Skip to main content
The Scenario Builder lets you describe how a conversation should go as a path graph. Each path through the graph becomes a Digital Human you can run against your agent. Use it when you want fine‑grained control over what your Digital Humans say, in what order, and on which branches.

​
What You’ll Learn

  • What a scenario is and how it produces Digital Humans
  • How to enumerate paths and assign counts per path
  • How scenarios connect to simulations

​
How Scenarios Work

A scenario is a graph of nodes (utterances, branches, decisions) connected by edges. Starting from the Start node, the builder enumerates every distinct path through the graph up to a per‑scenario cap. Each path becomes a single Digital Human whose enriched_playback follows that path turn by turn. In the example above the builder enumerates two paths from Start (Refill and Booking). Selecting this scenario in a simulation produces two Digital Humans, one per path.

​
Key Capabilities

  • Deterministic playback: each generated Digital Human follows its assigned path turn by turn
  • Branching coverage: every reachable branch yields its own Digital Human, up to the per‑scenario cap
  • Pasted JSON or LLM draft: author scenarios visually, paste a graph, or generate from a prompt
  • Reuse across simulations: one scenario can drive many simulation runs

​
Common Use Cases

  • Capture every meaningful branch of a workflow (happy path + edge cases) as one scenario
  • Reproduce specific incidents from production transcripts as a deterministic path
  • Build a regression suite of conversation paths an agent must consistently handle

​
Next Steps

Scenario Builder Cookbook

Practical examples and patterns for building scenarios.

Create Scenario API

Create and manage scenarios programmatically.