgrades on customer schedules. - Cost + RAI posture monitoring.

This accelerator gives partners the tools + patterns to do those jobs well. It does not replace them.

3. Microsoft field (when applicable)

Depending on your partner's commercial relationship with Microsoft, a field Solutions Architect may be engaged per-customer for co-delivery on lighthouse or strategic engagements. That is outside the scope of this repo; coordinate through your Microsoft partner manager.

How to file a useful issue

Include:

  1. What layer is the issue on?
  2. Content (pattern doc / guidance)
  3. Tooling (validator, baseline CLI, azd template)
  4. Example (reference scenario, example Spec)
  5. What version? — release tag, or commit SHA.
  6. Minimal repro — Spec excerpt, CI log tail, validator output.
  7. What you expected vs. what happened.
  8. Severity from your perspective — blocking a production deploy / blocking dev / nice-to-have.

Good issues, with clear reproduction steps, get triaged faster. Vague or off-topic issues may be closed without action.

What's actually shipped today

  • Flagship scenario (Sales Research & Outreach) is fully runnable under src/scenarios/sales_research/; loaded dynamically from accelerator.yaml -> scenario: at startup.
  • Scenario framework: src/workflow/registry.py + src/workflow/base.py. scripts/scaffold-scenario.py materializes new scenarios.
  • Infra: infra/ deploys Foundry (GA API) + default content filter + model deployment + Search + KV + ACA + App Insights via azd up.
  • Evals: evals/quality/ + evals/redteam/ run against the deployed endpoint; thresholds in accelerator.yaml -> acceptance.
  • Lint: scripts/accelerator-lint.py — ~30 deterministic AST-only checks; CI-gated.
  • Reference scenarios: Customer Service Actioning and RFP Response ship as walkthrough READMEs under docs/references/ (runnable code lands when a customer engagement motivates it).

See docs/getting-started/setup-and-prereqs.md for the exact onboarding path.

Security

Security vulnerabilities go to MSRC via SECURITY.md, not GitHub issues.

Escalation paths

Situation Where to go
Pattern bug or content error GitHub issue on this repo
Accelerator lint false positive / false negative GitHub issue with repro
Customer production incident Partner's own on-call; accelerator engineering is not on the pager
Security vulnerability in the accelerator MSRC per SECURITY.md
Need a new reference scenario or pattern variant GitHub issue + discussion
Disagree with a pattern recommendation PR with rationale; accelerator engineering reviews