Idea: graphify inspect to show semantic extraction inputs

While trying Graphify on a real-world project, I ran into an issue that took quite a while to diagnose.

My project initially produced:

found 2176 code, 94 docs, 26 papers, 360 images
error: no LLM API key found (480 doc/paper/image file(s) need semantic extraction).

I wanted to generate an AST-only graph, so I started removing directories and files until Graphify reported:

found 1324 code, 2 docs, 0 papers, 0 images

At that point I still had no idea which two files Graphify considered "docs". It took a fair amount of manual investigation to find the remaining semantic inputs.

Proposal

Add a new command:

graphify inspect

that lists the files Graphify classifies as semantic inputs before extraction.

Example:

Semantic Inputs
Documents (2)
 .cursor/rules/graphify.mdc
 README.md
Papers (1)
 docs/architecture.pdf
Images (3)
 docs/system.png
 docs/ui.png
 docs/logo.png

Optionally:

graphify inspect --json

for scripting and CI.

Error message improvement

When no LLM backend is configured, the current error could become:

No LLM API key found.
2 semantic files require semantic extraction.
Run:
 graphify inspect
to list the files requiring semantic extraction.

This would make debugging much easier without changing Graphify's existing behavior.

If this direction makes sense, I'd be happy to implement it and open a PR.