A guided pipeline that classifies, plans, executes, and verifies every case.
Pick the best case model from the prompt and domain signal.
Initialize the case and load all step templates.
Generate a plan, map dependencies, and stage required fields.
Skip completed steps, respect stop targets, and gate the run.
Run ready steps in parallel while enforcing dependency order.
Run processors, normalize values, and enrich field output.
Generate requests, trigger actions, and update fields.
Detect kill switches, validate completion, and surface gaps.
Persist outputs into steps and User Tables for future runs.