Make LangGraph Agents Durable with Dapr Workflow - Supply Chain Auditor
Duration: 30 minutes
Supported languages: Python. Requires an Anthropic API key.
Dependabot opens a pull request and its changelog says "documentation only" — but a changelog is written by whoever published the release, including an attacker. The classic supply-chain attack hides malicious code inside an update whose notes read as innocent. In this self-paced track you'll run an AI agent that checks a dependency bump's release notes against its real source changes, and see how Dapr Workflow makes that audit durable.
What you'll run
You'll work with the Supply Chain Auditor, a staged LangGraph pipeline — gather_evidence → analyze → render_report — that audits a real Dependabot PR. The gather_evidence node resolves the bumped package to its upstream repo, fetches the release notes and the actual source diff, and runs deterministic red-flag heuristics. Only analyze calls Claude, grounding its verdict in nonce-tagged untrusted evidence that can raise the risk score but never lower the heuristic floor. Every node runs as a checkpointed Dapr Workflow activity, so when you crash the process mid-audit it resumes from durable Redis state — replaying the completed steps instead of re-fetching from GitHub or paying for the Claude call twice. You'll need around 25 minutes to complete the 3 challenges.
Ensure you have your own Anthropic API key before you start this track.
In this self-paced track, you'll learn:
- Why durable execution matters when one node in a pipeline makes an expensive, non-idempotent LLM call.
- How each LangGraph pipeline node becomes a Dapr Workflow activity checkpointed to a Redis state store.
- How a real mid-run crash resumes from durable state instead of re-fetching from GitHub or calling Claude again.
When you've registered to access learning content, click the View fullscreen button in the course player below.
Having trouble loading the course? Open in a new tab