Usage statistics

Look at this node's schema and the first thing you'll notice is what's missing: no required inputs, no optional inputs beyond a UI widget, and no outputs at all. That's not a broken brief - LF_UsageStatistics genuinely doesn't take or emit anything. It's a pure display node, and it belongs to LF Nodes' Analytics category, which the pack itself describes as tools to "visualize and track data, like checkpoint/LoRA usage or image histograms." Drop it on your canvas and it renders a bar chart via its own KUL_TAB_BAR_CHART widget.

Because there's no wiring involved, the honest answer to "how does it know what to chart" is: we can't tell you from the schema alone, and the pack doesn't document the exact data source at the node level. The most plausible read, going by the category description and how the rest of this pack works, is that it's reading from usage data LF's own selector and configuration nodes (checkpoint pickers, LF_CivitAIMetadataSetup, and similar) log in the background as you use them - meaning this node is more useful the more of the LF ecosystem you're already running, not something you can drop into a checkpoint-loader-only workflow and expect numbers from. Treat that as an educated guess rather than a documented fact.

There's genuinely nothing to configure here beyond letting the node exist on your canvas. The optional ui_widget slot is the chart rendering itself - not an input you set.

Installing it. Same as every node in this pack: ComfyUI Manager, search "LF Nodes," install, restart - or manually cd ComfyUI/custom_nodes && git clone https://github.com/lucafoscili/lf-nodes and restart. Use that URL, not comfyui-lf - that's the pack's old, now-archived home; the author migrated everything to lf-nodes and the archived repo won't get updates. No models or heavy dependencies here either way; this whole pack is Python plus a bundled web-components frontend, nothing GPU-hungry.

Where people get tripped up. The most common confusion is expecting this node to accept an input and chart it directly - it doesn't have one to wire, so if you're hoping to visualize a specific dataset you produced elsewhere in your graph, this isn't that node; look at LF_SaveJSON or the dataset/analytics_dataset outputs other LF nodes expose if you want to inspect raw numbers instead. The second is simpler: if the chart looks empty, it may just be that nothing tracked yet has run in this session - since there's no explicit input, there's also no explicit way to force data into it, so give it a real workflow to observe first.