FlowTrace Getting Started

FlowTrace is moving from manually-authored test runs into a deterministic mapping and probing system. The operating model is now: start from one entry URL, map the funnel, derive pathways, and probe them with a real browser.

What the operator does

1. Define the funnel entry URL.
2. Turn Protect on.
3. Let the Cartographer discover nodes and edges.
4. Review the generated map and pathways.
5. Run traces and inspect issues by node and pathway.

What the system is trying to do

• discover the full funnel graph from a single URL
• classify nodes and edges
• detect CTA groups and pathway structure
• probe the live funnel in a browser instead of trusting static configuration
• store verdicts in SQL and heavier evidence in blob or lake storage

Current implementation reality

The new Funnel Map, node, edge, pathway, and trace concepts exist. The repo still contains older FlowTrace concepts centered on prompt-authored step plans and run history. That legacy surface is being displaced, but operators should expect mixed UI language while the transition is still underway.

What to look for after setup

• did discovery produce the expected pages
• did CTAs connect the right pages
• are there orphan nodes or missing edges
• do the traced pathways match how a real buyer moves through the funnel
• are issues grouped by node in a way that makes remediation obvious