LangGraph Resilient Agent

The purpose

Treats approval, interrupted-run recovery and side-effect uniqueness as testable application invariants.

What the project exposes

LangGraph checkpoints, idempotency keys, uniqueness constraints, tenant isolation and audit events without sensitive payloads.

How to verify it

Ruff, mypy and pytest cover safe, sensitive, resumed and rejected paths.

Provenance

The public murillo-consulting repository is the source for this presentation. Its history, license and files are the reference for understanding the exact provenance of contributions and dependencies.

Limits to understand

Header identity and SQLite are local adapters that must be replaced for production.

This demonstrator is neither a production validation for your system nor evidence of a client engagement. Adoption requires a review appropriate to your environment.