fix(docs): pin Fern VLM model slugs

[lbliii]

Summary

• pin the five VLM page slugs that Fern currently derives differently from their authored model-coverage links
• apply the same explicit slugs to nightly, latest, and v0.4 navigation
• redirect the previous auto-generated versioned and unversioned URLs to preserve existing bookmarks
• document the Fern site-root link requirement, GitHub source-view limitation, explicit-slug criteria, and version/redirect checklist in AGENTS.md, the Fern README, and the contributor skill

Root cause

The Fern migration converted working Sphinx file links into version-agnostic URL paths based on the source filenames. The corresponding navigation entries did not receive explicit slug: values, so Fern instead generated routes from display names such as InternVL and LLaVA-OneVision. Those generated routes (intern-vl, l-la-va-one-vision, and similar) did not match the authored links (internvl, llava-onevision), causing the unresolved links to render without a version prefix and return 404.

Impact

The VLM overview links for Ministral3 VL, InternVL, SmolVLM, LLaVA, and LLaVA-OneVision resolve within the reader's selected documentation version. Existing links to Fern's previously generated routes continue to work through redirects.

Future contributors and agents are explicitly warned not to replace Fern inter-page routes with GitHub-friendly file-relative links, and to keep linked slugs aligned across every applicable version navigation file.

This addresses the VLM model-coverage subset reported in NVBug 6179245.

Validation

• make -C docs/fern docs-check
  • stitched the frozen v0.4 pages from docs-archive
  • validated 290 MDX files
  • fern check: 0 errors (the optional historical-redirect check warned because the remote Fern registry was unavailable)
• targeted consistency check confirming each authored VLM link has the same explicit slug in nightly, latest, and v0.4 navigation
• git diff --check

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-lbliii-fix-fern-vlm-slugs.docs.buildwithfern.com/nemo/automodel