Headless Conversational Engine Development in .NetCore

I need a headless, .NET Core–based finite-state-machine that can run an entire conversational flow for user onboarding and easily adapt to other scenarios later. The solution must persist each state in SQL while storing the raw conversational payload as JSON so we can replay or analyze it at any time.

Functionality you must wire in from day one includes user authentication, real-time updates (SignalR or equivalent), and comprehensive error logging with clear reporting hooks.

A JSON-driven rule engine will determine every transition, guard, and side effect. To balance power and ease of use I want a hybrid authoring model: non-technical staff can tweak parameters through a Razor Pages GUI, while developers may still drop full JSON definitions straight into source control.

Key technical pieces
– .NET Core FSM engine (Stateless or a library you recommend)
– SQL storage + JSON payload archiving
– Razor Pages front-end for admin, live logs, and visual debugging
– C# domain-mapping layer that keeps DTOs and domain objects cleanly separated

Deliverables
• Web API endpoints (e.g., /converse, /admin) compiled and running in .NET Core
• based on Db First Approach - So that we can integrate this module in existing project.
• Razor Pages admin portal with single Super User Login guarded with oAuth2.0, live log stream, and workflow editor
• Sample onboarding workflow defined completely in JSON, executing end-to-end
• Unit/integration tests covering persistence, real-time events, and failure paths
• Docker-compose for single-command local spin-up

Show me how you would structure the FSM, what libraries or patterns you favor, and a rough timeline so we can kick this off quickly.
If you have read it all, start your proposal with "FSM"