Mindwalk setup: map code and review AI sessions

This Portuguese-language walkthrough shows how Mindwalk turns a codebase and AI agent session logs into a navigable 3D map. The presenter describes a Go binary with a local server and browser interface, and says Go must be installed. The map records file activity so users can inspect what an agent viewed, read or edited. Its colors indicate the deepest kind of interaction with a file rather than the number of visits.

The setup demonstration uses an existing e-commerce project on Windows. The presenter adapts the repository's Linux-oriented commands for PowerShell and mentions WSL as another installation option. He then asks the coding assistant to add an editable free-shipping threshold and follows its activity through the map, file inspection and session timeline. Although he expects edits to appear live, some displayed numbers require a reload during the demonstration.

A second request addresses inconsistent shipping totals by sharing the calculation between the cart, checkout and saved order. The session evaluation returns a Portuguese report. It identifies repeated cart edits and notes that the second round ran lint and tests without another type check. The presenter uses these findings to discuss planning and verification, while stressing that developers should test the generated code themselves. This is a walkthrough of locally hosted session analysis; it does not demonstrate running a local LLM.