OpenAI Chatbot exposes a chat widget that calls an OpenAI-compatible endpoint and stores each session and message in custom tables. The admin gives a recent-session list and a per-session view of the transcript. That covers spot-checking the widget. It does not cover the operational job of comparing sessions across pages, models or dates, or of bulk-reviewing a week of conversations for quality and prompt regression.

SleekView reads the same sessions and messages tables and presents them as one filterable, sortable table. Each row is a session (or a message, depending on the view) with model, tokens_used, role, entry_url and started_at as real columns. Filter to a single page, sort by token count, jump to the rows that matter without scrolling through a chronological feed.

Save the column set per role: support sees session length and entry URL, ops sees model and tokens, the product owner sees session counts scoped to this month. One dataset, several lenses, all inside WP admin.

OpenAI Chatbot ships with the screen its core job needs: a recent-sessions list and a per-session transcript. That is the right product for debugging. It is the wrong product for running the bot as a measured surface, where the operational questions are which pages drive the longest sessions, which model handles the spend, and how this week compares to last.

A table joining sessions and messages, with model, tokens_used and entry_url as columns, makes those questions answerable in seconds. Capability-gated column visibility keeps message bodies scoped to the right role. Same tables, same plugin, completely different operating posture.